ISSN 1451 - 9372(Print) ISSN 2217 - 7434(Online ... No1_p1-104...čevi, Novi Sad, Serbia (45 20’ N, 19 51’ E). The content of water in grain was determined according to the method

ISSN 1451 - 9372(Print)ISSN 2217 - 7434(Online)JANUARY-MARCH 2020Vol.26, Number 1, 1-104

www.ache.org.rs/ciceq

Journal of the Association of Chemical Engineers of Serbia, Belgrade, Serbia

EDITOR-In-Chief Vlada B. Veljković

Faculty of Technology, University of Niš, Leskovac, Serbia E-mail: [email protected]

ASSOCIATE EDITORS Jonjaua Ranogajec

Faculty of Technology, University of Novi Sad, Novi Sad, Serbia

Srđan PejanovićDepartment of Chemical Engineering, Faculty of Technology and Metallurgy,

University of Belgrade, Belgrade, Serbia

Milan Jakšić ICEHT/FORTH, University of Patras,

Patras, Greece

EDITORIAL BOARD (Serbia) Đorđe Janaćković, Sanja Podunavac-Kuzmanović, Viktor Nedović, Sandra Konstantinović, Ivanka Popović

Siniša Dodić, Zoran Todorović, Olivera Stamenković, Marija Tasić, Jelena Avramović, Goran Nikolić, Dunja Sokolović

ADVISORY BOARD (International) Dragomir Bukur

Texas A&M University, College Station, TX, USA

Milorad Dudukovic Washington University,

St. Luis, MO, USA Jiri Hanika

Institute of Chemical Process Fundamentals, Academy of Sciences of the Czech Republic, Prague, Czech Republic

Maria Jose Cocero University of Valladolid,

Valladolid, Spain Tajalli Keshavarz

University of Westminster, London, UK Zeljko Knez

University of Maribor, Maribor, Slovenia

Igor Lacik Polymer Institute of the Slovak Academy of Sciences,

Bratislava, Slovakia Denis Poncelet

ENITIAA, Nantes, France

Ljubisa Radovic Pen State University,

PA, USA Peter Raspor

University of Ljubljana, Ljubljana, Slovenia

Constantinos Vayenas University of Patras,

Patras, Greece Xenophon Verykios University of Patras,

Patras, Greece Ronnie Willaert

Vrije Universiteit, Brussel, Belgium

Gordana Vunjak Novakovic Columbia University,

New York, USA Dimitrios P. Tassios

National Technical University of Athens, Athens, Greece

Hui Liu China University of Geosciences, Wuhan, China

FORMER EDITOR (2005-2007) Professor Dejan Skala

University of Belgrade, Faculty of Technology and Metallurgy, Belgrade, Serbia

Journal of the Association of Chemical Engineers of Serbia, Belgrade, Serbia

Vol. 26 Belgrade, January-March 2020 No. 1

Chemical Industry & Chemical EngineeringQuarterly (ISSN 1451-9372) is published

quarterly by the Association of ChemicalEngineers of Serbia, Kneza Miloša 9/I,

11000 Belgrade, Serbia

Editor:Vlada B. Veljković

[email protected]

Editorial Office:Kneza Miloša 9/I, 11000 Belgrade, Serbia

Phone/Fax: +381 (0)11 3240 018E-mail: [email protected]

www.ache.org.rs

For publisher:Tatijana Duduković

Secretary of the Editorial Office:Slavica Desnica

Marketing and advertising:AChE Marketing Office

Kneza Miloša 9/I, 11000 Belgrade, SerbiaPhone/Fax: +381 (0)11 3240 018

Publication of this Journal is supported by theMinistry of Education, Science and

Technological Development of the Republic of Serbia

Subscription and advertisements make payableto the account of the Association of Chemical

Engineers of Serbia, Belgrade, No. 205-2172-71, Komercijalna banka a.d., Beograd

Computer typeface and paging:Vladimir Panić

Printed by:Faculty of Technology and Metallurgy,

Research and Development Centre of Printing Technology, Karnegijeva 4, P.O. Box 3503,

11120 Belgrade, Serbia

Abstracting/Indexing:Articles published in this Journal are indexed inThompson Reuters products: Science Citation

Index - ExpandedTM - access via Web of Science®, part of ISI Web of KnowledgeSM

CONTENTS

Dragan Živančev, Bojan Jocković, Novica Mladenov, Alek-sandra Torbica, Miona Belović, Branka Mijić, Jordana Ninkov, The effects of wheat cultivars on the pro-duction of different types of wheat flours of precisely defined magnesium content ................................................... 1

Pejman Roohi, Esmaeil Fatehifar, Enhanced treatment of 2-methylpropane-2-thiol contaminated soil using magno-modified Fenton process ............................................ 9

Karla Raphaela, Braga de Melo, Gabriela Cantarelli Lopes, Dayana de Gusmão Coêlho, João Inácio Soletti, Liquid-liquid equilibrium for systems composed by biodiesel from Catolé oil (Syagrus cearensis), methanol and glycerol .......................................................... 21

Ana Elisa Achiles, Vádila Giovana Guerra, Performance of a cyclone scrubber in removal of fine particulate matter ......... 31

Marija Lješević, Jelena Milić, Gordana Gojgić-Cvijović, Tat-jana Šolević Knudsen, Mila Ilić, Jelena Avdalovic, Miroslav M. Vrvić, Evaluation of assays for screening polycyclic aromatic hydrocarbon-degrading potential of bacteria ................................................................................. 41

Mohd Azahar Mohd Ariff, Norfazilah Abdullah, Optimization of reflux extraction for cat’s whiskers leaves extract using response surface methodology ................................... 49

Eduardo Ramos Braga, George de Souza Mustafa, Danilo de Aguiar Pontes, Luiz Antônio Magalhães Pontes, Eco-nomic analysis and technicalities of acrylic acid production from crude glycerol ............................................. 59

Erhan Sulejmani, Muhamet Demiri, Volatile compounds of Macedonian fermented sausage as affected by ripening process using SPME/GC-MS ................................. 71

Marija Jokanović, Bojana Ikonić, Predrag Ikonić, Vladimir Tomović, Tatjana Peulić, Branislav Šojić, Snežana Škaljac, Maja Ivić, Jelena Ivetić, Towards reproducibility of traditional fermented sausages: Texture profile analyses and modelling ................................ 79

Evgeny Akulinin, Oleg Golubyatnikov, Dmitry Dvoretsky, Stanislav Dvoretsky, Optimization and analysis of pressure swing adsorption process for oxygen production from air under uncertainty .................................... 89

Activities of the Association of Chemical Engineers of Serbia are supported by:

- Ministry of Education, Science and Technological Development, Republic of Serbia - Hemofarm Koncern AD, Vršac, Serbia - Faculty of Technology and Metallurgy, University of Belgrade, Belgrade, Serbia - Faculty of Technology, University of Novi Sad, Novi Sad, Serbia - Faculty of Technology, University of Niš, Leskovac, Serbia - Institute of Chemistry, Technology and Metallurgy, University of Belgrade, Belgrade, Serbia

Chemical Industry & Chemical Engineering Quarterly

Available on line at Association of the Chemical Engineers of Serbia AChE www.ache.org.rs/CICEQ

Chem. Ind. Chem. Eng. Q. 26 (1) 1−7 (2020) CI&CEQ

1

DRAGAN ŽIVANČEV1

BOJAN JOCKOVIĆ1 NOVICA MLADENOV1

ALEKSANDRA TORBICA2

MIONA BELOVIĆ2

BRANKA MIJIĆ1 JORDANA NINKOV1

1Institute of Field and Vegetable Crops, Novi Sad, Serbia

2University of Novi Sad, Institute of Food Technology, Novi Sad,

Serbia

SCIENTIFIC PAPER

UDC

THE EFFECTS OF WHEAT CULTIVARS ON THE PRODUCTION OF DIFFERENT TYPES OF WHEAT FLOURS OF PRECISELY DEFINED MAGNESIUM CONTENT

Article Highlights • Increase in the production of brown flour in the milling industry • Differentiation in Mg mineral content between white and brown flours • Daily requirements for magnesium fulfilled by consumption of brown bread Abstract

Whole kernels of cereals are the most important source of magnesium. This mineral has several positive effects on human health. However, the human body cannot absorb 100% of magnesium from plant sources during digestion. Additionally, the wheat flour milling process usually produces refined flour with a substantially lower content of magnesium. The aim of this study was to examine the effect of milling of two wheat cultivars on total and soluble mag-nesium distribution in milling fractions, with the purpose of creating wheat bread with a precisely defined magnesium content. Ash content, thousand grain weight (TGW), and kernel size were analysed in five wheat cultivars. Two most statistically distinguished wheat cultivars according to these analyses (Moma and Todorka) were milled in a laboratory mill to gain as many flour frac-tions as possible (eleven). Magnesium was extracted from the flour and its content was measured by inductively coupled plasma. The results showed that the level of soluble magnesium in refined white flours (T-500 with ash content between 0.46-0.60d.m.%) is approximately 17% and is available for uptake in the human body. Also, by milling the Moma cultivar in an industrial mill with a capacity of 100 t per day gave 6.6 t more brown flour (T-1000 with ash content between 1.05-1.15 d.m.%) than the Todorka cultivar. Furthermore, the daily consumption of brown bread (produced from brown flour) in Serbia would satisfy about 60% of the daily magnesium requirements.

Keywords: magnesium, wheat, flour, milling process.

Cereals are known as the most important dietary source of the mineral magnesium (Mg). Several studies [1-2] hypothesised that the consumption of food rich in magnesium is linked to a decreased risk of diabetes. Moreover, this was confirmed in the endocrinological surveys of Liu et al. [3] and Fung et al. [4] who examined the influence of whole-grain food

Correspondence: D. Živančev, Small Grains Department, Institute of Field and Vegetable Crops, Maksima Gorkog 30, 21101 Novi Sad, Serbia. E-mail: [email protected] Paper received: 4 October, 2018 Paper revised: 19 April, 2019 Paper accepted: 3 July, 2019

https://doi.org/10.2298/CICEQ181004019Z

intake and the risk of diabetes mellitus type 2. Also, it is well known since the beginning of the 20th century [5] that magnesium is an effective treatment against depression. Furthermore, Wacker et al. [6] showed that magnesium deficiency could cause numerous neuromuscular symptoms which could be prevented by magnesium intake. Magnesium supplementation is necessary for people living in modern society where they are exposed to chronic stress. Cernak et al. [7] showed that chronic stress decreased free and total plasma ionized magnesium in humans.

Almost every mineral in cereal grains is com-plexed by phytic acid, which results in forming salt compounds known as phytates. Degradation of phyt-ates in the human stomach is connected with calcium

D. ŽIVANČEV et al.: THE EFFECTS OF WHEAT CULTIVARS… Chem. Ind. Chem. Eng. Q. 26 (1) 1−7 (2020)

2

intake by food diet [8]. Also, according to Simpson and Wise [9] these salts could be soluble or insoluble depending on the degree of metal bonding to phos-phate groups. During digestion in the stomach and the small intestine it is possible to degrade from 1/3 to 2/3 of total phytate content consumed in a plant rich diet [10,11].

Milling is the most important technological pro-cess affecting the quantity of essential mineral ele-ments in foods obtained from wheat. One of the prime aims of the milling process is to segregate wheat endosperm from bran and germ as much as possible. Considering the fact that approximately 60% of the total mineral content of the wheat kernel is concen-trated in the aleurone layer [12], it is obvious that better separation will cause lesser content of mineral elements in refined or white flour. Also, Pomeranz [13] showed the distinction in the mineral content between the inner part of wheat endosperm and the outer part of wheat endosperm. Furthermore, the ker-nel geometry, which depends on wheat cultivar, also affects the mineral content of the obtained final pro-duct - flour.

Therefore, the aim of this study was to examine the effect of two wheat cultivars on total and soluble magnesium distribution in milling fractions, with the purpose of creating of wheat bread with precisely def-ined magnesium content.

MATERIALS AND METHODS

In order to select the suitable cultivars for the analysis, five wheat cultivars (Moma, Azra, Vlajna,

Efrosinija and Todorka) from the breeding program of the Institute of Field and Vegetable Crops, Novi Sad, Serbia were examined. Cultivation of these cultivars was performed at the experimental field Rimski šan-čevi, Novi Sad, Serbia (45°20’ N, 19°51’ E).

The content of water in grain was determined according to the method SRPS EN ISO 712:2012 [14] and the ash content in grain was determined accord-ing to SRPS EN ISO 2171:2012 [15]. Grain size was determined using Pfeuffer device Sortimat (Kitzingen, Germany) by sifting 100 g of clean grain sample for 1 min through three sieves with openings of 20 mm×2.8 mm, 20 mm×2.5 mm and 20 mm×2.2 mm [16]. Thousand grain weight (TGW) was determined according to SRPS EN ISO 520:2012 [17].

The moisture of chosen two grain cultivars was measured and first adjusted to 13.5% for 24 h and after that to 15.0% half an hour before milling in a Bühler pneumatic laboratory mill MLU 202 (Uzwil, Switzerland) and using a Bühler Bran Duster (Uzwil, Switzerland) (Figure 1). The 11 flour streams of the chosen wheat cultivars were gained from three break passages (B1-B3); three reduction passages (M1-M3) whereas an additional flour stream B4 and an addit-ional flour stream M4 were derived by passing bran and shorts through a bran duster. The remaining bran obtained after using the bran duster was reground by breaking rolls of MLU 202 and all three break pas-sages were used for making an additional flour stream B5. The remaining shorts gained after using the bran duster were reground by reduction rolls of MLU 202 and all three reduction passages were used for making and additional flour stream M5. The last

Figure 1. Flow diagram for a used Bühler laboratory mill.


3

flour stream M6 was gained by passing shorts obtained after bran regrinding trough the bran duster. Additionally, by using the milling scheme (Figure 2), one bran fraction and two shorts fractions (Shorts 4 and 5) were gained. All 14 streams were well mixed in a cubic mixer and prepared for further analyses.

The water and ash content in streams were det-ermined according to the above-mentioned methods [14,15] in duplicate. The samples of ash content were made in two replicates, and the standard deviation between the repetitions was within the range of 0.0 to 2.0% of the obtained values.

The flour stream samples, shorts and bran and whole grain of two wheat cultivars, were analysed for total contents of magnesium (MgT) after microwave digesting the samples in concentrated HNO3 and H2O2 (0.5 mg samples in digestion solution 10 ml HNO3 + 2 ml H2O2, Vu = 50 ml) by stepwise heating up to 180 °C using a Milestone Vario EL III for 35 min. The concentration of magnesium was determined by inductively coupled plasma-optical emission spectro-

scopy (ICP-OES, Vista Pro-Axial, Varian). Quality assurance and quality control (QA/QC) were con-ducted by the results of successfully passed AFPS Animal Feeds PT Scheme provided by LGC and by standard reference material NIST SRM 1515 - apple leaves. The obtained results achieved analytical pre-cision of 0.116 mg kg-1 Mg. The accuracy was within the interval 96.14 to 101.75%, and recovery was 98.52%. The limit of quantification (LOQ) in our study was 7.417 mg kg-1 Mg and limit of detection (LOD) was 2.45 mg kg-1 Mg, which provided adequate sen-sitivity to analyse.

The extraction of soluble minerals from flour fractions, shorts and bran from gained milling streams and whole grain of each grain cultivar was extracted with Tris-HCl buffer according to procedure described by Guttieri et al. [18] with some minor modifications. Namely, after vortexing, 70 mg of bran or 140 mg of flour was extracted with 7 mL of sterile Tris-HCl buffer (50 mM, pH 7.5) in a sterile tube at room temperature (22 °C) for 22 h with gentle shaking (100 rpm at

Figure 2. Cumulative mineral content curves for ash (d.m.% basis), total and soluble Mg (mg kg−1 d.m. basis) for wheat cultivars

Moma (A) and Todorka (B).


4

orbital shaker PSI-10i, Boeco, Germany). Samples were again vortexed and then centrifuged (15000 g, 20 min) at Eppendorf centrifuge 5804 R, Hamburg, Germany. An aliquot of the supernatant was removed and stored at -20 °C prior to analysis. Supernatants were filtered through 0.2 μm cellulose acetate syringe filters (VWR International) and diluted in an equal vol-ume of 10% nitric acid for analysis by ICP. Magne-sium concentrations were measured with duplicate instrumental analyses of two technical replicates. Immediately prior to detection, the samples were fil-tered once again and the concentration of soluble Mg was determined by ICP-OES (Vista Pro-Axial, Vari-an). The samples of total and soluble magnesium were made in two replicates, and the standard devi-ation between the repetitions was within the range of 0.45 to 9.31% of the obtained values.

The results of ash content, TGW and grain size were statistically processed by analysis of variance (ANOVA). The statistical analyses were performed by InfoStat (Info Stat Student, Version 2016e) software [19]. Cumulative curves of ash, total and soluble mag-nesium of wheat cultivars were calculated on a dry matter basis (d.m. basis), following closely the pro-cedure described by Fišteš and Tanović [20] for cal-culation of the cumulative flour ash curves.

RESULTS AND DISCUSSION

In order to select wheat cultivars which will be milled according to flow diagram (Figure 1) ash con-tent, TGW and grain sizes of five wheat cultivars were examined (Table 1). The one-way ANOVA showed that Moma wheat cultivar possessed significantly the highest ash content and percentage of grains >2.5 mm of all cultivars, whereas Todorka cultivar pos-sessed the smallest significant values of these two parameters. In the case of percentage of grains > 2.8 mm the two cultivars had the opposite trends. Also, the Todorka cultivar had the highest TGW value of all wheat cultivars and the Moma cultivar had the small-est TGW value. Ash content is strongly connected with the presence of aleurone cells in milling fractions [21]. On the other hand, according to Cochrane and

Duffus [22] there is a positive correlation between endosperm cell number and TGW. Because of the above-mentioned results, it was assumed that there was the largest difference in quantity of aleurone layer and endosperm between these two cultivars. There-fore, they were chosen for milling and further exam-ination of magnesium distribution in milling fractions.

The total and soluble Mg content in whole grain are reported in Table 2. Those values were in the range (1215–2165 mg kg−1) reported by Zhang et al. (2010) [23], who examined 265 winter wheat genotypes from breeding programs in the Northern China. The soluble Mg content in wheat grains was almost three times lower than the content of total Mg.

Table 2. Content (mg/kg) of total and soluble magnesium in whole wheat grain

Wheat cultivar

Total magnesium content, mg kg−1

Soluble fraction of magnesium content, mg kg−1

Moma 1906.00 694.50 Todorka 1754.00 643.50

The largest fraction of flour in both cultivars was obtained from the first reduction passage M1 (31.6 and 30.2% of product yield for Moma and Todorka cultivar, respectively). Additionally, the ash content of the M1 flour stream of Moma (0.567% d. m. basis) and Todorka (0.536% d. m. basis) cultivars was slightly above the ash content of flour streams M3 and B1 (0.505 and 0.533% d.m. basis for Moma and 0.462 and 0.431% d.m. Todorka cultivar, respect-ively), as shown in Tables 3 and 4. Four flour streams (M3, B1, M1 and M2) obtained by milling of Moma cultivar, comprised 50.8% of total milled products. Also, the calculated ash content of mixture of these four flour streams according to the equation for cum-ulative flour ash curves [20] was 0.558% d.m. basis and the calculated total and soluble Mg content were 858.0 and 347.8 mg kg-1 d.m. basis, respectively. The calculated ash content of gained mixed flour was in the range of the refined white T-500 flour (from 0.46 to 0.60% d.m. basis) [24] which is usually used for bread production in Serbia. Milling the Todorka cul-tivar gained five flours streams (B1, M3, M1, M2 and

Table 1. Mean values of ash content, TGW and grain sizes of five wheat cultivars analysed by one-way ANOVA; Duncan test, different letters indicate significant difference at 0.05 probability level

Wheat cultivar Ash content (%) TGW (g dm. basis) > 2.8 mm (%) > 2.5 mm (%) > 2.2 mm (%) Bottom (%) Moma 1.70a 33.15a 36.70d 48.43a 8.19a 6.76b Azra 1.65b 33.97a 46.26b 36.24e 7.21a 10.33a Vlajna 1.63bc 33.99a 39.80c 42.11c 7.98a 10.11a Efrosinija 1.60cd 34.39a 37.73cd 45.21b 7.66a 9.37a Todorka 1.58d 36.22a 50.47a 38.03d 4.57b 6.95b


5

B2), comprised 57% of total milled products. Also, the calculated ash content of the mixture of these four flour streams according to the equation for cumulative flour ash curves [20] was 0.520% d.m. basis, in the range of the refined white T-500 flour, and calculated total and soluble Mg content 799.8 and 300.8 mg kg-1 d.m. basis. Calculated total Mg content of the mix-tures with value of ash parameters in the range of ref-ined white T-500 flour represents approximately 45% of the total Mg content in grain of both examined wheat cultivars. This is almost three times higher than the level of 16% which Anonymous (2002) [25] found in refined flour when compared to whole grain flour. Also, the calculated level of soluble Mg content of the mixtures with value of ash parameters in the range of refined white T-500 flour represents approximately 17% and it is available for uptake in the human body. By mixing five flour streams of Moma cultivar (B2, M4, B3, B5 and M5), which makes 26.50% of extraction rate, the gained flour would be with ash content 1.118% d.m. basis, total and soluble Mg content 1403.1 and 797.0 mg kg-1 d.m. basis (calculated according to equation for cumulative flour ash curves [20]). This kind of flour belongs to the refined T-1100 flour (brown flour) since the ash content was in the range from 1.05 to 1.15% d.m. basis [24]. By mixing four flour streams of Todorka cultivar (M4, B3, B5 and M5), which makes 19.90% of extraction rate, flour will be gained with ash content 1.158% d.m. basis (upper limit for refined dark T-1100 flour), total and soluble Mg content 1423.0 and 852.0 mg kg-1 d.m. basis (cal-culated according to the equation for cumulative flour ash curves [20]).

In order to assess the nutritional values of white and dark flours and Mg concentrations, Mg content of

a standard 100 g slice of commercially prepared white and brown bread was calculated. According to the USDA National Nutrient Database for Standard Refer-ence (NDB 18069), the bread slice contains 36.7% water. In a standard Basic Straight-Dough Bread-Bak-ing Method-Long Fermentation [26] the flour com-prises 86% of the dry ingredients. Therefore, a stan-dard 100 g slice of white bread prepared from the Moma cultivar white mixture of four flour streams (M3, B1, M1 and M2) in this study contained 59.7 mg of total Mg and 24.2 mg of soluble Mg, whereas 100 g slice of white bread prepared from a white mixture of five flour streams (B1, M3, M1, M2 and B2) obtained from the Todorka cultivar contained a total of 55.6 mg Mg and 20.9 mg of soluble Mg. A standard slice of brown bread prepared from Moma cultivar brown mix-ture of five flour streams (B2, M4, B3, B5 and M5) in this study contained 97.5 mg of total Mg and 55.4 mg of soluble Mg, whereas a standard slice of brown bread prepared from the Todorka cultivar dark mix-ture of four flour streams (M4, B3, B5 and M5) in this study contained 98.9 mg of total Mg and 59.2 mg of soluble Mg. Taking into account that the average daily recommended value for Mg for adults between the ages 19 to 65 is 220 mg for women and 260 mg for men [27], the consumption of only 400 g of brown bread would fulfil total Mg requirements (only soluble Mg is taken into account). Average bread con-sumption in the EU per year is 59.4 kg, whereas Bal-kan countries such as Greece, Bulgaria and Turkey eat bread even more (68.0, 95.0 and 104.0 kg per year) [28]. The average bread consumption in Serbia is at the level of the Balkan countries, 83.7 kg per year [29] or 229 g per day. It could be concluded that

Table 3. Millstream yields (%), estimated ash (d.m. basis), total and soluble magnesium content with the streams sorted in the order of increasing ash content for wheat cultivar Moma

Stream Yield, % Ash, % d.m. basis Total magnesium, mg kg−1 d.m. basis Soluble magnesium, mg kg−1 d.m. basis M3 6.00 0.505 857.07 320.49 B1 7.40 0.533 862.14 353.16 M1 31.60 0.567 856.49 347.72 M2 5.80 0.594 862.10 369.58 B2 8.20 0.703 1034.07 464.96 M4 13.40 1.166 1474.36 829.58 B3 1.60 1.560 1802.90 1258.02 B5 1.60 1.633 1834.72 1194.81 M5 1.70 1.833 1838.52 1334.18 M6 1.50 3.555 3275.82 2660.08 B4 1.10 3.970 3992.72 3413.72 Shorts5 7.30 5.781 5354.49 3326.11 Bran 2.00 6.418 5614.76 2749.54 Shorts4 8.00 6.765 6199.06 4591.15


6

by the average daily consumption of bread about 60% of magnesium requirements would be met.

The calculated cumulative curves for ash total and soluble Mg of wheat cultivars Moma (A) and Todorka (B), identical to the cumulative ash curve, are shown in Figure 2. This figure is the most usual presentation of data in the milling industry since it is one of the best ways to evaluate performance of the used mill and allow precise selection of gained flour streams for making the final products (flour) for dif-ferent purposes [30]. The curve of both cultivars showed that concentration of total and soluble Mg in the lowest ash flour streams was relatively low, whereas it was high in the shorts and bran. Hence, it confirmed that most of the soluble Mg is stored in the external parts of the kernel.

CONCLUSIONS

According to the results of refined white flour, it could be concluded that the level of soluble mag-nesium in refined white flours (T-500 with ash content between 0.46-0.60 d.m.%) represents approximately 17% and it is available for uptake in human body. Also, by milling the Moma cultivar it was possible to obtain 26.50% of the extraction rate of brown flour, whereas the Todorka cultivar gained about 7% lower extraction rate of the brown flour. This means that the industrial mill with the capacity of 100 t per day will pro-duce approximately 6.6 t more of this valuable nutri-tive product by using the Moma cultivar as a raw mat-erial than by using the Todorka cultivar. Furthermore, consuming brown bread on the average level in Serbia would satisfy about 60% of magnesium requirements.

Acknowledgments

This work was funded by the Serbian Ministry of Education, Science and Technological Development (project TR31066 and project TR31007).

REFERENCES

[1] R. Jiang, J.E. Manson, M.J. Stampfer, S. Liu, W.C. Willett, F.B., Hu, JAMA 288 (2002) 2554-2560

[2] E.J. Feskens, C.H. Bowles, D. Kromhout, Am. J. Clin. Nutr. 54 (1991) 136-140

[3] S. Liu, J. Am. Coll. Nutr. 21 (2002) 298-306

[4] T.T. Fung, F.B. Hu, M.A. Pereira, S. Liu, M.J. Stampfer, G.A. Colditz, W.C. Willett, Am. J. Clin. Nutr. 76 (2002) 535-540

[5] P.G. Weston, Am. J. Psychiatry 78 (1921) 637-638

[6] W.E. Wacker, A.F. Parisi, N. Engl. J. Med. 278 (1968) 712-719

[7] I. Cernak, V. Savic, J. Kotur, V. Prokic, B. Kuljic, D. Grbo-vic, M. Veljovic, Magnes. Res. 13 (2000) 29-36

[8] A.R. Walker, F.W. Fox, J.T. Irving, Biochem. J. 42 (1948) 452–62

[9] C.J. Simpson, A. Wise, Br. J. Nutr. 64 (1990) 225-232

[10] A.S. Sandberg, H. Andersson, J. Nutr. 118 (1988) 469-473

[11] R.A. McCance, E.M. Widdowson, Biochem. J. 29 (1935) 2694-2699

[12] J.J.C. Hinton, Cereal Chem. 36 (1959) 19-31

[13] Y. Pomeranz, in Wheat Chemistry and Technology. Chemical composition of kernel structures, Vol. I, Y. Pomeranz (Ed.), American Association of Cereal Chem-ists, St. Paul, MN, 1988, p. 97

[14] SRPS EN ISO 712:2012. Cereals and cereal products —Determination of moisture content Reference method. Institute for standardization of Serbia, Belgrade, Serbia

Table 4. Millstream yields (%), estimated ash (d.m. basis), total and soluble magnesium content with the streams sorted in the order of increasing ash content for wheat cultivar Todorka

Stream Yield, % Ash, % d.m. basis Total magnesium, mg kg−1 d.m. basis Soluble magnesium, mg kg−1 d.m. basis

B1 7.40 0.431 765.50 269.24

M3 6.00 0.462 777.01 241.44

M1 30.20 0.536 800.06 306.53

M2 5.80 0.541 758.73 301.51

B2 7.60 0.578 881.62 354.94

M4 14.90 0.984 1287.91 719.04

B3 1.80 1.354 1607.67 1012.71

B5 2.00 1.533 1733.45 1157.13

M5 1.20 2.396 2306.03 1753.65

M6 1.30 3.974 3668.16 2992.59

B4 1.20 4.105 3936.68 3337.30

Shorts5 7.30 6.198 6466.45 3247.53

Shorts4 7.70 6.725 5965.07 3771.38

Bran 1.30 7.233 5843.93 2815.47


7

[15] SRPS EN ISO 2171:2012. Cereals, pulses and by-pro-ducts – Determination of ash yield by incineration. Insti-tute for Standardization of Serbia, Belgrade, Serbia

[16] G. Kaluđerski, N. Filipović, In Methods of testing the qual-ity of wheat, flour and finished products, M. Žeželj, R. Vukobratović (Eds.), University of Novi Sad, Faculty of Technology, 1998, p. 25

[17] SRPS EN ISO 520:2012. Cereals and pulses – Determin-ation of the mass of 1000 grains, Institute for Stan-dardization of Serbia, Belgrade, Serbia

[18] M.J. Guttieri, B.W. Seabourn, C. Liu, P.S. Baenziger, B.M. Waters, J. Agric. Food Chem. 63 (2015) 10681- –10688

[19] InfoStat (Info Stat Student, 2016, Version 2016e, UNC, Argentina (www.infostat.com.ar)

[20] A. Fišteš, G. Tanović, In Practice book in milling techno-logy, Z. Zavargo, (Ed.), University of Novi Sad, Faculty of Technology, 2014, p. 75

[21] S. Jensen, L. Munck, H. Martens, Cereal Chem. 59 (1982) 477-484

[22] M.P. Cochrane, C.M. Duffus, Ann. Appl. Biol. 102 (1983) 177-181

[23] Y. Zhang, Q. Song, J. Yan, J. Tang, R. Zhao, Y. Zhang, Z. He, C. Zou, I. Ortiz-Monasterio, Euphytica 174 (2010) 303-313

[24] Regulation on the quality of grains, milling and bakery products and pasta, Official Gazette of the Republic of Serbia, 68, 2018

[25] Anonymous, Harv. Heart Lett. 13 (2002) 2-4

[26] American Association of Cereal Chemists (AACC), Approved Methods of Analysis, 10th ed., Methods (AACC Method 10-09), AACC, St. Paul, MN, 2000

[27] Report of a joint FAO/WHO expert consultation Bangkok, Thailand (2001) Human Vitamin and Mineral Require-ments (http://www.fao.org/3/a-y2809e.pdf)

[28] A. Eglite, D. Kunkulberga, Foodbalt, in Proceedings of 11th Baltic conference on food science and technology “Foodbalt 2017”, Jelgava, Latvia, 2017, p. 178

[29] B. Gulan, Macroeconomy, Bread and its consumption in Serbia, 2016 (in Serbian) https://www.makroekono-mija.org/poljoprivreda/hleb-i-potrosnja-u-srbiji2016/

[30] S.D. Sakhare, A.A. Inamdar, J. Food Sci. Technol. 51 (2014) 795-799.

DRAGAN ŽIVANČEV1 BOJAN JOCKOVIĆ1

NOVICA MLADENOV1

ALEKSANDRA TORBICA2

MIONA BELOVIĆ2 BRANKA MIJIĆ1

JORDANA NINKOV1 1Institut za ratarstvo i povrtarstvo,

Maksima Gorkog 30, 21101 Novi Sad, Srbija

2Univerzitet u Novom Sadu, Institut za prehrambene tehnologije, Bulevar Cara

Lazara 1, 21000 Novi Sad, Srbija

NAUČNI RAD

UTICAJ MLINSKE SIROVINE NA PROIZVODNJU TIPSKOG BRAŠNA TAČNO DEFINISANOG SADRŽAJA MAGNEZIJUMA

Zrna žita su najznačajniji izvor magnezijuma. Ovaj mineral ima nekoliko pozitivnih efekata na zdravlje ljudi. Međutim, ljudski organizam nema mogućnost da prilikom pro-cesa varenja iz hrane biljnog porekla apsorbuje celokupnu količinu magnezijuma. Mle-venjem pšenice uobičajeno se proizvode bela brašna, koja se i najčešće konzumiraju, što dodatno smanjuje količinu magnezijuma u ishrani. Cilj ovog istraživanja je bio da se ispita kako izbor mlinske sirovine može uticati na proizvodnju brašna i hleba od istog brašna sa tačno definisanim sadržajem magnezijuma. Da bi se odabrala odgovarajuća sorta pšenice prvo je na pet sorti analiziran sadržaj pepela, masa hiljadu zrna i veličina zrna. Dve sorte (Moma i Todorka) koje su se statistički najviše razlikovale po ovim oso-binama su samlevene na laboratorijskom mlinu tako da se dobije maksimalan broj pasažnih brašna (jedanaest). Iz ovih pasaža ekstrahovan je magnezijum, a njegov sadr-žaj je određen tehnikom induktivne spregnute plazme. Rezultati su pokazali da je nivo rastvorljivog magnezijuma u belom brašnu 17% tako da bi se prilikom procesa varenja ova količina mogla u celosti abpsorbovati. Takođe, mlevenjem sorte Moma u industrij mlinu kapaciteta 100 t dnevno bi se proizvelo 6,6 t više crnog brašna nego što je to slučaj sa sortom Todorka. Na ovaj način bi se u Srbiji konzumiranjem crnog hleba zadovoljilo 60% dnevnih potreba za magnezijumom.

Ključne reči: magnezijum, pšenica, brašno, mlevenje.



Chem. Ind. Chem. Eng. Q. 26 (1) 9−20(2020) CI&CEQ

9

PEJMAN ROOHI

ESMAEIL FATEHIFAR

Environmental Engineering Research Center (EERC), Faculty of Chemical Engineering, Sahand University of Technology, Sahand

New Town, Tabriz, Iran

SCIENTIFIC PAPER

UDC 502.521:631.41:66.094.3

ENHANCED TREATMENT OF 2-METHYLPRO-PANE-2-THIOL CONTAMINATED SOIL USING MAGNO-MODIFIED FENTON PROCESS

Article Highlights • Magno-modified Fenton process is used for remediating 2-methylpropane-2-thiol pol-

luted soil • H2O2, NaOH, Fe3O4 and ethanol concentrations in the magnetic field are the influ-

encing factor • The batch reactor exposed to the magnetic field is a good way of separating Fe3O4 • H2O2/nanoFe3O4 and NaOH/nanoFe3O4 are the influencing interaction with a positive

role • NanoFe3O4/ethanol, H2O2/NaOH and magnetic field/nanoFe3O4 have a negative role

in the remediation yield Abstract

In this work, the nano-magno-modified Fenton process for remediation of soil polluted with 2-methylpropane-2-thiol is investigated. In this two-step oxidation process, a nanocatalyst system is used, including Fe2+/Fe3+ coupled with etha-nol solvent exposed to the magnetic field. The data analysis resulting from analysis of variance shows that H2O2, NaOH, Fe3O4 nanoparticle and ethanol concentration with magnetic field intensity are effective on 2-methylpropane-2--thiol removal efficiency (26.99, 4.36, 3.42, 17.63 and 8.66%, respectively). Moreover, H2O2/nanoFe3O4 and NaOH/nanoFe3O4 were the influencing inter-action with a positive role and nanoFe3O4/ethanol, H2O2/NaOH and magnetic field/nanoFe3O4 with a negative role. The maximum level of initial H2O2 con-centration, a minimum level of ethanol and certain levels of NaOH initial con-centration (8.64 mass%), Fe3O4 nanoparticles to soil ratio and magnetic field intensity (0.11 mass% w/w and 3585 Oe, respectively) were suggested for the maximum contaminant remediation efficiency in the studied interval (%R == 92.02). The magno-modified Fenton process could be used for both, inc-reasing the removal efficiency and minimizing the vaporization of similar org-anosulfur compounds. Experimental results show that the solvent addition for diluting thiol contaminant could reduce the removal efficiency while the coup-ling of Fe3O4 and magnetic field is a good way of separation and reuse to solve the recovery problem of the nanoparticles.

Keywords: advanced oxidation, environmental remediation, mathem-atical modeling, optimization, volatile organic compounds (VOCs).

2-methylpropane-2-thiol (tert-butyl mercaptan) is widely used for various purposes in the rubber ind-

Correspondence: E. Fatehifar, Environmental Engineering Res-earch Center (EERC), Faculty of Chemical Engineering, Sahand University of Technology, Sahand New Town, Tabriz, Iran. E-mail: [email protected] Paper received: 17 April, 2019 Paper revised: 13 June, 2019 Paper accepted: 10 July, 2019

https://doi.org/10.2298/CICEQ190417020R

ustry and natural gas odorization due to its repulsive odor. The release of this organosulfur compound has hazardous effects on humans [1]. Moreover, there are several environmental concerns with secondary social consequences due to air and soil pollution. According to these impacts, the entry of 2-methylpropane-2-thiol into the soil and its vaporization is one of the environ-mental concerns.

Over the years, Fenton treatment has been modified to achieve adequate degradation [2-4] using

P. ROOHI, E. FATEHIFAR: ENHANCED TREATMENT OF 2-METHYLPROPANE-... Chem. Ind. Chem. Eng. Q. 26 (1) 9−20 (2020)

10

H2O2 coupled with a catalyst [5] such as magnetite [6,7]. Moreover, chelating agents and stabilizers [8,9] or goethite [10] is used for increasing the efficiency of the modified Fenton process. In this way, nano-scale iron catalysts are widely applied to the remediation of contaminated soil [11,12]. Magnetite nanoparticles (nFe3O4) contain both Fe2+ and Fe3+ which react as catalysts [6]. These nanoparticles pass through the porous media and remain suspended in aquifer sys-tems that increase reactivity. Fe3O4 nanoparticles tend to collect in a magnetic field [13]. Therefore, the use of magnetite is a good way for separation and reuse to solve the recovery problem of the nanopar-ticles and decreasing remediation cost.

The hydrocarbon remediation from the soil using modified Fenton treatment could have an unlimited range of efficiency. The physicochemical character-istics of the soil, pollutant and reactants, pollutant aging, reactant dosage, acidity or alkalinity of a sol-ution, soil texture, type of stabilizers and catalyst and reaction time are the various influence factors [14]. However, hydroxyl radicals play a key role in the rem-oval efficiency based on the main reactions which are represented as follows [15,16]:

H2O2+Fe2+→Fe3++●OH+OH¯ (1) ●OH+RH→●R+H2O (2) ●OH+Fe2+→Fe3++OH (3) Fe3++H2O2→Fe2++●HO2

+H+ (4) ●OH+H2O2→H2O+●HO2 (5) ●OH+●OH→H2O2 (6) ●OH+●HO2→H2O+O2 (7) ●R+H2O2→ROH+●OH (8) Fe2++●HO2→Fe3++HO2 (9) Fe3++●HO2→Fe2++O2+2H+ (10) ●HO2+H2O2→H2O+●OH+O2 (11) ●HO2+

●HO2→H2O+O2 (12) ●HO2+O●

2¯+H2O→H2O2+O2+OH (13)

O●2

¯+Fe3+→Fe2++O2 (14) O●

2-+Fe2++2H+→Fe3++H2O2 (15)

The overall reaction for oxidation of thiol by hydrogen peroxide can also yield sulfonic acids [17]:

RSH+3H2O2→RSO3H+3H2O (16) Because 2-methylpropane-2-thiol is volatile, it

vaporizes when it released on the soil. So, the dev-elopment of methods offering low vaporization and oxidation of 2-methylpropane-2-thiol is required. One

of the methods for decreasing the vaporization of thiol is caustic addition.

The caustic solution reacts with thiol and dec-reases their vaporization. The following mechanism represented by Wallace [18] and Rollmann [19]:

OH¯+RSH→RS¯+H2O (17) 2RS¯+O2→2RS●+O2

2¯ (18) 2RS●→RSSR (19) O2

2¯+H2O→2OH¯+1/2O2 (20) The overall reaction in the presence of a catalyst

represented is as follows [20]:

2RSH+1/2O2→RSSR+H2O (21) Although there is some published research on

2-methylpropane-2-thiol removal from soil with 82% to 94% removal efficiency [21-23], they do not work on soil remediation and vaporization prevention of this contaminant simultaneously. Furthermore, even though there are some studies which couple the Fen-ton process with other process methods like micro-waves for increasing the remediation yield [24], there are a few studies which use a combination of a mag-netic field and chemical oxidants. The aim of this work was to study the combined system, in order to mini-mize vaporization and accomplish fast 2-methylpro-pane-2-thiol remediation from the polluted soil. Modi-fied Fenton process using sodium hydroxide and magnetite nanoparticles exposed to the magnetic field is studied. The effects of H2O2, NaOH, Fe3O4 nano-particles and ethanol concentration and magnetic field intensity are five experimental factors under study. Initial soil temperature is investigated in two blocks.

Central composite design was used as an expe-rimental design model, with response surface meth-odology used for the design of the experiment. Math-ematical modeling resulted from the applied method to predict the optimal experimental condition [25].

To the best of our knowledge, experimental ana-lysis of 2-methylpropane-2-thiol degradation from soil using modified Fenton process coupled with NaOH and Fe3O4 nanoparticles exposed to the magnetic field has not been published yet.

EXPERIMENTAL

Polluted soil

The sample soil was prepared from Sahand New Town, East Azerbaijan province in Iran which contain organic carbon (0.63%), cation exchange capacity (2.34 cmol/kg), crystalline and amorphous iron (19625 and 47.4 mg/kg, respectively) and crys-talline and amorphous magnesium (528 and 21.9


11

mg/kg, respectively) with 6.8 in pH. This soil is a mix-ture of clay (19.4%), sand (20.3%) and silt (60.3%) sampled from 20 cm depth of the soil.

The tubular reactor and chemical reactants

The materials used for the experiments were NaOH (purity ≥ 97.0%), C4H10S (98%) and H2O2 (35%) purchased from Merck (in reagent grade). Fe3O4 nanopowder (97% trace metals basis, 50-100 nm particle size and average surface area >60 m2/g) was supplied from Sigma-Aldrich. The “Pars Azma” deionizer was used to produce deionized water. Et-hanol was purchased in reagent grade from the “Razi” company (Iran). Magnetic field generation was pro-duced by a “Farayand Gostar” magnet (Iran).

All 100 ml cylindrical glass batch reactor parts were washed by deionized water and sterilized for 3 times (in 120 °C for 45 min) to eliminate undesirable contaminants (Figure 1). The first and final 2-methyl-propane-2-thiol concentrations in soil and air media were measured from one point of air and five points of soil around the reactor. These five points were aver-aged and recorded as the final result.

Figure 1. Schematic of the bench scale reactor used for

magno-modified Fenton process.

Material analyzing

Iron and manganese oxyhydroxides (crystalline and amorphous) were analyzed using citrate-bicarbo-nate-dithionite extractions process [26]. For the deter-mination of organic carbon content, sample soil was burned (at 900 °C) with emitted carbon dioxide trap-ped in potassium hydroxide and measured by back titration of unreacted hydroxyl ions [27]. Cation exchange capacity was analyzed using sodium ace-tate saturation (at pH 8.2).

2-Methylpropane-2-thiol concentration in the sample soil was analyzed by a GC device (Agilent

7890, TCD and FID detector, HP-Plot Q column) with methanol extract techniques. The fuel and carrier gas was H2 (99.999%) and He (99.999%) was used as make-up gas. The zero-grade dry air was used as support gas which was obtained from “Sabalan” Com-pany. The temperature program began at 105 °C for 2 min, ramped up at 12 °C/min to 240 °C, and was then held at 240 °C for 4 min. The detector temperatures and injection port were at 270 and 170 °C, respect-ively. 2-methylpropane-2-thiol was analyzed by inject-ing of extracted liquid manually using a syringe.

Procedure

At first, soil (25 g) polluted with 2-methylpro-pane-2-thiol (10% pollution by weight) was poured into the reactor, and pollution concentration in air and soil media of the reactor is measured. Then, NaOH (1.5 ml) is poured on the surface of the polluted soil in order to minimize thiol vaporization. Fe3O4 nanopar-ticles with ethanol are injected 5 mm below the sur-face of the soil. After that, contaminated soil which was remediated with 25 ml of H2O2, was agitated at 20 rpm and left for progress oxidation (20 min). Limited experimental time interval leads to minimizing the effects of biodegradation. The contaminant rem-oval efficiency (%R) against time is given by:

( ) −= 1 2

1

Removal efficiency % 100C C

C (22)

where C1 and C2 are the initial and final 2-methylpro-pane-2-thiol concentrations (ppm), respectively.

Design of experiments

In this study, a central composite design using response surface methodology was used to deter-mine the factors and their interaction effects.

In this remediation process, initial H2O2 concen-trations, initial concentrations of NaOH, Fe3O4 nano-particles and ethanol portion of the sample soil (mass%) and magnetic field intensity are the main investigated factors. Factor levels are specified due to legal consideration, safety, and pre-experiments. The factor levels are shown in Table 1.

Central composite design method

The experiments were divided into two blocks in order to study the effects of initial soil temperature (15 and 30 °C). Moreover, in order to eliminate the effects of uncontrolled factors on the results, the random-ization technique was used.

Coded variables are used for statistical calcul-ations for simplicity in experimental design as follows:


12

−=δ

cii

X XxX

(23)

where xi is the coded value of the independent vari-ables as a function of Xi (actual values of the inde-pendent variables), Xc (actual values of the indepen-

dent variables at the center point) and δX (step change).

Design of experiment, which is summarized in Table 1, includes 33 experiments (16 cubic points, 10 axial points (α = ±2) and 7 replication at the center point (α = 0)). This 5-factor 5-level matrix, was divided

Table 1. Experimental factors and their ranges and levels and the 5-factors 5-level CCD matrix with the observed and predicted responses

Independent variable Symbol Factor code

Ranges and levels

-2 -1 0 1 2

nFe3O4 to soil ratio, mass% (nFe3O4:S)0 X1 0.04 0.06 0.08 0.10 0.12

Magnetic field, Oe H(Oe) X2 700 2800 4900 7000 9100

Initial H2O2 concentration, % [H2O2]0 X3 1 3 5 7 9

Ethanol to soil ratio, mass% (Et:S)0 X4 5 10 15 20 25

Initial NaOH concentration, mass% [NaOH]0 X5 5 7.5 10 12.5 15

Standard order

Run order

Block Coded value Removal efficiency, %

(nFe3O4:S)0 H(Oe) [H2O2]0 (Et:S)0 [NaOH]0 Observed Predicted

13 1 1 -1.00 -1.00 1.00 1.00 1.00 63.310 62.707

4 2 1 1.00 1.00 -1.00 -1.00 1.00 66.501 67.193

5 3 1 -1.00 -1.00 1.00 -1.00 -1.00 73.370 71.407

9 4 1 -1.00 -1.00 -1.00 1.00 -1.00 60.780 57.564

20 5 1 0.00 0.00 0.00 0.00 0.00 70.510 71.021

12 6 1 1.00 1.00 -1.00 1.00 -1.00 44.056 44.694

18 7 1 0.00 0.00 0.00 0.00 0.00 71.380 71.021

6 8 1 1.00 -1.00 1.00 -1.00 1.00 81.801 82.196

1 9 1 -1.00 -1.00 -1.00 -1.00 1.00 63.070 65.203

15 10 1 -1.00 1.00 1.00 1.00 -1.00 64.022 63.072

19 11 1 0.00 0.00 0.00 0.00 0.00 68.660 71.021

14 12 1 1.00 -1.00 1.00 1.00 -1.00 74.140 74.031

2 13 1 1.00 -1.00 -1.00 -1.00 -1.00 64.455 61.599

16 14 1 1.00 1.00 1.00 1.00 1.00 68.151 65.292

7 15 1 -1.00 1.00 1.00 -1.00 1.00 70.610 69.212

10 16 1 1.00 -1.00 -1.00 1.00 1.00 71.262 69.258

21 17 1 0.00 0.00 0.00 0.00 0.00 73.010 71.021

22 18 1 0.00 0.00 0.00 0.00 0.00 73.181 71.021

11 19 1 -1.00 1.00 -1.00 1.00 1.00 59.850 56.869

17 20 1 0.00 0.00 0.00 0.00 0.00 71.347 71.021

3 21 1 -1.00 1.00 -1.00 -1.00 -1.00 62.610 64.069

8 22 1 1.00 1.00 1.00 -1.00 -1.00 75.232 71.967

31 23 2 0.00 0.00 0.00 0.00 -2.00 60.350 62.295

30 24 2 0.00 0.00 0.00 2.00 0.00 61.431 63.601

29 25 2 0.00 0.00 0.00 -2.00 0.00 79.910 78.441

27 26 2 0.00 0.00 -2.00 0.00 0.00 57.871 57.752

32 27 2 0.00 0.00 0.00 0.00 2.00 69.550 69.676

33 28 2 0.00 0.00 0.00 0.00 0.00 73.325 71.021

25 29 2 0.00 -2.00 0.00 0.00 0.00 71.340 72.266

23 30 2 -2.00 0.00 0.00 0.00 0.00 57.760 58.335

26 31 2 0.00 2.00 0.00 0.00 0.00 60.721 61.866

28 32 2 0.00 0.00 2.00 0.00 0.00 73.920 76.110

24 33 2 2.00 0.00 0.00 0.00 0.00 63.372 64.867


13

into blocks 1 and 2 for initial soil temperature of 15 and 30 °C, respectively. All experiments were per-formed in duplicate, which have deviation from ±2.3 to ±10.2%.

For the prediction of the pollutant removal efficiency, the following quadratic polynomial res-ponse was used (Eq. (24)):

β β β β ε−

= = = == + + + +

≠

1

20

1 1 1 2,

k k k k

i i ii i ij i ji i i j

Y x x x x

i j (24)

where Y (predicted removal efficiency) is represented as a function of xi and xj (coded independent vari-ables of i and j, respectively) and xixj (coded inter-action term). ε is an additive random error. β0 (inter-cept term), βi, βii and βij (the linear, quadratic and interaction effects in the prediction model, respect-ively) are determined from data analysis. k represents the number of input variables. i and j are index num-bers (i<j for the interaction term) [28].


In order to predict the removal efficiency as a function of the main factors and their interactions, a mathematical model (Eq. (25)) was made in coded values based on Eq. (24):

= + − + −

− + − − +

+ + − − ++ − + ++ − − + −− −

1 2 3

2 24 5 1 2

2 2 23 4 5 1 2

1 3 1 4 1 5

2 3 2 4 2 5

3 4

71.0207 1.6330 2.5999 4.5896

3.7099 1.8454 2.3550 0.9887

1.0225 0.1713 1.2855 2.14241.7531 0.5424 2.11060.7463 1.6043 0.78130.6687 1

Y x x x

x x x x

x x x x xx x x x x xx x x x x xx x −3 5 4 5.9792 0.8287x x x x

(25)

Analysis of variance

Analysis of variance based on the results from Table 1, is summarized in Table 2. Because the sig-nificance level (α) is 0.05 with a specified degree of freedom (DF), P-value smaller than 0.05 shows that the factor is effective on the removal efficiency [29].

According to Table 2, P-value of blocks (0.069), indicates that initial soil temperature is not an effect-ive factor in the studied interval. Also, the lack of fit is not significant (P-value > 0.05) around the prediction model (Eq. (25)) with a suitable correlation coefficient (R2 = 0.9851). Moreover, P-values represent that the main factors (linear and square) with their interactions are effective on the removal efficiency (P-value < < 0.05).

Table 2 shows P and t values for the estimated regression coefficients. This table indicates that all 5

main factors and interaction effects of H2O2/nFe3O4, H2O2/NaOH, nFe3O4/magnetic field, ethanol/magnetic field, and NaOH/nFe3O4 are significant in the model.

The Pareto graph analysis (Figure 2) resulted from Eq. (26) shows the factors and interaction effects on the removal efficiency:

ββ

= × ≠

2

2 100, 0ii

i

P i (26)

In order to obtain a simpler empirical model, insignificant terms are omitted from Eq. (25) and the final model in coded form is made (Eq. (27)):

= + − ++ − + −

− − + −

− − + ++ − −

1 2

3 4 5

2 2 21 2 3

25 1 2 1 3

1 5 2 4 3 5

71.0207 1.6330 2.59994.5896 3.7099 1.8454

2.3550 0.9887 1.0225

1.2855 2.1424 1.75312.1106 1.6043 1.9792

Y x xx x x

x x x

x x x x xx x x x x x

(27)

The main factors and their interaction effects

The effects of operational parameters on 2-meth-ylpropane-2-thiol degradation by modified Fenton treatment are shown in Figure 3.

As demonstrated in Figure 3a, the removal effi-ciency of the 2-methylpropane-2-thiol from the sample soil rises as Fe3O4 nanoparticles-to-soil ratio rises in constant amounts of the other main parameters ([H2O2]0 = 5%, H = 4900 Oe, [NaOH]0 = 10%, (Et:S)0 = = 15 mass%). This figure shows that increasing the nFe3O4 up to a certain level (0.08 mass% to soil ratio) leads to higher 2-methylpropane-2-thiol degradation. But the thiol removal decreases with the additional nFe3O4 due to the side and propagation reactions (Eqs. (3), (4) and (9)). The use of Fe3O4 nanoparticles as a catalyst which could degrade hydrocarbons in the wide range of pH [30], react with NaOH and H2O2 to form powerful radicals.

Figure 3b shows that the magnetic field supplied for nFe3O4 recovery, decreases the removal effici-ency. The magnetic field could be having influence on the water and soil’s physical and chemical properties, such as removing excess soluble salts, dissolving slightly soluble components such as sulfates, phosphates and carbonates, and decreasing pH values of the soil [31]. It was reported that sulfate could have a negative effect on the Fenton process efficiency by forming an iron complex such as Fe2+

4Fe3+2(OH)12SO4·nH2O which removes iron ions

[32]. Moreover, phosphate and carbonate are hydro-xyl radical scavengers [33,34]. Also, phosphate can form iron phosphate salts which are insoluble [35]. Despite the negative effect of sulfates, phosphates and


14

Table 2. Analysis of variance (ANOVA) table for 2-methylpropane-2-thiol removal from soil

Parameter DF SS MS F-value P-value

Blocks 1 15.72 15.718 4.06 0.069

Regression 20 1733.69 86.685 22.37 0.000

Linear 5 1143.83 228.766 59.03 0.000

Square 5 250.59 50.118 12.93 0.000

Interaction 10 339.27 33.927 8.75 0.001

Residual Error 11 42.63 3.875

Lack-of-Fit 6 28.58 4.764 1.70 0.289

Pure error 5 14.05 2.809

Total 32 1792.04 R2 = 97.62%

Terms removal of 2-methylpropane-2-thiol

Term Coefficient t-Value P-value

β0 70.0207 96.061 0.000

β1 1.6330 4.064 0.002

β2 -2.5999 -6.470 0.000

β3 4.5896 11.421 0.000

β4 -3.7099 -9.232 0.000

β5 1.8454 4.592 0.001

β11 -2.3550 -6.576 0.000

β22 -0.9887 -2.761 0.019

β33 -1.0225 -2.855 0.016

β44 0.1713 0.478 0.642

β55 -1.2588 -3.515 0.005

β12 -2.1424 -4.353 0.001

β13 1.7531 3.562 0.004

β14 -0.5424 -1.102 0.294

β15 2.1106 4.289 0.001

β23 0.7463 1.516 0.158

β24 -1.6043 -3.260 0.008

β25 0.7813 1.587 0.141

β34 -0.6687 -1.359 0.201

β35 -1.9792 -4.022 0.002

β45 0.8287 1.684 0.120

Figure 2. Pareto graphic analysis of the main factors and interactions.


15

Figure 3. a) The effect of nFe3O4 to soil ratio (mass%); b) the effect of magnetic field intensity of H (Oe); c) the effect of effect of initial

H2O2 concentration (%) on 2-methylpropane-2-thiol removal efficiency.

carbonates, pH reduction leads to increased effici-ency. Moreover, the static magnetic field changes the hydrogen bond network, disturbing liquid/gas inter-face from the air nanobubbles in the water which could increase mass transfer between the reactants [36].

With respect to magnetite attraction in the mag-netic field, this catalyst collected near the reactor side surface and leads to bad distribution in the soil media. Also, ethanol, water, and hydrogen peroxide as dia-magnetic materials are repelled by a magnetic field due to an opposite direction induced magnetic field creation, causing a repulsive force. On the other hand, the magnetic field helps to slightly move react-ant molecules. Therefore, due to the presence of a magnetic field, reactant movement, lowering pH (slightly) and disturbing liquid/gas interface of nano-bubbles have a positive effect, and magnetite attract-ion in the magnetic field and dissolving sulfate and carbonates have a negative effect on the removal effi-ciency, which is in opposition. When constant amounts of the other main parameters are applied at the center point, the negative role of the magnetic field is domi-nant.

Figure 3c illustrates that the contaminant deg-radation enhanced with the hydrogen peroxide con-centration (with constant other factors: H = 4900 Oe, [NaOH]0 = 10%, (nFe3O4:S)0 = 0.08 mass% and (Et:S)0 = 15 mass%. As shown in this figure, with further increasing of H2O2 concentration, the contam-inant degradation efficiency slope decreases. This result is due to propagation reactions and scavenging which generate less reactive reactants such as HOO• and OH- based on the reactions (3) and (5)–(7) [37].

When initial organic matter such as ethanol con-centration increases, 2-methylpropane-2-thiol dis-solved in ethanol from the sample soil particles and the pollutant is more exposed to the oxidizing agent. On the other hand, some possible zones of accumul-ated ethanol are created which hinders the contact between the target organic pollutants and lead to less accessibility for oxidation remediation. Moreover,

ethanol is a hydroxyl radical scavenger [38-40] and a portion of hydroxyl radical forms react with ethanol instead of 2-methylpropane-2-thiol, based on react-ions 28-30 [41,42]. These two phenomena will lead to two conflicting results that will be in competition. Fig-ure 4a, shows that the addition of ethanol reduces the removal efficiency (with decreasing in its slope):

CH3CH2OH+OH●→H2O+CH3CH2O (28)

CH3CH2OH+OH●→H2O+CH3CHOH (29)

CH3CH2OH+OH●→H2O+CH2CH2OH (30)

Figure 4. a) The effect of ethanol-to-soil ratio (mass%); b) the

effect of initial NaOH concentration on 2-methylpropane-2-thiol removal efficiency.

As demonstrated in Figure 4b, the removal effi-ciency of 2-methylpropane-2-thiol from polluted soil enhances with NaOH level up to an optimum concen-tration (about 12.5%). As NaOH concentration inc-reases, OH− reacts with the thiol due to reactions 17 to 20 on the one hand, and on the other hand, inc-reases the pH of the slurry. Increase in pH leads to lower efficiency of H2O2 oxidation power [43]. It should be noted that an acidic pH is favorable in order to keep the soluble form of Fe(II) for increasing effici-ency of traditional Fenton treatment method [44]. These two opposite results are in competition, therefore an optimum concentration of NaOH should be used to maximize 2-methylpropane-2-thiol degradation.

According to Table 2 and the mathematical model (Eq. (27)), an interaction between hydrogen


16

peroxide concentration and nFe3O4 to the sample soil ratio is an effective factor for the contaminant degrad-ation. As shown in Figure 5a, increasing hydrogen peroxide concentration or nFe3O4:soil ratio raising up to an optimum concentration could raise the degrad-ation efficiency due to OH generation which are very powerful oxidizing radicals [45], but the greater con-centration of nFe3O4 drops the removal efficiency due to a series of side and propagation reactions like (5)– -(8) and (15) [37]. These phenomena will not occur for H2O2, because this oxidation agent remediates 2-meth-ylpropane-2-thiol, ethanol and other organic matter in soil media.

Table 2 illustrates that nFe3O4 and the magnetic field have an effective interaction which decreases the contaminant degradation. As indicated in Figure 5b, at low nFe3O4 levels, the pollutant degradation rises as the magnetic field intensity increases. Magnetite nanoparticles in the presence of the magnetic field have positive and negative effects on the removal efficiency.

As the first positive effect, the magnetic field lowers the pH of the soil which increases Fenton rem-ediation efficiency. Also, the magnetic field could change the hydrogen bond network, disturbing liquid/ /gas interface from the air nanobubbles in the water which could increase mass transfer between the

reactants [36] which is the second positive effect. The nFe3O4/magnetic field tends to distribute magnetite nanoparticles uniformly injected at the center of the reactor and remove magnetite nanoparticles in the accumulated ethanol due to the magnetic properties of nFe3O4 and diamagnetic properties of ethanol, as a third positive effect. But with further increase in mag-netic field power, especially with higher magnetite concentration, the negative effects become more powerful. As a first negative effect, excess magnetic field intensity leads to making nFe3O4 and soil par-ticles aggregated colonies. Therefore, lowering the reactant contact area and poor distribution of magne-tite decreases the removal efficiency. Moreover, in the presence of the magnetic field, water solubility increases and sulfates, carbonates and phosphates tend to emigrate from the soil particle to water. As illustrated in Figure 3b, sulfates, carbonates and phosphates could have a negative effect on the Fen-ton process based on the removing of iron ions by complex formation, radical scavenging and iron phos-phate participation, respectively [32,35], which is another negative effect. Therefore, due to the oppo-site effects of the magnetic field intensity and nFe3O4, they should be at the optimum level.

Figure 5c and Table 2 show that NaOH and nFe3O4 interaction has a positive role in the removal

Figure 5. Counter plot of 2-methylpropane-2-thiol removal efficiency as a function of: a) H2O2 and nFe3O4 interaction; b) magnetic field and nFe3O4; c) nFe3O4 and NaOH; d) magnetic field and ethanol in constant amounts of the other main parameters at the center point.


17

efficiency. Magnetite nanoparticles in couple with NaOH make a series of catalytic reactions as follows [20]:

OH¯+RSH→RS¯+H2O (31) 2Fe2++O2→2Fe3++O2

2¯ (32)

RS¯+Fe3+→Fe2++RS● (33)

2RS●→RSSR (34)

O22¯+H2O→2OH¯+1/2O2 (35)

In this catalyzed system, as a new method, the initial ionization of 2-methylpropane-2-thiol occurs due to its reaction with sodium hydroxide. In the pre-sence of Fe3+, the mercaptide ion changes to mercap-tide radical and then forms disulfide. Peroxidess are generated from reaction (32), leading to the regener-ation of the hydroxyl ion, based on reaction (35).

As represented in Figure 5c, increase in NaOH and nFe3O4 levels up to an optimum concentration enhances 2-methylpropane-2-thiol removal, but further increase in these two reactants reduces the removal efficiency. Higher NaOH concentration leads to rising pH which decreases Fenton reaction efficiency [46]. Moreover, sodium peroxide could react with Fe2+ and Fe3+ and Fe3O4 [47], Fe(OH)3 and Fe(OH)2 form. By following the reactions (33)-(35), Fe(OH)3 and Fe(OH)2 participate [48] and Fe3O4 cores grew and became larger [49,50]:

Fe2++2Fe3++8OH¯→Fe3O4+H2O (36)

Fe2++2OH¯→Fe(OH)2 green participate (37)

Fe3++3OH-→ Fe(OH)3 brown participate (38)

Comparing reactions (1)-(20) and (31)-(38), it is concluded that in the catalyst system of H2O2/NaOH/ /nFe3O4, the couples of NaOH/nFe3O4 and H2O2/ /nFe3O4 attack to degrade thiol by OH¯/Fe3+ and hyd-roxyl radical, respectively. Therefore, the addition of Fe3O4 nanoparticles has a positive effect on the catal-ytic system at the optimum level. But raising nFe3O4

concentration greater than the optimum level has a negative effect on the removal efficiency in a different mechanism. Excess addition of magnetite in NaOH/ /nFe3O4 usage decreases the removal efficiency due to the participation of Fe2+ and Fe3+, but in H2O2/ /nFe3O4 system, hydroxyl radical scavenging lowers 2-methylpropane-2-thiol removal efficiency.

Table 2 demonstrates that interaction between ethanol and the magnetic field has a negative effect on the removal efficiency. Figure 5d shows that in low magnetic field intensity, ethanol addition decreases the removal efficiency. As the magnetic field becomes more powerful, the removal efficiency drops with a higher slope. Ethanol is a diamagnetic material and

its uniformity in the slurry is affected in the magnetic field. Moreover, the magnetic field enhances the hydrogen bonding ability between water, ethanol, and water-ethanol molecules. These phenomena could decrease the self-diffusion coefficient of ethanol and water which decrease the mobility of the water and ethanol molecules [51]. Moreover, an increase in hyd-rogen bonds leads to an increase in the ordered structure of ethanol and water formed around hydro-phobic molecules and colloids [52] and also enhances the dissolution rate of the oxygen and other gases such and thiol vapors in water and ethanol.

The P value of β35 (P = 0.002) from Table 2 shows that H2O2 and NaOH interaction is effective on the removal efficiency. Surprisingly, Figure 6 illus-trates that sodium hydroxide/hydrogen peroxide inter-action has a negative role in the removal efficiency. The mixture of H2O2 and NaOH produces a strong oxidizer with exothermic mist from the onset of high temperatures and heat (Eq. (39)). This reaction result-ing in a mist could decrease the contact area between reagents [53]:

2NaOH+H2O2+nH2O→Na2O2⋅8H2O+(n-6)H2O (39)

Figure 6. Counter plot of 2-methylpropane-2-thiol removal effi-ciency as a function of NaOH and H2O2, in constant amounts of

the other main parameters at the center point.

Furthermore, NaOH addition leads to rising of slurry pH which is not favored in the Fenton process [43,54]. But NaOH addition to the process has some advantages, such as minimizing 2-methylpropane-2- -thiol vaporization and increasing the removal effici-ency due to the positive effect of NaOH initial con-centration and NaOH/nFe3O4 interaction. The com-parison of NaOH addition effects is represented as follows: NaOH/nFe3O4 (+)>NaOH/H2O2 (-)>NaOH(+).

Figure 7 shows the effect of H2O2 or/and NaOH addition to the sample soil with 10 mass% 2-meth-ylpropane-2-thiol polluted soil. As represented in this figure, H2O2 and NaOH addition to the remediation


18

process, minimize 2-methylpropane-2-thiol vapor-ization at 24 °C, pH 6.6, and contaminant-to-soil ratio of 10% by weight.

Figure 7. 2-methylpropane-2-thiol vaporization in ppm from soil

(▲) H2O2, (●) NaOH and (■) H2O2 and NaOH addition.

The optimum levels for 2-methylpropane-2-thiol degradation

Prediction models resulted from ANOVA show that the maximum removal efficiency is 91.50% at the optimum conditions in the studied intervals (Table 3). The optimum conditions suggest that the concen-trations of nFe3O4 and NaOH must be at a certain level in the presence of maximum H2O2 concentration and minimum ethanol concentration for remediation in the studied intervals. Also, magnetic field intensity should be applied at a certain level.

The verification experiment shows that the max-imum removal efficiency is 92.02% according to the

optimum conditions. This efficiency higher compared to some methods like sonication and a modified Fen-ton process to degrade 2-methylpropane-2-thiol. As represented in Figure 8, a magno-modified Fenton process has a higher removal efficiency compared to sonication and modified Fenton, minimizing the con-taminant vaporization and Fe3O4 recovery simultane-ously [21,22,55].

CONCLUSION

2-methylpropane-2-thiol has several human, animal and social consequences and its removal from polluted soil is important to study. Modeling and opti-mization of this pollutant removal efficiency from pol-luted soil using NaOH coupled with the nano-modified Fenton process were the aims of this study. Pareto analysis and analysis of variance indicated that all investigated factors were effective on the removal effi-ciency. Moreover, the interactions of H2O2/nFe3O4 and nFe3O4/NaOH were significant with a positive role, and H2O2/NaOH, nFe3O4/magnetic field and ethanol/magnetic field with a negative role. Surpris-ingly, with nFe3O4, NaOH and magnetic field intensity at optimum levels, 2-methylpropane-2-thiol degrad-ation enhances and at higher levels, the degradation efficiency decreases due to scavenging, side react-ions, pH rising and attracting/repelling of the reactant. Moreover, ethanol concentration should be at a mini-mum level and hydrogen peroxide should be applied at a maximum level in the studied interval. The impact of soil temperature was not significant, based on the experiment results and the ANOVA table.

Table 3. Optimal values of the main independent factors for the maximum removal efficiency

(nFe3O4:S)0

mass% H

Oe [H2O2]0

% (Et:S)0

mass% [NaOH]0

mass%

Removal efficiency (%R)

Predicted Observed

0.11 3585 9.00 5.00 8.64 91.50 92.02

Figure 8. 2-methylpropane-2-thiol remediation rate from soil (▲) sonication, (■) modified Fenton, and (●) magno-modified Fenton.


19

This method can be used for increasing the removal efficiency and minimizing the vaporization of similar organosulfur compounds, because of the posi-tive interaction between reactants.

REFERENCES

[1] R. Karthikeyan, S.L.L. Hutchinson, L.E. Erickson, J. Biorem. Biodegrad. 3(2012) 2-7

[2] F. Pardo, J.M. Rosas, A. Santos, A. Romero, J. Chem. Technol. Biotechnol. 90 (2015) 754-764

[3] L. Wang, J. Yang, Y. Li, J. Lv, G. Li, J. Chem. Technol. Biotechnol. 92 (2017) 552-561

[4] G. Dalgic, F.I. Turkdogan, K. Yetilmezsoy, E. Kocak, Chem. Ind. Chem. Eng. Q. 23 (2017) 177-186

[5] A.C. Kumoro, R. Ratnawati, D.S. Retnowati, Asia‐Pac. J. Chem. Eng. 10 (2015) 648-658

[6] M. Pereira, L. Oliveira, E. Murad, Clay Miner. 47 (2012) 285-302

[7] S.-P. Sun, A.T. Lemley, J. Mol. Catal. 349(2011) 71-79

[8] F. Vicente, J. Rosas, A. Santos, A. Romero, Chem. Eng. J. 172 (2011) 689-697

[9] M. Lu, Z. Zhang, W. Qiao, Y. Guan, M. Xiao, C. Peng, J. Hazard. Mater. 179 (2010) 604-611

[10] S.R. Kanel, B. Neppolian, H. Choi, J-W. Yang, Soil Sedi-ment Contam. 12 (2003) 101-117

[11] M. Danish, X. Gu, S. Lu, M. Naqvi, Environ. Sci. Pollut. Res. 23 (2016) 13298-307

[12] N. Ezzatahmadi, G.A. Ayoko, G.J. Millar, R. Speight, C. Yan, J.H. Li, S.Z. Li, J.X. Zhu, Y.F. Xi, Chem. Eng. J. 312 (2017) 336-350

[13] Q. Han, Z. Wang, J. Xia, S. Chen, X. Zhang, M. Ding, Talanta 101 (2012) 388-395

[14] S. Gan, H.K. Ng, Chem. Eng. J. 213 (2012) 295-317

[15] S. Gan, H.K. Ng, Chem. Eng. J. 180 (2012) 1-8

[16] A. Karimi, F. Mahdizadeh, M. Eskandarian, Chem. Ind. Chem. Eng. Q. 18 (2012) 89-94

[17] L.A. Van Bergen, G. Roos, F. De Proft, J. Phys. Chem., A 118 (2014) 6078-84

[18] T.J. Wallace, A. Schriesheim, J. Org. Chem. 27 (1962) 1514-1516

[19] L.D. Rollmann, J. Am. Chem. Soc. 97 (1975) 2132-2136

[20] T. Wallace, A. Schriesheim, H. Hurwitz, M. Glaser, Ind. Eng. Chem. Proc. Des. Dev. 3 (1964) 237-241

[21] P. Roohi, E. Fatehifar, R. Alizadeh, Asia‐Pac. J. Chem. Eng. 11 (2016) 743-56

[22] P. Roohi, E. Fatehifar, R. Alizadeh, Iran. J. Environ. Technol. 2 (2016) 11-26

[23] P. Rouhi, Phys. Chem. Res. 7 (2019) 53-64

[24] A.M. Dugandzic, A.V. Tomasevic, D.M. Dabic, N.Ž. Šekuljica, M.M. Radisic, S.D. Petrovic, D.Ž. Mijin, Chem. Ind. Chem. Eng. Q. 24 (2018) 201-208

[25] N. Mokhtarani, S.Y. Yasrobi, H. Ganjidoust, Ozone: Sci. Eng. 37 (2015) 279-286

[26] M.L. Jackson, C.H. Lim, L.W. Zelazny, Method of soil analysis, Part 1, American Society of Agronomy and Academic Press, New York, 2003, p. 113

[27] D. Nelson, L.E. Sommers, Methods of soil analysis, Part 2, American Society of Agronomy and Academic Press, New York, 1982, p. 539

[28] A. Aghaeinejad-Meybodi, A. Ebadi, S. Shafiei, A. Khataee, M. Rostampour, Environ. Technol. 36 (2015) 1477-1488

[29] M. Hosseini, A.B. Khoshfetrat, E. Sahraei, S. Ebrahimi, Process Saf. Environ. Prot. 92 (2014) 869-878

[30] W. Wang, Q. Mao, H. He, M. Zhou, Water Sci. Technol. 68 (2013) 2367-2373

[31] V. Zlotopolski, Int. Soil Water Conserv. Res. 5 (2017) 253-257

[32] K. Henke, Environmental Chemistry, health threats and waste treatment, Wiley Online Library, 2009

[33] R.J. Watts, A.L. Teel, J. Environ. Eng. 131 (2005) 612-622

[34] E. Lipczynska-Kochany, G. Sprah, S. Harms, Chemo-sphere 30 (1995) 9-20

[35] M.A. Tarr. 2003, In Chemical Degradation Methods for Wastes and Pollutants, CRC Press, Boca Raton, FL, pp. 159-190

[36] A. Szcześ, E. Chibowski, L. Hołysz, P. Rafalski, Chem. Eng. Process. 50 (2011) 124-127

[37] A. Karimi, F. Mahdizadeh, D. Salari, F. Vahabzadeh, A. Khataee, Chem. Ind. Chem. Eng. Q. 18 (2012) 431-439

[38] Z. Kozmér, E. Takács, L. Wojnárovits, T. Alapi, K. Her-nádi, A. Dombi, Rad. Phys. Chem. 124 (2016) 52-57

[39] W. Zhao, J. Zhang, X. Zhu, M. Zhang, J. Tang, M. Tan, Y. Wang, Appl. Catal., B 144 (2014) 468-477

[40] M. Nie, Q. Wang, G. Qiu, Ultrason. Sonochem. 15 (2008) 222-226

[41] D. Grosjean, J. Brazil. Chem. Soc. 8 (1997) 433-442

[42] C. Burchill, I. Ginns, Canadian J. Chem. 48 (1970) 2628- -2632

[43] H. Mosmeri, F. Gholami, M. Shavandi, E. Alaie, S.M.M. Dastgheib, Environ. Sci. Pollut. Res. 25 (2018) 31051- -31061

[44] M. Usman, O. Tascone, V. Rybnikova, P. Faure, K. Hanna, Environ. Sci. Pollut. Res. 24 (2017) 14748-14757

[45] R. Baciocchi, Water, Air, Soil Pollut. 224 (2013) 1-11

[46] H.-F. Lu, H.-F. Chen, C.-L. Kao, I. Chao, H.-Y. Chen, Phys. Chem. Chem. Phys. 20 (2018) 22890-228901

[47] G. Gnanaprakash, S. Mahadevan, T. Jayakumar, P. Kalyanasundaram, J. Philip, B. Raj, Mater. Chem. Phys. 103 (2007) 168-175

[48] H. Tamura, Corrosion Sci. 50 (2008) 1872-1883

[49] A. Ebrahiminezhad, Y. Ghasemi, SR. Amini, J. Barar, S. Davaran, Bull. Korean Chem. Soc. 33 (2012) 3957-3962

[50] F. Balacianu, A. Nechifor, R. Bartos, S. Voicu, G. Nechi-for, Optoelectron. Adv. Mater., Rapid Commun. 3 (2009) 219-222

[51] K.-T. Chang, C.-I. Weng, J. Appl. phys. 100 (2006) 1-6

[52] S. Ozeki, C. Wakai, S. Ono, J. Phys. Chem. 95 (1991) 10557-10559


20

[53] J. Balej, Collect. Czech. Chem. Commun. 36 (1971) 426- -438

[54] Y. Qian, X. Zhou, Y. Zhang, W. Zhang, J. Chen, Chemo-sphere 91 (2013) 717-723

[55] P. Roohi, E. Fatehifar, R. Alizadeh, Adv. Environ. Technol. 3 (2016) 137-146.

PEJMAN ROOHI

ESMAEIL FATEHIFAR

Environmental Engineering Research Center (EERC), Faculty of Chemical

Engineering, Sahand University of Technology, Sahand New Town,

Tabriz, Iran

NAUČNI RAD

POBOLJŠANI TRETMAN ZEMLJIŠTA KONTAMINI-RANOG 2-METILPROPAN-2-TIOLOM PRIMENOM MAGNO-MODIFIKOVANOG FENTONOVOG PROCESA

U ovom radu je istraživan nano-magno modifikovani Fentonov procesa za remedijaciju zemljišta zagađenog 2-metilpropan-2-tiolom. U ovom dvostepenom oksidacionom pro-cesu korišćen je sistem sa nanokatalizatorom koji uključuje Fe2+/Fe3+ etanol u magnet-nom polju. Analiza podataka dobijena analizom varijanse je pokazala da H2O2, NaOH, Fe3O4-nanočestice, koncentracija etanola i intenzitet magnetnog polja utiču na efikas-nosti uklanjanja 2-metilpropan-2-tiola (26,99, 4,36, 3,42, 17,63 i 8,66 %, redom). Šta-više, sistemi H2O2/nanoFe3O4 i NaOH/nanoFe3O4 su imali pozitivnu ulogu dok su nano-Fe3O4/etanol, H2O2/NaOH i magnetno polje/nanoFe3O4 imali negativnu ulogu. Preporu-čeni su maksimalna početna koncentracija H2O2, minimalna koncentracija etanola i određeni nivoi početne koncentracije NaOH (8,64 mas.%), odnos nanočestice Fe3O4/ /zemljište i intenzitet magnetnog polja (0,11 mas.% i 3585 Oe, redom) koji obezbeđuju maksimalnu efikasnost remedijacije nečistoća u ispitivanom intervalu (%R = 92,02). Magno-modifikovani Fenton postupak može se koristiti i za povećanje efikasnost ukla-njanja i minimiziranje isparavanja sličnog organo sumpora. Eksperimentalni rezultati su pokazali da dodavanje rastvarača za razblaživanje tiolnog zagađivača može smanjiti efikasnost uklanjanja dok je kombinovanje Fe3O4 i magnetnog polja dobar način za nje-govo odvajanje i ponovnu upotrebu nanočestica.

Ključne reči: Napredna oksidacija, remedijacija životne sredine, matematičko modelovanje, optimizacija, isparljiva organska jedinjenja.




21

KARLA RAPHAELA BRAGA DE MELO1

GABRIELA CANTARELLI LOPES1

DAYANA DE GUSMÃO COÊLHO2

JOÃO INÁCIO SOLETTI2 1Department of Chemical

Engineering, Federal University of São Carlos, Rodovia Washington

Luís, São Carlos - São Paulo, Brazil

2Laboratory of Separation and Process Optimization Systems,

Technology Center, Federal University of Alagoas, s/n Maceió -

Alagoas - Brazil

SCIENTIFIC PAPER

UDC 66.063.6:665.353.4: 665.75

LIQUID-LIQUID EQUILIBRIUM FOR SYSTEMS COMPOSED BY BIODIESEL FROM CATOLÉ OIL (Syagrus cearensis), METHANOL AND GLYCEROL

Article Highlights • Characterization of Catolé oil and biodiesel according to Brazilian standards • Liquid-liquid equilibrium of biodiesel-methanol-glycerin systems at three temperatures• Validation of tie-line data through Othmer-Tobias and Hand correlations • Determination of distribution and selectivity coefficients Abstract

The biodiesel inclusion in the Brazilian energy matrix still faces challenges due to the lack of diversity of raw material and the high costs associated with the stages of production, purification and phase separation of the biofuel. In this work, the potential of Catolé (Syagrus cearensis) was evaluated as an alternat-ive source for biodiesel production, through methyl transesterification by alka-line catalysis. Given the absence of reported data on this oilseed, the aim of the present paper is to characterize the oil and biodiesel in terms of density, kinematic viscosity, and acid value. A liquid-liquid equilibrium (LLE) study of Catolé biodiesel-methanol-glycerin systems at 298.15, 308.15 and 323.15 K at atmospheric pressure was also carried out. Results showed that Catolé seed oil is a viable alternative for biodiesel production, with parameters found in accordance with the quality standards and an ester yield of 97.1%. Ternary diagrams showed that the solubility of the components lowers at low tempe-ratures. Tie-lines evidenced higher affinity of methanol with the glycerin-rich phase. Distribution and selectivity coefficients were determined and the Othmer-Tobias and Hand correlations were applied to verify the quality of the experimental data. The determination indices (>0.97) proved the thermodyn-amic consistency of the data.

Keywords: biodiesel, Syagrus cearensis, methanol, liquid-liquid equilib-rium.

The search for alternative sources to fossil fuels favors the development of fuels from renewable raw materials. Among them, biodiesel stands out, which is generally defined as a mixture of mono-alkyl esters of vegetable oils or animal fats. They are produced mainly by transesterification via alkaline catalysis [1].

Correspondence: K.R.B. de Melo, Department of Chemical Eng-ineering, Federal University of São Carlos, Rodovia Washington Luís, km 235 - SP-310, P.O. Box: 13565-905, São Carlos - São Paulo,Brazil. E-mail: [email protected] Paper received: 8 May, 2019 Paper revised: 6 July, 2019 Paper accepted: 12 July, 2019

https://doi.org/10.2298/CICEQ190508021B

However, several works propose alternative routes in order to optimize biodiesel production, such as reactions through acid catalysis, with ionic liquids, under supercritical conditions, enzymatic processes, ultrasonic-assisted transesterification, and hydroes-terification, among others [2-6].

Transesterification is the step of converting oils or fats to alkyl esters due to the reaction with a short-chain alcohol (methanol or ethanol) in the presence of a catalyst [7]. In this reaction, the triglycerides react with alcohol in the mole ratio of 1:3. The reaction mechanism occurs in three stages, in which mono- and di-glycerides are formed as intermediates and, finally, esters of fatty acids and glycerin. In general, this reaction is considered fast and promotes high

K.R.B. DE MELO et al.: LIQUID-LIQUID EQUILIBRIUM FOR SYSTEMS COMPOSED... Chem. Ind. Chem. Eng. Q. 26 (1) 21−29 (2020)

22

conversion to esters, even under moderate condit-ions. However, the effectiveness and feasibility of this process, although relatively simple, is dependent on the characteristics of the raw material used.

According to the Statistical Yearbook 2018 of National Petroleum, Natural Gas and Biofuels Agency (NPA) [8], 71.6% of biodiesel produced in Brazil comes from soybean oil, followed by 16.8% from ani-mal fat, indicating a lack of diversification of the Bra-zilian biodiesel production chain and suggesting a dependence on a single raw material. This scenario encourages the search for alternative sources for bio-diesel production. Several studies have evaluated alternative raw materials considering the country's rich culture and diversity and have shown a great pot-ential for energetic production from several olea-ginous plants. The exploration of these cultures not only represents an important addition to the Brazilian bioenergy matrix but could also boost the economic growth of less developed areas [9-12].

Syagrus cearensis, popularly known as Catolé and abundant in the Brazilian semi-arid and Atlantic Forest, is a promising source for biofuel production. This palm tree has a large annual production of almonds with high lipid content. The fat found in the almonds is formed predominantly by saturated fatty acids of low molecular weight and high resistance to oxidative rancidification − factors that ensure bio-diesel quality and durability. The Catolé seed oil has been investigated for different applications related mainly to its use as a food source given its nutritional value [13-15].

The biodiesel quality and viability depend not only on the physical and chemical properties of the raw material but mainly on the purification process. Following transesterification, a biphasic system com-posed predominantly of esters and glycerol is formed. In addition to water, excess alcohol, tri-, di- and unconverted monoacylglycerols and unwanted com-ponents are distributed between the phases and may compromise engine performance and power. Once these two immiscible phases reach steady state, a liquid-liquid equilibrium (LLE) study is essential to understand the phase separation, measure the com-position of each phase, and verify which factors influ-ence the equilibrium. These are key points to be det-ermined, seeking to optimize operational conditions and, thus, reach a cost-effective production process [16,17].

1. Several liquid-liquid equilibrium systems composed of biodiesel-methanol-glycerin have been studied (Table 1), mainly focusing on the effect of temperature on solubility curves and the influence of

oleaginous and alcohol-type properties on the separ-ation process. In these systems, methanol exhibits greater affinity toward the glycerin-rich phase and temperature variation significantly influences the solu-bility of the components of equilibrium systems. How-ever, there is no data in the literature reported for systems containing biodiesel from Catolé, which encouraged the development of the present study.

Table 1. Raw material and temperature applied in studies of systems composed of biodiesel, methanol, and glycerin

Raw material Temperature, K

Soybean [18] 298, 323

Corn and frying oil [19] 293.2, 303.2, 313.2

Canola and Sunflower [20] 303.15, 313.15, 323.15

Brazilian nut [21] 303.15, 323.15

Fish oil [22] 298.15, 313.15, 328.15

2. In this context, this study evaluates the Cat-olé potential as an oleaginous source for biodiesel production. The aim is to assess the viability of bio-diesel production from the transesterification reaction via alkaline catalysis of Catolé oil (S. cearensis). The ternary liquid-liquid equilibrium system containing bio-diesel-methanol-glycerin was also investigated, seek-ing to evaluate and promote its potential as a bio-diesel source. To this end, ternary diagrams at 298.15, 308.15 and 323.15 K were determined at atmospheric pressure. The tie-lines data were used in the study of the selectivity and component distribution coefficient that were validated by the Othmer-Tobias and Hand correlations.


Raw material

This study used the fruits of Syagrus cearensis palm tree (Figure 1), collected in Barra de São Miguel, state of Alagoas, Brazil. The fruits were dried in a kiln at 343.15 K for 24 h and ground to obtain the

Figure 1. Catolé coconut (Syagrus cearensis) in different stages

of maturation.

K.R.B. DE MELO et al.: LIQUID-LIQUID EQUILIBRIUM FOR SYSTEMS COMPOSED… Chem. Ind. Chem. Eng. Q. 26 (1) 21−29 (2020)

23

almonds. The oil was extracted from the almonds by mechanical extraction, purified by vacuum filtration and oven dried for 12 h at 343.15 K for the removal of impurities and moisture.

Catolé oil and biodiesel characterization

The fatty acid profile was obtained by gas chro-matography. In addition, the acid value, density and kinematic viscosity of Catolé oil and biodiesel were measured. The acid values were determined accord-ing to the AOCS (American Oil Chemists’ Society) standard Cd3d-63, using the colorimetric method with phenolphthalein as an indicator. The density mea-surements were taken at the temperature of 293.15 K using a digital densimeter, according to ASTM (Amer-ican Society for Testing and Materials) D4052. The kinematic viscosity was determined using the method-ology of ASTM 445 with an Ostwald viscometer with a capillary of 200 mm2/s.

Biodiesel production

Esterification Since Catolé oil presented a high acid value

content during the characterization stage, the ester-ification reaction was used as a pre-treatment for the biodiesel production.

Methyl alcohol and hydrochloric acid (HCl) were used in the esterification reaction, with the latter as an acid catalyst. The fixed parameters were: the mole ratio of oil/alcohol 1:7, 1% by mass of catalyst, the temperature of 338.15 K and a reaction time of 2 h. After the reaction, the methyl ester was washed with 0.01 M NaOH solution and distilled water at 343.15 K until it reached neutral pH. To eliminate moisture, the sample was placed in an oven at 343.15 K.

Transesterification The transesterification reaction occurred in the

presence of an alkaline catalyst (NaOH) and meth-anol as the reagent alcohol. This reaction occurred at 323.15 K for 1 h, with a 1:7 oil/alcohol mole ratio and catalyst amount of 0.75% (mass basis). The biodiesel was purified from the washing process with 0.01 M hydrochloric acid solution with pH 2 and distilled water until it reached neutral pH. Magnesium sulfate was used as a desiccant to remove moisture. The sample was allowed to stand for 24 h and then it was subjected to vacuum filtration.

Chromatographic analysis

Qualitative and quantitative analyses of fatty acid were performed by gas chromatography: Shimadu GC-Plus® gas chromatograph with flame ionization detector (FID) and a 2.2 m column with

injector temperature of 523.15 K, detector tempe-rature of 613.15 K, column temperature of 323.15 K and pressure of 6 kPa. Hydrogen, nitrogen and syn-thetic air were employed as entrainment gases. The internal standard used was glyceryl trioctanoate (tri-caprylin), in the concentration of 0.8 g/10 mL of hex-ane. Biodiesel samples weighed approximately 0.15 g and were diluted in 1 mL of the standard solution (tri-caprylin and hexane). 1 μL of the sample was injected into the chromatograph. Injections were made in duplicate and the yield of esters was calculated according to Eq. (1):

= Tricaprylin s

Tricaprylin s

100m A f

nA m

(1)

where mTricaprylin corresponds to the weight of the inter-nal standard, As - the sum of peak areas referring to the esters, f - the response factor, ATricaprylin - peak area referring to the internal standard, and ms - the sample weight.

ELL data

Binodal curves: Catolé biodiesel-methanol-glycerin The binodal curve was obtained from titration,

for temperatures of 298.15, 308.15 and 323.15 K. For the biodiesel-rich phase, the glycerin P.A. was titrated dropwise with a completely miscible binary mixture consisting of biodiesel and methyl alcohol P.A. with a total mass of approximately 10 g. The same method-ology was used for the glycerin-rich phase for which the biodiesel titration of a binary mixture consisting of glycerin + methyl alcohol PA was performed. Com-ponent masses of the binary mixture were weighed in Gehaka® analytical balance AG200, as well as the initial and final masses of titrants. The diagrams were plotted using OriginPro® software, version 8.0.

Tie-lines: Catolé biodiesel-methanol-glycerin For the determination of the tie-lines, points

were selected below the binodal curve, called mixing points. Samples of approximately 12 g were pre-pared, then stirred for an hour and left to rest until the complete separation of phases. The tie-lines were obtained using the drying scale, at a temperature of 358.15 K. Initially, an aliquot of the upper phase, rich in biodiesel, was weighed to obtain the mass percent-age of methyl alcohol in this phase. Then, the methodology was repeated with the lower glycerin-rich aliquot. This procedure was performed for the other mixing points samples, totaling three tie-lines for each binodal curve constructed.


24


Characterization of Catolé oil

The fatty acid profile of the Catolé oil was obtained by gas chromatography, consisting of 36.1% of lauric acid (12:0), 32% of palmitic acid (16:0) and 24.6% of oleic acid (18:1). Lauric and palmitic acids predominance is a characteristic of palm. However, since it is a vegetable and non-commercial raw mat-erial, this composition may vary depending on the period of fruits harvesting, the seed quality, the stor-age conditions and the way of extracting the oil. Other parameters obtained from the physicochemical char-acterization of Catolé oil are shown in Table 2.

Table 2. Characterization of Catolé oil

Properties Results

Oil content, % 41.0

Acid value, mg KOH g-1 12.00±0.01

Density, kg m-³ 916±1

Kinematic viscosity, mm² s-1 47.448±0.005

Catolé oil shows a high lipid content for bio-diesel production. The Catolé oil kinematic viscosity and density are close to those found for other non- -commercial oleaginous sources used for biodiesel production in literature, as shown in Table 3.

Table 3. Density and kinematic viscosity data of different non-commercial oils

Raw material Density kg m-³

Kinematic viscositymm² s-1

Syagrus cearensisa 916±1 47.448±0.005

Terminalia catappa L. [23] 910 39.8

Moringa oleifera [24] 912 43.4

Sterculia striata [25] 924 53.66 aAuthor data

The acid value is determinant for the biodiesel production process. Since the Catolé oil showed high acidity, 12 mgKOH/g (>3 mgKOH/g), esterification was used as a pretreatment. This is necessary to avoid undesired secondary reactions, such as sapon-ification of free fatty acids promoted by the basic reaction medium. In addition to that, the esterification directly influences the parameters of kinematic visco-sity, purity, and biodiesel quality.

Characterization of Catolé biodiesel

The characterization of the Catolé biodiesel was performed according to the NPA Resolution No. 45, published in 2014. The parameters were found to be

satisfactory and complying with the criteria of the ASTM D and EN 14214 regulatory standards, as shown in Table 4.

Table 4. Characterization of biodiesel from Catolé

Properties Results ASTM D EN

Acid value, mgKOH g-1 0.493±0.020 0.50 0.50

Density, kg m-3 859±1 850–900 860-900

Kinematic viscosity, mm² s-1 3.445±0.010 3.0-6.0 3.50-5.0

Ester content, % 97.10 ± 0.05 96.5 96.5

Results indicate that the esterification and trans-esterification steps employed here were satisfactory for the biodiesel production according to the quality standards. However, the kinematic viscosity and den-sity were lower than those established by the Euro-pean standard of biodiesel EN 14214. Transester-ification converts triglycerides into smaller fatty acid ester molecules, thus reducing kinematic viscosity. The kinematic viscosity of Catolé biodiesel was close to that of methyl biodiesel from oilseeds usually used in fuel production, such as soybean (4.0-4.1 mm2/s) and canola (4.4 mm2/s) [26].

The chromatographic analysis provided a con-version to esters equal to 97.1%, a result higher than the minimum yield established by NPA, which is 96.5%. A predominance of the esters, linoleic (18:2), oleic (18:1) and palmitic (16:0), in proportions of 14, 26.2 and 52%, respectively, was identified in the transesterified oil. A good similarity between Catolé oil and soybean oil [27] (the main source of raw mat-erial used for biodiesel production) in terms of com-position can be noticed. Studies such as one by Sagiroglu et al. [2] demonstrate that oils rich in these acids have higher biodiesel yield, and indicate the quality of the biodiesel to be produced, according to Knothe [28].

ELL data

Binodal curves Binodal curves for each of the biodiesel + meth-

anol + glycerin systems, shown in Figure 2, demon-strate that the systems exhibit well-characterized behavior, with large regions of immiscibility. The bin-odal curves of Catolé biodiesel (1) + methanol (2) + glycerin (3) plotted at temperatures of 298.15 and 308.15 K did not vary significantly regarding the com-ponent’s solubility; however, at 298.15 K, the large region of miscibility observed for 298.15 K indicates that the separation process is favored by low tempe-ratures.

The effect of temperature on ELL is observed with the increase of solubility in ternary mixtures with


25

the increase in temperature. This is demonstrated by the significant decrease in the immiscibility region as a result of the increase in temperature in the system at 323.15 K.

Tie-lines The tie-lines behavior (Figure 3) demonstrates

that methanol is more soluble in the glycerin-rich phase than in the biodiesel-rich phase. Figure 3 shows a steep slope in the tie-lines, indicating a good separation of biodiesel from the other components. The low content of methanol and glycerin in the phase rich in biodiesel after the phase separation rep-resents an advantage of Catolé biodiesel in the indus-trial process. In the separation and purification stages, the biodiesel is easily separated from meth-anol and glycerin, reducing operational costs.

Thermodynamic consistency of experimental data

Othmer-Tobias [29] and Hand [30] correlations (Eqs. (2) and (3)) were applied to evaluate the ther-modynamic consistency of the tie-lines. These correl-ations consist of empirical equations that provide a linear representation of the ternary diagrams tie-lines, in which linearity (indicated by determination indexes

(R2) greater than 0.95) validates the quality of the experimental data:

− −= +II II

II II

1 3 1 1ln( ) ln( )

3 1w wA B

w w (2)

= +II I

II II

2 2ln ' ' ln

3 1w wA Bw w

(3)

Variable w3II in Eqs. (2) and (3) corresponds to the mass fraction of glycerin in the glycerin-rich phase and w1I, is the biodiesel mass fraction in the biodiesel rich phase; w2I and w2II correspond to methyl alcohol mass fractions in the biodiesel and glycerin-rich phases, respectively.

Parameters A, B, A', B' and determination coef-ficients (R2) are presented in Table 5. Othmer-Tobias and Hand correlations showed determination indexes above 0.97, demonstrating the quality of the expe-rimental data.

Distribution and selectivity coefficients

To determine the liquid extraction efficiency in ELL systems, the distribution and selectivity coef-ficients were computed.

Figure 2. Binodal curves for the systems Catolé biodiesel + methanol + glycerin at 298.15 (●), 308.15 (■) and 323.15 K (▲).


26

The distribution coefficient gives the distribution of components between the phases in equilibrium and is as such considered a key parameter. According to Sandler [31], it is defined as the concentration of a given component in each of the phases. Thus, meth-anol (β1) and glycerin (β2) coefficients were defined as the ratio between their respective concentrations in the biodiesel and glycerol phases, according to Eqs. (4) and (5):

β =I

1 II

22

ww

(4)

β =I

2 II

33

ww

(5)

The distribution coefficients indicate high meth-anol concentration in the glycerin-rich phase. The higher affinity of methanol with the glycerin-rich phase is caused by the hydrogen bonds formed between methanol and glycerin. This preference is also evi-denced by the tie-lines slope (Figure 3).

As observed in Figure 4, for lower temperatures (298.15 and 308.15 K), the increase in methanol mass fraction in the global composition results in an inc-rease in the distribution, which indicate that, for this temperature range, the excess of methanol may hinder the separation process. For lower methanol contents, although a decrease is observed at 323.15 K, the distribution coefficients increase with temperature.

Figure 3. Tie-lines for the systems Catolé biodiesel + methanol + glycerin systems at 298.15 (●), 308.15 (■ ▲) and 323.15 K ( ).

Table 5. A, B and R2 for Othmer-Tobias and Hand correlations

System Othmer-Tobias Hand

A B R2 A’ B’ R2

Biodiesel + methanol + glycerin, 298.15 K 1.339 1.117 0.978 1.378 1.051 0.970




27

Figure 4. Distribution coefficient as a function of the methanol

mass fraction in the global composition at ternary systems Catolé biodiesel + methanol + glycerin at 298.15 (●), 308.15 (■)

and 323.15 K (▲).

Another important parameter for the liquid-liquid extraction study is the selectivity coefficient (S), which is used to evaluate the solvent extraction capacity. The higher the value of S, the greater the extraction efficiency and, consequently, the yield related to this process. In biodiesel + methanol + glycerin ternary systems, biodiesel selectivity reveals the relationship between ethanol and glycerin solubilities in the bio-diesel-rich phase. This parameter is given by the ratio between methanol (β1) and glycerin (β2), according to Eq. (6):

ββ

= 1

2

S (6)

The selectivity coefficients obtained here are higher than 1, indicating that methanol is an excellent solvent for glycerin extraction in the biodiesel-rich phase. In Figure 5, it is possible to observe that the selectivity decreases with temperature for a global methanol composition between 0.3 and 0.4, indicating a higher content of solubilized glycerin in the bio-diesel-rich phase. For the ternary systems at 298.15 and 323.15 K, selectivity decreases as the amount of methanol increases. This behavior may be explained by the presence of excess methanol and its higher solubility in the biodiesel-rich phase due to the inc-rease in temperature.

CONCLUSIONS

The lipid content of Catolé almonds and the physicochemical characteristics of its oil suggest a potential for biodiesel production. Characterization

tests indicated that values of density, kinematic vis-cosity, and acid value for biodiesel are in compliance with the quality standards for marketing established by the NPA, reinforcing the potential of this olea-ginous plant and evidencing the quality of the bio-diesel produced. In addition, the methyl trancester-ification of the Catolé oil provided a high conversion to esters (equal to 97.1%), above the minimum per-centage required by the standard.

Binodal curves presented well-characterized behavior. The results show that at 298.15 and 308.15 K there is no significant variation in the solubility of components. At these temperatures, there is a broad two-phase region, which demonstrates that the puri-fication process is most effective at low temperatures. The influence of the temperature on the solubility of components is best observed with the data of the bin-odal curve at 323.15 K. The decrease in the region of immiscibility of the ternary diagram is due to the inc-rease in temperature, which indicates an increase in the solubility of the components and, consequently, a greater difficulty in the spontaneous formation of the phases in systems with higher temperatures.

The tie-lines behavior and the values obtained for the distribution and selectivity coefficients indicate higher solubility of methanol in the glycerin-rich phase for the different systems. Experimental data was validated by Othmer-Tobias and Hand correlations. High determination indexes validate the methodology applied to obtain the tie-lines.

Figure 5. Selectivity coefficient, S, as a function of the methanol mass fraction in the global composition at ternary systems

Catolé biodiesel + methanol + glycerin at 298.15 (●), 308.15 (■) and 323.15 K (▲).


28

REFERENCES

[1] G. Knothe, Prog. Energy Combust. Sci. 36 (2013) 364– –373

[2] A. Sagiroglu, S.S. Isbilir, H. Paluzar, N.M. Toprakkiran, Chem. Ind. Chem. Eng. Q. 17 (2011) 53-58

[3] Y. Tsai, H. Lin, M. Lee, Bioresour. Technol. 145 (2013) 362-369

[4] L. Changxia, Y. Jinlei, G. Huafeng, L. Chunqiao, Biocatal. Biotransform. 34 (2016) 283-290

[5] D. Ozçimen, M.O. Gulyurt, B. Inan, Chem. Ind. Chem. Eng. Q. 23 (2017) 367-375

[6] L.K. Santos, R.R. Hatanaka, J.E. Oliveira, D.L. Flu-mignan, Renew. Energy 130 (2019) 633-640

[7] F. Ferella, G. Di Mazziotti Celso, I. Michelis, V. Stanisci, F. Vegliò, Fuel 89 (2010) 36-42

[8] National Petroleum, Natural Gas and Biofuels Agency, Brazilian statistical yearbook, http://www.anp.gov.br/ /images/publicacoes/anuario-estatistico/2018/ /anuario_2018.pdf (accessed 20 February 2018)

[9] L. Lovato, B.L. Pelegrini, J. Rodrigues, A.J.B. Oliveira, I.C.P. Ferreira, Biomass Bioenergy 60 (2014) 247-251

[10] O.K. Iha, F.C.S.C. Alves, P.A.Z. Suarez, C.R.P. Silva, M.R. Meneghetti, S.M.P. Meneghetti, Ind. Crops Prod. 52 (2014) 95-98

[11] A.S. Sánchez, M.B. Almeida, E.A. Torres, R.A. Kalid, E. Cohim, A. Gasparatos, Renewable Sustain. Energy Rev. 81 (2018) 2744-2759

[12] F.A. Cajamarca, A.F. Lancheros, P.M. Araújo, I.Y. Mizubuti, S.M. Simonelli, E.I. Ida, C.L.B. Guedes, M.F. Guimarães, Ind. Crops Prod. 126 (2018) 113-119

[13] V.T. Nascimento, N.P. Moura, M.A.S. Vasconcelos, M.I.S. Maciel, U.P. Albuquerque, Food Res. Int. 44 (2011) 2112-2119

[14] W. Harand, R.S. Pinho, L.P. Felix, Braz. J. Bot. 39 (2016) 169-183

[15] G.D. Sousa, J. Kishishita, K.A.S. Aquino, O.A.F. Pres-grave, L.B. Leal, D.P. Santana, AAPS PharmSciTech 18 (2017) 1833-1842

[16] G. Knothe, J. Van Gerpen, The biodiesel handbook, AOCS Press, Champaign, 2005

[17] G. Santori, G.D. Nicola, M. Moglie, F. Polonara, Appl. Energy 92 (2012) 109–32

[18] J.D. Gonçalves, M. Aznar, G.R. Santos, Fuel 133 (2014) 292-298

[19] D.I.S. Silva, M.R. Mafra, F.R. Silva, P.M. Ndiaye, L.P. Ramos, L.C. Filho, M.L. Corazza, Fuel 108 (2013) 269- –276

[20] M. Maghami, J.S. Yousefi, S.M. Sadrameli, A. Haghtalab, Fluid Phase Equilib. 409 (2016) 124-130

[21] M. Hakim, H.A. Najafabadi, G. Pazuki, M. Vossoughi, Ind. Eng. Chem. Res. 53 (2014) 855-864

[22] M. Rostami, S. Raeissi, M. Ranjbaran, M. Mahmoodi, Fluid Phase Equilib. 353 (2013) 31-37

[23] I.C.F. Santos, S.H.V. Carvalho, J.I. Soletti, W.F. Salles, K.T.S. Salles, S.M.P. Meneghetti, Bioresour. Technol. 99 (2008) 6545-6549

[24] J.P.V. Silva, T.M. Serra, M. Gossmann, C.R. Wolf, M.R. Meneghetti, S.M.P Meneghetti, Biomass Bioenergy 34 (2010) 1527-1530

[25] J.C.G. Filho, A.S. Peiter, W.R.O. Pimentel, J.I. Soletti, S.H.V. Carvalho, L. Meili, Ind. Crops Prod. 74 (2015) 767- –772

[26] G. Knothe, Top. Catal. 53 (2010) 714-720

[27] G. Simonelli, C. Moraes, C.A.M. Pires, L.C.L. Santos, Chem. Ind. Chem. Eng. Q. 25 (2019) 183–192

[28] G. Knothe, Energy Fuels 22 (2008) 1358–1364

[29] D.F. Othmer, P.E. Tobias, Ind. Eng. Chem. Res. 34 (1942) 693–696

[30] D.B. Hand, J. Phys. Chem. 34 (1930) 1961-2000

[31] S.I. Sandler, Models for Thermodynamics and Phase Equilibrium Calculations, Marcel Dekker, New York (1994).


29

KARLA RAPHAELA BRAGA DE MELO1

GABRIELA CANTARELLI LOPES1

DAYANA DE GUSMÃO COÊLHO2

JOÃO INÁCIO SOLETTI2 1Department of Chemical Engineering,

Federal University of São Carlos, Rodovia Washington Luís, São Carlos -

São Paulo, Brazil 2Laboratory of Separation and Process

Optimization Systems, Technology Center, Federal University of Alagoas,

s/n Maceió - Alagoas - Brazil

NAUČNI RAD

RAVNOTEŽA TEČNO-TEČNO ZA SISTEM BIODIZEL NA BAZI ULJA PALME Siagrus cearensis-METANOL-GLICEROL

Uključivanje biodizela u brazilsku energetsku matricu је i dalje suočеnо sа izazovima usled slabe raznovrsnosti sirovina i visokih troškova povezanih sa fazama proizvodnje, prečišćavanja i separacije. U ovom radu, procenjen je potencijal palme Siagrus cear-ensis kao alternativni izvor za proizvodnju biodizela bazno katalizovanom transeste-rifikacijom sa metanolom. S obzirom na to da nema podataka o primeni ove uljarice, cilj ovog rada je da se karakterišu ulje i biodizel u pogledu gustine, kinematske viskoznosti i kiselinskog broja. Takođe, sprovedeno je istraživanje ravnoteže tečno-tečno sistema biodizel-metanol-glicerol na 298,15, 308,15 i 323,15 K pri atmosferskom pritisku. Rezul-tati su pokazali da je ovo palmino ulje održiva alternativa za proizvodnju biodizela čiji su parametri u skladu sa standardima kvaliteta i prinosom estera od 97,1%. Ternarni dija-grami su pokazali da se rastvorljivost komponenata smanjuje na niskim temperaturama. Vezivne linije su pokazale veći afinitet metanola prema fazi bogatoj glicerolom. Odre-đeni su koeficijenti raspodele i selektivnosti i primenjene su korelacije Othmer-Tobiasa i Handa. Koeficijenti determinacije (> 0,97) su potvrdili termodinamičku konzistentnost podataka.

Ključne reči: biodizel, Siagrus cearensis, metanol, ravnoteža tečno-tečno.




31

ANA ELISA ACHILES

VÁDILA GIOVANA GUERRA

Department of Chemical Engineering, Federal University of

São Carlos, Brazil

SCIENTIFIC PAPER

UDC 621.928.3+641.513:66.074.2

PERFORMANCE OF A CYCLONE SCRUBBER IN REMOVAL OF FINE PARTICULATE MATTER

Article Highlights • Cyclone spray scrubber is an effective device for the removal of fine particles • Great performance on fine particulate removal at low L/G flow ratio and pressure drop

conditions • Inlet air velocity and water injection position influence the collection efficiency • Liquid-to-gas flow ratio strongly affects the performance of cyclone spray scrubber • Smaller droplet sizes increase the collection efficiency of fine solid particles Abstract

Cyclones are not classified as effective devices for removing fine particles, while high efficiency wet scrubbers usually have high operational costs. In order to achieve better performance, the aim of this study is to evaluate, for the first time, a cyclone scrubber design based on the dimensions of a Stairmand cyclone separator with the inclusion of liquid injection nozzles located in differ-ent positions to improve the separation of fine particles. Given the lack of stu-dies considering the effect of liquid injection and other operational conditions in the removal performance of a cyclone scrubber with Stairmand dimensions, the present paper provides a complete evaluation of these effects for the separation of sugar cane bagasse ash from air. The parameters investigated were inlet gas velocity, liquid injection position, liquid-to-gas flow ratio and droplet size distribution. The cyclone scrubber performance was evaluated considering collection efficiency and pressure drop. Overall efficiency of almost 99% and low-pressure drop was achieved by employing a liquid-to-gas flow ratio of 0.43 L/m³ for the collection of ash from the combustion of sugar cane bagasse. Grade efficiencies revealed that injecting droplets into cyclones sig-nificantly improved the removal of fine particles with an aerodynamic diameter less than 2.5 µm.

Keywords: cyclone scrubber, collection efficiency, fine particle separ-ation, liquid injection.

Cyclones and wet scrubbers are applied in many industrial processes to separate the particles laden in gas stream. An example of a process that uses such equipment is the production of sugar and ethanol. The bagasse is generated in the production process and this waste is used for cogeneration of electricity by burning it in boilers and at the end, it produces sugar cane bagasse ash. The inadequate Correspondence: V.G. Guerra, Federal University of São Carlos. Via Washington Luiz, km 235, 13565-905, São Carlos, SP, Brazil. E-mail: [email protected] Paper received: 20 December, 2018 Paper revised: 11 July, 2019 Paper accepted: 6 August, 2019

https://doi.org/10.2298/CICEQ181220022A

removal of this particulate matter from gas streams can lead to negative consequences, including chronic respiratory problems in humans [1]. Consequently, the removal of particulate matter from these industrial exhaust streams is essential in order to avoid air pollution [2,3].

Dry cyclones are frequently employed in indus-trial applications, due to their low operating costs, simple construction and maintenance, compactness, and suitability for use under harsh conditions [4]. However, dry cyclones are inefficient in collecting very fine particles smaller than 5 μm [5], so they are mainly used for initial pretreatment, upstream of other more expensive air pollution control equipment [6].

A.E. ACHILES, V.G. GUERRA: PERFORMANCE OF A CYCLONE SCRUBBER... Chem. Ind. Chem. Eng. Q. 26 (1) 31−40 (2020)

32

Wet scrubber techniques are applied in many industrial processes, due to their better capacity to remove fine particles and soluble pollutants from gas streams [7]. Nonetheless, major drawbacks of these scrubbing systems are related to the high energy cost associated with successful removal of submicron par-ticles. Venturi scrubbers exhibit significant pressure drops, ranging from 2942 to 8826 Pa [8], while some types of scrubbers require the use of high liquid-to-gas ratios, resulting in the generation of large amounts of sludge [8,9].

In attempting to improve the collection efficiency of dry cyclones for smaller particles and decrease the operational cost associated with wet scrubbers, the cyclone spray scrubber combines the particle collect-ion mechanisms of these techniques, in a single dev-ice. This apparatus essentially consists of a dry cyc-lone adapted with nozzles that inject atomized liquid (usually water) into the inner chamber. The addition of droplets within the equipment leads to improvement of its performance, because in addition to the centri-fugal force acting on the particles, other collection mechanisms (impaction, interception, and Brownian diffusion) also operate, hence increasing the probabil-ity of removing the particles from the gas stream. In general, the removal rate of the small particles is con-trolled by Brownian diffusion, while impaction and interception are significant for removal of large par-ticles [3,10].

The performance of the cyclone spray scrubber, including its collection efficiency and pressure drop, depends mainly on the method of liquid injection, the droplet size distribution, the droplet number, the liquid flow distribution and the initial liquid momentum [3,11]. Several previous studies have investigated the influence of these parameters on the performance and applications of cyclone spray scrubbers.

Krames and Büttner [12] investigated a cyclone scrubber with dimensions based on the Barth [13] and Muschelknautz [14] equations. Droplets were gener-ated by a pneumatic atomization nozzle that was arranged at the cyclone inlet in the direction of the flow. A key finding of this study was an empirical rel-ation between the residence time and the separation result. When the cyclone was operated in dry mode, the separation efficiency declined with decreasing inlet velocity. In contrast, operation in wet mode res-ulted in the best separation at the lowest volumetric flow rate. According to the authors, the cyclone spray scrubber could be classified as a high efficiency separator.

In order to improve understanding of submicron particle removal, Kim et al. [15] conducted a num-

erical investigation of the effects of the different col-lection mechanisms (Brownian diffusion, interception and impaction) on the separation efficiency of a gravitational wet scrubber. It was found that droplet residence time, liquid-to-gas flow ratio, and droplet size distribution had significant effects on the wet scrubber efficiency. Therefore, the water droplet size is one of the most important considerations in design-ing a wet scrubber.

Mohan and Meikap [11] reported on detailed experimental studies on the removal of dust particles from hot gases by using a spray-cum-bubble column with twin-fluid atomizers using water as the scrubbing medium. The results indicated that the spray-cum- -bubble column achieved almost 75-99% removal effi-ciency of particulates. In addition, a comparison of the experimental and theoretical efficiencies was made and analyzed.

Lee et al. [8] developed a novel system con-sisting of a cyclone and a swirl scrubber with an imp-act cone and plates, and evaluated the particle col-lection efficiency and the applications of this appar-atus. The parameters investigated were the plate angles, nozzle size and pressure, and volumetric flow rate of the scrubbing medium. The results demon-strated that particle separation efficiency improved with decrease of the plate angle, increased pressure of the scrubbing medium at the nozzle tip, and inc-reased volumetric flow rate of the scrubbing medium.

Ahuja [2] studied a novel type of wetted wall cyclone scrubber, where wetting of the cyclone walls led to an improvement of up to 33% in the collection efficiency for fine particles in the size range from 1 to 3 μm. Higher inlet air velocity resulted in greater separation efficiency for all particles, under both dry and wet conditions.

Yang et al. [16] studied an innovative type of cyclone, called a “cyclone splitter”, using numerical simulation to evaluate its performance, considering the pressure drops obtained with pure and droplet-laden gasses. The effect of droplets loading on the pressure drop was determined using droplet volume concentrations in the inlet gas ranging from 0.1 to 1.0%. Experimental results showed that the two-phase Euler numbers decreased with increasing droplets loading, and a new pressure drop model for a gas-liquid cyclone was established by introducing a liquid phase correction coefficient.

The aim of the present work was to evaluate the performance of a Stairmand-type cyclone adapted with water injection nozzles, considering the effects of inlet air velocity, droplet size distribution, number of droplets, droplet initial velocity, liquid injection posit-


33

ion, and liquid-to-gas flow ratio. To this end, several experiments were performed under defined conditions in order to investigate the separation efficiency and pressure drop behaviors of this apparatus used to remove sugarcane bagasse fly ash from gas streams.

EXPERIMENTAL

The equipment used in the experiments was a tangential inlet cyclone constructed of stainless steel. The dimensions of the cyclone scrubber were based on the Stairmand-type cyclone. This geometry was chosen because it indicates adequate values of pres-sure drop and collection efficiency using correlations to predict the performance of cyclones. In order to enhance its collection efficiency, the device was adapted with nozzles placed in variable positions along of the cylindrical cyclone body. A schematic illustration of the apparatus and its main dimensions are shown in Figure 1.

As it can be seen, there are six nozzles installed on the cylindrical cyclone body: three located at the upper wall (U configuration) and three located at the side wall (S configuration). The nozzles with U con-figuration are arranged along the cross section of cylindrical body, as drawn in Fig. 1a. The nozzles with S configuration are arranged along the side wall of cylindrical at different distances from the upper wall, which are 0.13 m for the 1S and 3S nozzles and 0.24 m for 2S nozzle, as demonstrated in Fig. 1b. Besides the positions, another distinction among them is their spray orientations. The nozzles placed at upper con-figuration were installed in the vertical direction, so

that their sprays are oriented down. Considering the nozzles located at side configuration were in the hori-zontal direction: 1S and 3S nozzles are facing each other so that their sprays are oriented towards the front and back side walls of the cylindrical body, res-pectively; the 2S nozzle was placed with its spray oriented toward the front side of the cylinder.

In addition to the cyclone, the experimental sys-tem included a blower that delivered atmospheric air into the cyclone at three different inlet velocities (8.0, 10.7 and 13.6 m/s), a rotary plate to inject particulate matter continuously into the inlet gas, and a positive displacement pump to feed water from a storage tank to the scrubber nozzles arranged on the cyclone wall.

During passage through the cylindrical section of the cyclone, the inlet gas stream contacted the atomized water droplets, generating a mixture zone of gas and droplets. Subsequently, the flow of air and liquid droplets was transferred to the chamber walls by centrifugal force, with the liquid draining down to the bottom cone of the cyclone. The air stream, free from the majority of the particulates, exited through a top outlet, while the water and the solid particles were collected in a cylindrical discharge container con-nected to the bottom of the cyclone.

The overall efficiency was quantified by means of simultaneous isokinetic sampling in the gas inlet and outlet of the cyclone by gravimetric analysis: probes connected to vacuum pumps were inserted in the inlet and upper outlet pipes, and the gas/particle mixtures were isokinetically transferred to particle col-lectors containing 0.8 µm pore size membranes. The vacuum provided by the pumps was adjusted using

Figure 1. Dimensions of the cyclone used in the experiments and the nozzle configurations: a) upper view; b) front view.


34

an air flow meter, so that the suction rate was equal to the air velocity within the pipes. Eq. (1) was applied to calculate overall efficiency (ηoverall):

η −= in outoverall

in

C CC

(1)

where Cin and Cout are the mass concentrations of particles at the inlet and upper outlet of the cyclone, respectively.

The grade efficiency was determined by using an aerodynamic particle sizer (APS) spectrometer capable of measuring the particle size distribution - with aerodynamic diameters in terms of mass concen-tration per volume. During this procedure, isokinetic sampling conditions were also maintained and mea-surements were made separately in the inlet and outlet gas pipes of the cyclone. Grade collection efficiency, η(dp), for each particle with aerodynamic diameter (dp) was calculated using Eq. (2):

η−

= p p

pp

in( ) out( )( )

in( )

d dd

d

C C

C (2)

in which Cin(dp) and Cout(dp) are the mass concentration of particles at the inlet and upper outlet of the cyclone, respectively, of particles with aerodynamic diameter dp.

Since the pressure drop has a significant effect on the performance of particle removal equipment, it was measured for all the experimental conditions, using U-shaped water manometers connected at the gas inlet and outlet of the cyclone.

The positions where the isokinetic samplings (Cin and Cout) and pressure measurements (Pin and Pout) were performed are also demonstrated in Figure 1. For all these locations, a Pitot tube was used to measure the fluid velocity profile along the cross-sect-ion of the pipes in order to determine the gas velocity in the exact point where isokinetic samplings were evaluated and verify the gas stream uniformity.

Particle characterization is extremely important for obtaining accurate estimates of the energy requirements of air pollution control devices [17]. The APS spectrometer (TSI Instruments) was also used to determine the particle size distribution. Figure 2 shows the particle size distribution of raw ash used in the collection efficiency tests. The particle density was 2.5 g/cm³ and its concentration in the inlet gas stream was kept below 1.0 g/m³ in all the experi-ments, in order to avoid the influence of this para-meter on the separation characteristics.

The size of the droplets generated by atomizing liquid in spray nozzles is a considerable parameter for

efficient and economical operation of a scrubber [16]. Thus, the influence of the droplet size distribution on the performance of the cyclone scrubber was evalu-ated by using two different nozzles in the experi-ments. Both were of the “pressure swirl” type, where the fluid pressure is converted into kinetic energy [18]. However, one of the nozzles generated a “hollow cone” spray, while the other produced a “full cone” spray.

Figure 2. Cumulative volumetric particle size distribution.

The droplet size distributions of the sprays generated by the nozzles were determined empirically with a Spraytec® system (model RTS5134, Malvern Instruments) based on a laser diffraction technique. As shown in Figure 3, the measurements of droplet size distribution for each nozzle were conducted out of the cyclone in an open environment, so that the liquid spray could be generated freely, without being affected by the cyclone wall or the gas velocity. The nozzle was positioned perpendicularly at the same level as the laser beam, and the liquid flow rate was kept fixed. In order to characterize the size distri-bution of the droplets generated by the spray, the measurements were carried out at a point next to the nozzle (x).

In summary, the adjustable experimental para-meters in this study were the inlet air and water flow rates, the position of the water injection within the cyclone, and the type of scrubber nozzle. To account for any possible empirical errors, the tests were per-formed at least three times over a certain period of time, for each experimental condition. Table 1 shows the experimental conditions investigated.


Droplet size distribution

Figure 4 shows the droplet size distributions for the “hollow cone” and “full cone” nozzles, measured in


35

position x, with the same liquid flow rates. The initial liquid velocity in each nozzle was obtained using Eq. (3):

ρ Δ=

0.5

LL

L

2 PU (3)

derived from the mechanical energy equation for stationary incompressible fluid systems, where UL is the initial liquid velocity; ΔPL is the nozzle liquid pres-

sure; and ρL is the liquid density. The number of droplets was obtained using the correlation of Kach-hwaha et al. [19], shown in Eq. (4):

π= L

g 332

6QND

(4)

where Ng is the number of the droplets generated; QL is the liquid volumetric flow rate; and D32 is the Sauter mean diameter.

Figure 3. Measurement of droplet size distribution: a) general view of the experimental equipment; b) position evaluated in the liquid spray.

Table 1. Experimental conditions investigated

Experiment Type of nozzle Number of nozzles Water injection position vi / m s-1 (L/G) / L m-³

E-1 Hollow cone 1 1U 13.6 0.10



E-4/E-5/E-6/E-7/E-8 Hollow cone 1 2U/3U/1S/2S/3S 13.6 0.10

E-9 Hollow cone 2 1U + 1S 13.6 0.26

E-10 Hollow cone 4 1U + 1S + 3S + 2U 13.6 0.43

Figure 4. Droplet size distributions: a) “hollow cone” nozzle; b) “full cone” nozzle.


36

Effect of inlet air velocity on efficiency

Experiments E-1, E-2 and E-3 were performed to investigate the effect of the inlet air velocity on the grade efficiency of the cyclone scrubber. The results obtained were compared to dry mode operation under the same experimental conditions, as shown in Figure 5.

As shown in Figure 5, the injection of water within the cyclone substantially improved the grade efficiency, for all the inlet gas velocities employed. The droplets generated inside the equipment acted as additional collectors of particles and provided a resist-ance to the gas stream. Interaction of the fine par-ticles with the droplets, by direct collision on or adher-ence into the droplet surfaces, acted to increase their density and, consequently, inertia. Therefore, the effect of centrifugal force on the particles became greater, enabling their separation. In addition, the droplets wetted the inner wall of the equipment, hence decreasing the energy of impact of the par-ticles on the wall and reducing rebound of the par-ticles into the outlet air stream.

Higher inlet air velocity increased grade effi-ciencies under both dry and wet conditions, since it also enhanced the centrifugal force that is mainly res-ponsible for separating particles from the gas stream in conventional cyclones. However, under wet oper-ation, the efficiency of removal of particles smaller than, approximately, 2 µm improved with lower inlet air velocity. Table 2 exhibits these results with more details. As it can be observed, the experimental con-dition with lower air inlet velocity, even performing a lower overall efficiency, was more efficient in collect-ing particles smaller than 1.84 µm.

This behavior associated with smaller particles occurs due to two main reasons:

• turbulence; • residence time. Firstly, higher turbulence, caused by greater gas

flow velocities, increases the separation of larger par-ticles since they have difficulty following the high-speed spiral motion of the gas and the vortex, so the particles hit the inside walls of the container and drop down into a collection chamber. Nonetheless, with regard to fine particles, since they have less inertia, the vortex influence on them is more considerable and, because of this, they are dragged with the clean gas.

Table 2. Grade efficiencies (%) of particles lower than 2.0 µm under constant L/G

Air inlet velocity m/s

Overall efficiency, %

Particle diameter, µm

1.49 1.60 1.72 1.84 1.98

Dry operation

8.0 75.25 47.56 48.19 50.59 52.16 55.91

13.6 81.00 44.83 49.74 53.09 57.64 63.58

Wet operation

8.0 90.18 72.52 75.59 77.52 79.71 81.25

13.6 95.50 61.68 67.92 72.41 77.94 83.87

Another important effect is the residence time (τ), defined by Eq. (5), which is longer as the gas inlet flow rate is lower. Therefore, it means that lower air inlet velocities in cyclone scrubbers are able to pro-vide a higher contact time between particles and droplets:

τ = Volume of the cyclonegas inlet flow rate

(5)

Thus, the lower turbulence associated with longer residence time, both occasioned by the dec-

Figure 5. Effect of inlet air velocity on grade efficiency under dry and wet operation.


37

reasing of the air inlet velocity, are responsible to keep the smaller particles that are not dragged by the clean gas in contact with the droplets by a longer period. Then, the collision on or adherence of these particles into the droplet surfaces are enhanced, inc-reasing their inertia and, consequently, their separ-ation.

Effect of water injection position on efficiency

Experiments E-1 and E-4 to E-8 were performed to examine the influence of water injection position on the overall efficiency of the cyclone scrubber, with the other adjustable conditions being kept constant. The positions tested are shown in Figure 1 and the results are provided in Table 3.

Table 3. Overall efficiency for each water injection position

Experiment Water injection position Overall efficiency, %

E-1 1U 95.50

E-4 2U 93.06

E-5 3U 92.82

E-6 1S 95.48

E-7 2S 88.39

E-8 3S 95.22

The values obtained for the overall efficiency indicated that this parameter was affected by the position of water injection within the cyclone scrubber. When the nozzles were positioned in the upper wall of the cylindrical body, the closer the water injection position was to the gas inlet, the higher the collection efficiency was, as demonstrated by the nozzles at positions 1U and 2U. Regard to the nozzles that were located in the side wall, the 2S configuration resulted in the worst separation efficiency, because it was

placed at a lower height, so the contact time between the droplets generated and the gas stream was shorter.

Considering all the water injection positions, it was found out the nozzles which were at positions 1U, 2U, 1S and 3S showed a quite similar effect on the overall efficiency, with a maximum difference of 2.4% between them. Higher efficiencies were achieved when the sprays had closer contact with the inlet gas/particle stream and, consequently, provided greater probability of collision.

Effect of L/G flow ratio on efficiency

Since the liquid-to-gas flow ratio is crucial in the performance of all kinds of wet scrubbers, the influ-ence of this parameter was determined by varying it in the range from 0.10 to 0.43 L/m³ (experiments E-1, E-9 and E-10). This L/G range was obtained by alter-ing the number of nozzles operating within the equip-ment and the results obtained were compared to those for dry operation under the same conditions. The nozzles liquid injection used in these tests (Table 1) were those that provided higher overall efficiencies, as it can be observed in Table 3. The decision criteria that influenced this choice was the fact that these nozzles did not present a noticeable discrepancy among their overall efficiency results. Thus, this sel-ection was a way to ensure that water injection pos-ition would not be promote a significant effect on the results.

A higher L/G ratio increased the number of drop-lets within the cyclone, so the mean distance among the droplets was reduced. Hence, the probability of impaction and interception between the droplets and the solid particles was higher, leading to important improvements in the overall and grade efficiencies, as shown in Figure 6 and Table 4, respectively.

Figure 6. Effect of L/G ratio on grade efficiency.


38

Table 4. Effect of liquid-to-gas flow ratio on overall efficiency and pressure drop

(L/G) / L m-3 Overall efficiency, % Pressure drop, Pa

0 81.00 657.0

0.1 95.50 627.6

0.26 97.51 490

0.43 98.44 392

Although the increase of the L/G ratio substan-tially improved the separation efficiency, the improve-ment was not linear (Table 4). When the L/G ratio was doubled and quadrupled, the increments in overall efficiency were only 2.01 and 0.93%, respectively. Therefore, this parameter should be evaluated in order to determine whether the increased separation efficiency might justify the higher water consumption and the operational costs associated with wastewater treatment, since an optimum value of liquid-to-gas flow ratio is that one which provides the maximum effi-ciency with a minimum scrubbing liquid application.

On the other hand, the experimental cyclone spray scrubber provided a substantial increment of 17.44% on overall efficiency, compared to dry oper-ation. High fine particle collection performance of 98.44% was achieved using a low L/G ratio of 0.43 L/m³, which was equivalent to the performance other high-efficiency wet scrubbers, such as Venturi scrub-bers that employ L/G ratios from 0.5 up to 5.0 L/m³ [12].

It is also important to point out that the injection of atomized water into the cyclone mainly improved the removal of fine particles with aerodynamic dia-meter less than 2.5 µm (PM2.5). The average increase achieved for this particle size category (experiment

E-10) was approximately 27%, which was a satis-factory result considering the potential harmful effects of these fine particles on human health.

Effect of droplet size distribution on efficiency

Figure 7 shows the grade efficiencies obtained for each type of nozzle, as well as for dry operation under the same experimental conditions.

The highest collection efficiency was observed for the “hollow cone” nozzle that applied greater pressure on the liquid, as found also by Lee et al. [8]. Higher pressure nozzles generate a larger number of smaller droplets that can achieve higher initial velo-cities, as shown in Figure 5, which improved the col-lection mechanism between the particles and droplets since the greater kinetic energy, the higher the impact among them and, therefore, more particles are penet-rated in the surface tension of the liquid scrubbing. In addition, smaller droplets have a larger surface area and can remain in contact with the gas stream for longer times due to their lower inertia.

All these characteristics increase the probability of collision between the droplets and the particles, consequently improving the separation efficiency. The results obtained here were in agreement with the findings of Kim et al. [15], who reported that for the entire particle size range, the collection efficiency inc-reased as the droplet size decreased.

Pressure drop

Besides the collection efficiency, the pressure drop is another significant parameter that affects cyc-lone spray scrubber performance, consequently influ-encing the operational costs. Therefore, the pressure drop was also evaluated in the experimental tests under dry and wet conditions.

Figure 7. Effect of liquid injection nozzles (droplet size) on grade efficiency.


39

As shown in Table 4, the pressure drop of the cyclone scrubber decreased from 627.6 to 392 Pa when the L/G ratio was increased from 0.10 to 0.43 L/m³.

The decrease in the pressure drop when liquid droplets are injected into the air streams of cyclones can be explained by a decrease of the tangential component of the gas velocity, resulting from the iner-tia of the particles/droplets contained in the air stream and increased friction at the walls of the equipment. According to Yang et al. [16], the main mechanism underlying this pressure drop reduction is enhance-ment of the friction factor, as a result of the hydraulic roughness of the liquid film produced on the inner wall of the cyclone by the injection of droplets. The higher friction factor is then responsible for reducing the maximum tangential velocity and the swirling effect.

Table 5 shows pressure drops for the cyclone spray scrubber under dry and wet operation. The results of experiments E-1, E-2 and E3 showed that when the inlet air velocity was increased from 8.0 to 13.6 m/s, the pressure drop under wet conditions inc-reased from 196.1 to 627.6 Pa. Similar behavior was observed when the cyclone was operated under dry conditions; however, the injection of water within the cyclone caused a slight reduction of approximately 6.5% in the pressure drop, for all the inlet air velo-cities employed.

Table 5. Pressure drops (Pa) for the cyclone scrubber under dry and wet operation

Experiment Operation

Wet Dry

E-1 627.6 657.0

E-2 387.0 412.0

E-3 196.0 215.8

The cyclone spray scrubber investigated here showed excellent performance when compared with other high-performance wet scrubbers, such as the Venturi scrubber which usually operates at a signific-ant pressure drop of 2942-8826 Pa and L/G range from 0,5 to 5 L/m3. The cyclone spray scrubber pro-vided a highly satisfactory collection efficiency of 98.44% using a reasonable L/G ratio (experiment E-10), but at a low applied pressure drop of 392 Pa, confirming its effectiveness as a device for particle control.

CONCLUSIONS

In order to overcome the drawbacks of dry cyc-lones and wet scrubbers, in this study a cyclone

scrubber, designed based on dimensions of a Stair-mand cyclone separator, was investigated in terms of collection efficiency and pressure drop. The conclus-ions can be summarized as follows:

• A higher inlet air velocity improved the collect-ion efficiency for larger particles, but decreased the separation of smaller particles.

• The water injection position influenced the overall collection efficiency.

• The liquid-to-gas flow ratio strongly influenced the performance of the cyclone spray scrubber, con-sidering both separation efficiency and pressure drop.

• Smaller droplet sizes could increase the col-lection efficiency of the solid particles.

• The experimental cyclone spray scrubber tested here was effective in the removal of fine par-ticles, especially those with aerodynamic diameter less than 2.5 µm. Separation of 98.44% of the solid particles present in the gas stream was achieved employing a low L/G ratio of 0.43 L/m³ and a low pressure drop of 329 Pa.

Nomenclature

APS Aerodynamic Particle Sizer Cin, Cout Particle concentrations (g/m³) in the cyc-

lone inlet and outlet, respectively D32 Sauter mean diameter (µm) dp Diameter of particle (µm) L/G Liquid to gas flow ratio (L/m3) Ng Number of droplets (-) x Measuring position of the droplets (cm) ΔPL Liquid pressure at the nozzle (Pa) PM2.5 Particulate matter with aerodynamic dia-

meter less than 2.5 µm (-) QL Volumetric liquid flow rate (m³/s) S Side configuration of the nozzles within the

cyclone (-) U Upper configuration of the nozzles within

the cyclone (-) UL Initial liquid velocity (m/s) vi Inlet gas velocity (m/s)

Greek letters η Collection efficiency (-) ρL Liquid density (kg/m³) τ Residence time (s)

Acknowledgements

The authors are grateful to Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) (grant number 454754/2014-0) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) (Finance Code 001) for the financial support provided for this study.


40

REFERENCES

[1] A. Bhatnagar, K.K. Kesari, N. Shurpali, Water Air Soil Pollut. 227 (2016) 1–30

[2] S.M. Ahuja, Powder Technol. 204 (2010) 48–53

[3] K.S. Lim, S.H. Lee, H.S. Park, J. Aerosol Sci. 37 (2006) 1826–1839

[4] B. Zhao, Chem. Eng. Process. 44 (2005) 447–451

[5] D.L. Iozia, D. Leith, Aerosol Sci. Technol. 10 (1989) 491– –500

[6] E. Ashbee, W.T. Davis, in Air Pollution Engineering Manual, Air & Waste Management Association, A.J. Buonicore, W.T. Davis (Eds.), Van Nostrand Reinhold, New York, 1992, pp. 71–78

[7] T. Mi, X.M. Yu, Chem. Eng. Process. 62 (2012) 159–167

[8] B.K. Lee, K.R. Jung, S.H. Park, J. Aerosol Sci. 39 (2008) 1079–1088

[9] S. Zarei, E. Jamshidi, A. Afshar Ebrahimi, Chem. Eng. Process. 49 (2010) 1193–1198

[10] S.H. Park, B.K. Lee, J. Hazard. Mater. 164 (2009) 315– -321

[11] B.R. Mohan, B.C. Meikap, Chem. Eng. Res. Des. 87 (2009) 109–118

[12] J. Krames, H. Büttner, Chem. Eng. Technol. 17 (1994) 73–80

[13] W. Barth, Brennst. Warme Kraft. 8 (1956) 1–9

[14] E. Muschelknautz, Auslegung von Zyklonabscheidern in der technischen Praxis, Staub – Reinhalt. Luft. 30 (1970) 187–195

[15] H.T. Kim, C.H. Jung, S.N. Oh, K.W. Lee, Environ. Eng. Sci. 18 (2001) 125–136

[16] J. Yang, C. Liu, S. Li, B. Sun, J. Xiao, Y. Jin, Chem. Eng. Process. 95 (2015) 256–266

[17] K.C. Schifftner, H.E. Hesketh, Wet Scrubbers, 2nd ed., Technomic Publishing Company, Lancaster, PA, 1996

[18] A.H. Lefebvre, Gas Turbine Combustion, Hemisphere Pub. Corp., Washington, DC, 1983

[19] S.S. Kachhwaha, P.L. Dhar, S.R. Kale, Int. J. Heat Mass Transfer. 41 (1998) 447–464.

ANA ELISA ACHILES VÁDILA GIOVANA GUERRA

Department of Chemical Engineering, Federal University of São Carlos, Brazil

NAUČNI RAD

EFIKASNOST CIKLONSKOG SKRUBERA U UKLANJANJU SITNIH ČESTICA

Cikloni se ne smatraju efikasnim uređajima za uklanjanje sitnih čestica, dok jako efi-kasni vlažni skruberi obično imaju velike operativne troškove. Cilj ovog rada je da se prvi put proceni ciklonski skruber dizajniran na osnovu dimenzija Stairmand ciklona uz uklju-čivanje mlaznica za ubrizgavanje tečnosti koje se nalaze na različitim položajima radi poboljšanja odvajanja sitnih čestica. Imajući u vidu nedostatak istraživanja koja bi uzele u obzir uticaj ubrizgavanja tečnosti i ostalih radnih uslova na performanse ciklonskog skrubera sa Stairmandovim dimenzijama, ovaj rad daje potpunu procenu ovih uticaja na odvajanje pepela otpadaka šećerne trske iz vazduha. Istraživani parametri su brzina ulaznog gasa, položaj ubrizgavanja tečnosti, odnos protoka tečnosti i gasa i raspodela veličine kapljica. Efikasnost prečišćavanja ciklonskog skrubera je procenjena uzimajući u obzir efikasnost odvajanja i pad pritiska. Ukupna efikasnost od gotovo 99 % i mali pad pritiska postignuti su korišćenjem odnosa tečnosti i gasa od 0,43 l/m³ za odvajanje pepela nastalog sagorevanjem otpadaka šećerne trske. Stepen efikasnost je pokazao da je ubrizgavanje kapljica u ciklone značajno poboljšalo uklanjanje sitnih čestica preč-nika manjim od 2,5 um.

Ključne reči: ciklonski skruber, efikasnost odvajanja, razdvajanje sitnih čestica, ubrizgavanje tečnosti.




41

MARIJA LJEŠEVIĆ1

JELENA MILIĆ1

GORDANA GOJGIĆ-CVIJOVIĆ1

TATJANA ŠOLEVIĆ KNUDSEN1

MILA ILIĆ1 JELENA AVDALOVIC1

MIROSLAV M. VRVIĆ1,2 1National Institute of Chemistry,

Technology and Metallurgy, University of Belgrade, Belgrade,

Serbia 2Brem Group Ltd., Belgrade,

Serbia

SCIENTIFIC PAPER

UDC 547.53+665.654:561.23

EVALUATION OF ASSAYS FOR SCREENING POLYCYCLIC AROMATIC HYDROCARBON- -DEGRADING POTENTIAL OF BACTERIA

Article Highlights • Assays can be used reliably for strain selection with a high potential for bioremedi-

ation procedures • Dehydrogenase activity assay correlated positively with the hydrocarbon growth assay • Dehydrogenase activity of Rhodococcus RNP05 was significantly higher than Plano-

microbium RNP01 • Planomicrobium RNP01 had the lowest ability of growth on pyrene • Rhodococcus RNP05 had the highest ability of growth on dibenzothiophene Abstract

Within a 30-day incubation laboratory study, the polycyclic aromatic hydro-carbon (PAH) degradation profile of two bacteria, Planomicrobium sp. RNP01 and Rhodococcus sp. RNP05 were studied by three microtiter plate assays to reveal the combination of certain biological and biochemical characteristics which are reliable indicators in evaluation of bacterial biodegradation abilities. The three assays, which are hydrocarbon growth assay, 2,6-DCPIP assay and dehydrogenase activity assay revealed that Rhodococcus sp. RNP05 exhibited better potential for PAH degradation than Planomicrobium sp. RNP01. Differ-ences between initial and final optical density and specific growth rate cons-tants were significantly higher (r = 0.995, P < 0.05) in case of Rhodococcus sp. RNP05 on all tested substrates, as compared to Planomicrobium sp. RNP01. This was confirmed by GC-FID analyses. Dehydrogenase activity of Rhodo-coccus sp. RNP05 was higher (r = 0.9995, P < 0.05) than Planomicrobium sp. RNP01 and correlated positively with the hydrocarbon growth assay (r = 0.999, P < 0.05, for Rhodococcus sp. RNP05, r = 0.986, P < 0.05 for Planomicrobium sp. RNP01). This study has shown that the combination of these assays could be used for determining the bioremediation potential of PAHs in petroleum con-taminated soil with the ability of screening a large number of bacterial strains.

Keywords: hydrocarbon growth, PAHs, screening assays, dehydrogen-ase activity, Planomicrobium, Rhodococcus.

Hydrocarbons contamination in the environment is caused by exploitation, transport, processing, storage and use of petroleum and its derivatives, as well as from combustion plants, motor vehicles and other gasoline-powered equipment. Polycyclic aro-matic hydrocarbons (PAHs) are of special concern

Correspondence: J. Milić, Nacional Institute of Chemistry, Tech-nology and Metallurgy, University of Belgrade, Njegoševa 12, 11000 Belgrade, Serbia. E-mail: [email protected] Paper received: 20 February, 2019 Paper revised: 2 August, 2019 Paper accepted: 10 August, 2019

https://doi.org/10.2298/CICEQ190220023L

since these compounds are considered potential health and environmental risks [1], and many of them have toxic, mutagenic and/or carcinogenic properties [2,3]. Sixteen PAHs have been specified as priority pollutants by the US Environmental Protection Agency as well as the European Commission [4,5]. Soil contaminated with hydrocarbons is classified as hazardous waste according to the European Waste Catalogue under index number 17 05 03* (soil and stones containing hazardous substances) and require remediation [6].

Because of their molecular stability and low sol-ubility, bioavailability of PAH compounds could be decreased, and the time to reach acceptable end-

M. LJEŠEVIĆ et al.: EVALUATION OF ASSAYS FOR SCREENING POLYCYCLIC... Chem. Ind. Chem. Eng. Q. 26 (1) 41−48 (2020)

42

points for bioremediation treatments could be ext-ended [7-9]. One of the important factors for PAH biodegradation is the presence and activity of PAH- -degrading microorganisms [10,11].

To show that a bioremediation procedure will be effective, it is important to first demonstrate the bio-availability of substrate and ability of microorganisms to enhance the rate of hydrocarbon degradation in controlled conditions [12]. Thus, the assessment of bioremediation potential of bacterial communities is an important step when deciding the appropriate bio-remediation strategy.

Turbidity and colorimetric measurements are low-cost and rapid procedures to detect the occur-rence of microbial metabolism in both aerobic and anaerobic conditions. When used in microtiter plates, besides simplicity, the main advantages of these methods are rapid screening of large numbers of bac-terial isolates and the fact they can be used with hydrocarbon-degrading nonculturable bacteria.

2,6-Dichlorophenolindophenol (2,6-DCPIP) assay has a sufficient sensitivity to detect primary oxidation of hydrocarbons in the first four weeks of the biodeg-radation process and it can be used as a quick screening method [13]. Moreover, this method rev-ealed two extremely potent bacteria for degradation of high molecular weight PAH, Planomicrobium sp. RNP01 and Rhodococcus sp. RNP5, which was confirmed by gas chromatography analysis of biodegradation of the mixture of high molecular weight PAHs [13].

The aim of this study was to evaluate and com-pare the three screening methods in microtiter plates (hydrocarbon growth assay, 2,6-DCPIP assay and dehydrogenase activity assay) in order to reveal reli-able indicators in the evaluation of bacterial biodeg-radation potential. Obtained results could potentially lead to developing of a set of simple microbial assays for quick and reliable PAH bioremediation assess-ment in laboratory conditions, one of the critical fac-tors in deciding bioremediation strategies.

EXPERIMENTAL

Chemicals

Phenanthrene (PHE) and dibenzotiophene (DBT) were at >96% purity (Sigma-Aldrich, Germany), and pyrene (PYR) was at >97% purity (Fluka, Ger-many).

Microorganisms

Bacterial strains Planomicrobium sp. RNP01 (GenBank Accession No. JN683359) and Rhodococ-cus sp. RNP05 (GenBank Accession No. JQ065876)

were isolated from the soil taken from Pančevo Oil Refinery, (Serbia). Isolation and identification of bac-terial strains has been described by Milić et al. [13].

Screening assays

Isolated pure bacterial strains were grown indi-vidually in 100 ml of bacterial culture medium (BCM), which consisted of mineral medium (MM) containing NPK solution (0.1% NH4NO3 and 0.025% K2HPO4) with 2000 ppm of diesel fuel D2, for 48h on a rotary shaker at 120 rpm and 28 °C. After incubation, the cell cultures were centrifuged at 6000 rpm at 10 °C for 20 min. Bacterial inoculum of each culture was estab-lished by suspension of cell pellets (twice washed in saline solution) in sterile MM solution to an optical density of OD600 = 1. In all experiments, aliquots of 150 μl of bacterial inoculum were used. Biodegrad-ation tests were performed in polystyrene 24 well mic-rotiter plates in triplicate. This set-up was used in all three methods of assaying in vivo microbial degrad-ation of PAHs.

Absorbance data of the control wells were sub-tracted from the solutions prior to calculations, for all screening methods. Unless otherwise specified, all results reported are averages of triplicate determin-ations.

Hydrocarbon growth assay

Hydrocarbon growth assay was performed by adding aliquots of inoculum to 2 ml sterile MM sol-ution with individual PAH (PHE, PYR or DBT). PAHs (dissolved in ether) were added to achieve a final concentration of 300 ppm in each microtiter well. The control was prepared using 2 ml of MM solution without individual PAHs. The plates were incubated on a rotary shaker at 120 rpm and 28 °C for 30 days, and optical density was measured at 600 nm on days 0, 5, 10, 15, 20, 25 and 30 of the experiment, using an Elisa reader (LKB 5060-006). Log OD600 for each strain was plotted as a function of time. Constant growth rates were calculated by curve fitting using Origin Fit Tool (OriginPro 8, OriginLab software) [13].

2,6-DCPIP assay

It is possible to ascertain the ability of the micro-organism to utilize a hydrocarbon substrate by obser-ving the color change of DCPIP from blue (oxidized) to colorless (reduced). This assay is qualitative since absorbance of the blue color (595 nm) suffers from interference from the increasing turbidity of the medium during the incubation time.

The qualitative 2,6-DCPIP assay was performed using the same method as the hydrocarbon growth assay, except that 650 μl of 15.3 % 2,6-DCPIP sol-


43

ution (153 μg ml-1) was added to every well to achieve a final concentration of 50 μg ml-1 [14]. The color change was registered as a + (discoloration) and - (blue color) after 15 and 30 days of growth.

Dehydrogenase activity assay (DHA)

The DHA is measured by a colorimetric method using 2,3,5-triphenyltetrazolium chloride (TTC) as an electron acceptor for many dehydrogenase enzymes [15]. Reduction of this compound by dehydrogenase gives triphenylformazan (TPF) which has a charac-teristic reddish color. The intensity of color is mea-sured at 485 nm and it is a good indicator of microbial activities [16]. Van der Waarde et al. showed the parameter that had the best correlation with hydro-carbon removal and soil respiration was dehydrogen-ase activity [17].

The bacterial isolates were grown at 28 °C in MM solution with individual PAH (PHE, PYR or DBT, as the sole source of hydrocarbon) for 30 days in mic-rotiter plates, as described in the section “Hydro-carbon growth assay”. After 30 days, A 50 µl aliquot of an electron acceptor solution, triphenyl tetrazolium chloride (TTC) solution (0.25 g TTC in 100 ml 100mM tris buffer, pH 7), was added to each well and the plates were incubated at 28 °C for 48 h for color dev-elopment. The hydrolysis reaction product (TPF) was extracted for 2 h at 30 °C with acetone and absorb-ance at 490 nm was recorded on the ELISA reader.

Microcosm assays

Biodegradation of PAH by Planomicrobium sp. RNP01 and Rhodococcus sp. RNP05 was examined in microcosm assays, simultaneously with screening assays. Microcosm assays were set up in 500 ml Erlenmeyer flasks containing 100 ml MM solution (with PHEN, PYR or DBT in final concentration of 50 ppm) and inoculated with 1 ml of inoculum. The assays were incubated for 30 days, at 28 °C with rotary stirring at 100 rpm.

The remaining PAH was extracted from the whole medium three times with n-hexane (50 ml). The organic layer was collected and washed with 50 ml 2% NaCl. After dehydration over anhydrous Na2SO4, it was concentrated using a vacuum rotary evaporator until the organic solvent was completely removed. All extracts were dissolved in the same volume of solvent (1 ml), and each extract was analyzed instrumentally by injecting the same volume of the solution (1 μl).

GC-FID analysis

The ability of the bacterial isolates to degrade PAHs was confirmed by gas-chromatography as ana-

lytical evidence that they can metabolize these com-pounds.

A gas chromatograph (Agilent 4890D) with flame ionization detector (FID) and HP-5MS column (30 m×0.25 mm, 0.25 μm film thickness) was used for gas chromatography. The carrier gas was hydrogen with a constant flow rate of 1 ml min-1. Injector tempe-rature was constant (250 °C), as was the detector temperature (300 °C). The following temperature ramping was used: initial temperature 80 °C, and then heating at a rate of 10 °C min-1 up to a temperature of 300 °C.


Evaluation of bioremediation potential of bacterial strains for PAHs

Hydrocarbon growth assay Hydrocarbon growth assay demonstrated that,

in term of substrates, both bacterial strains had the high ability of growth on DBT substrate, and low on PYR substrate. Rhodococcus sp. RNP05 achieved higher efficiency of degradation on all tested sub-strates, i.e., PHE, PYR and DBT (A600 of 0.320, 0.240 and 0.430, respectively), comparing with Planomic-robium sp. RNP01 (A600 of 0.219, 0.186 and 0.340, respectively). Difference between initial and final opti-cal density (represented as ΔOD), and specific growth rate constants (K) were significantly enhanced (r = = 0.995, P < 0.05) in case of Rhodococcus sp. RNP05 on all tested substrates, as compared to Pla-nomicrobium sp. RNP01 (Figure 1).

2,6-DCPIP assay The biodegradability of petroleum hydrocarbon

compounds can be verified using the technique based on the 2,6-dichlorophenol indophenol (2,6-DCPIP) redox indicator [18]. 2,6-DCPIP is a redox indicator that detects oxidation of NADH to NAD+ during the bacterial degradation of hydrocarbons. The principle of this technique is that during the microbial oxidation of hydrocarbons, electrons are transferred to electron acceptors such as oxygen, nitrates and sulphate. This results in the change in color of 2,6-DCPIP from blue (oxidized) to colorless (reduced).

Both bacterial strains changed color of the 2,6- -DCPIP solution from blue to colorless, after two weeks of experiment (Table 1), which indicates these strains have the ability to degrade tested PAH com-pounds, which is in correspondence with earlier pub-lished results [13].


44

Figure 1. Parameters of: a) phenanthrene-PHEN, b) pyrene-

-PYR and c) dibenzothiophene-DBT biodegradation by Plano-microbium sp. RNP01 and Rhodococcus sp. RNP05; a) bacte-

rial growth (OD600), b) specific growth rate constant, K (h-1), c) difference between initial and final optical density

represented as ΔOD.

Table 1. Bacterial oxidative activity measured using qualitative 2,6-DCPIP assay; blue (-/no growth) colorless (+/growth)

Bacterium PHE PYR DBT

Planomicrobium sp. RNP01 + + +

Rhodococcus sp. RNP05 + + +

Dehydrogenase activity assay Dehydrogenase activity (DHA) typically occurs

in all intact, viable microbial cells. Oxidation of org-anic matter by microorganisms assumes the involve-ment of a dehydrogenase enzymatic system by trans-ferring hydrogen from the organic substrates to the electron acceptor, so that its activity is a good indi-

cator of the microbiological action in contaminated environments, as well as of the bioremediation pot-ential of microorganism and dynamics of bioremedi-ation within a period of time. The very low water solubility of PAHs and their slow mass transfer rates from solid phase may limit their availability to micro-organisms, thus hindering natural microbial proces-ses of attenuation. The DHA assay can be used as a simple method to examine the possible inhibitory effect of environmental contaminants on microbial activities.

Both bacterial strains generally had a good dehydrogenase activity on tested PAH compounds after 30 days of biodegradation (Figure 2). However, dehydrogenase activity was the highest during bio-degradation test with Rhodococcus sp. RNP05 with highest DHA values achieved after growth on DBT and PHEN, with A490 up to 0.397 and 0.323, respect-ively.

Dehydrogenase activity of Rhodococcus sp. RNP05 was significantly higher (r = 0.9995, P < 0.05) than Planomicrobium sp. RNP01 and correlated positively with the hydrocarbon growth assay (r = 0.999, P < 0.05 for Rhodococcus sp. RNP05, r = 0.986, P < 0.05 for Planomicrobium sp. RNP01).

Figure 2. Dehydrogenase activities (A490) of bacteria after

growth on PAHs.

The evaluation of bioremediation potential of contaminated soil by different bacteria is a fundam-ental task when deciding correct bioremediation stra-tegies. In order to accomplish this, colorimetric screening methods have many advantages, because they are simple, fast, robust, inexpensive and conve-nient methods, making them applicable in most lab-oratories. These methods provide a picture of the metabolic activity of a microorganism growing on the hydrocarbon substrate examined, and take into account the nonculturable bacteria when consortiums from soil extracts are used.

In order to evaluate and compare three screen-ing methods, the bioremediation assessment of two hydrocarbon-degrading bacteria were analyzed using


45

the hydrocarbon growth assay, 2,6-DCPIP assay, and dehydrogenase activity assay. The hydrocarbon growth assay successfully determined difference in biodegradation potential between two bacterial strains. Furthermore, the results of oxidation activity measured by the 2,6-DCPIP completely corres-ponded with results of hydrocarbon growth assay, indicating that the 2,6-DCPIP assay can be used to evaluate microbial PAH-degradation abilities in an accurate, sensitive and simple manner. However, even though these results indicate that these assays could be used separately, our recommendation is to use both assays, because the 2,6-DCPIP assay cor-responds to hydrocarbon concentration decreases while the increasing turbidity of the suspension in the hydrocarbon growth assay could be a consequence of the increasing bacterial numbers, but also the size of the cells, which could obscure the findings. The 2,6-DCPIP assay has been employed for evaluation of hydrocarbon-degradation abilities in several stu-dies [18-20]. Kubota et al. showed that biodegrad-ation profiles analyzed by 2,6-DCPIP assay com-pletely corresponded to GC analysis of hydrocarbon-degradation [14]. Furthermore, Mariano et al. [20] valuated the capability of different microorganisms to degrade butanol/gasoline and ethanol/gasoline blends using 2,6-DCPIP assay and respirometry assay, and verified the order of biodegradability as ethanol > butanol > gasoline by both experiments. Bidoia et al. qualitatively examined a reducing 2,6-DCPIP color as a result of biodegradation and concluded that 2,6- -DCPIP concentrations below 0.03 g L-1 could not be measured by this method [21]. In our previous exam-inations, 2,6-DCPIP assay has been successfully used to evaluate degradation of PAH by microorg-anisms in 30 days [13], yet the results obtained after 15 and 30-day experiments suggested that the time period of the assay can be set to 15 days.

The third assay, the dehydrogenase activity assay, was used to check if the examined micro-organisms just tolerate PAHs or metabolically use them as sole carbon sources. In this assay, the red-uction of compounds by dehydrogenase gives tri-phenylformazan (TPF) which has a characteristic reddish color, and the intensity of color is measured at 485 nm, which is a good indicator of microbial act-ivities [16]. The measurement of the dehydrogenase activity by microorganisms from soil contaminated by petroleum hydrocarbons has the potential to assess an effectiveness of various bioremediation proce-dures [17]. DHA results revealed that both bacteria were metabolically active on tested PAH substrates and confirmed the results from the previous two assays.

Gas chromatographic analysis of bioremediation efficiency of strains Planomicrobium sp. strain RNP01 and Rhodococcus sp. strain RNP05

In order to confirm the results of screening assays, Planomicrobium sp. RNP01 and Rhodococ-cus sp. RNP05 have been growing in microcosms on PHEN, PYR and DBT as individual substrates. After 30 days of microcosm assay, GC-FID analysis of PAH extracts was used to check the degradation of substrates.

According to GC analysis, both tested micro-organisms utilized almost all of the DBT substrate after 30 days of growth. GC analysis revealed lower utilization of PHE and PYR by both microorganisms, where Rhodoccocus sp. RNP05 showed higher biodegradation potential for PHE and DBT, compared to Planomicrobioun sp. RNP01 (Figures 3 and 4).

To assure the accuracy and sensitivity of screening assays, the results of bioremediation assessment of hydrocarbon-degradation pattern of two strains, Planomicrobium sp. strain RNP01 and Rhodococcus sp. strain RNP05, was confirmed by the GC-FID analysis. The results from the gas chromato-graph analysis confirmed the order of biodegradability which has been indicated by screening assays: DBT> PHE>PYR; this shows that the three screening assays used as microtiter plate method can be used for quick and reliable selection of microorganisms that have a high potential for soil bioremediation proce-dures.

To show that a bioremediation procedure is potentially useful, it is important to demonstrate the bioavailability of substrate as well as the ability of selected microorganisms to enhance the rate of hydrocarbon degradation in controlled conditions [12]. Measuring the success of bioremediation of petro-leum-contaminated soil is based on several para-meters, among which is the degradation of polycyclic aromatic hydrocarbons.

The potential of microorganisms to metabolize hydrophobic substrates under the defined conditions is used as a screening parameter for choosing the strains with multiplicity of catabolic pathways for hyd-rocarbon compounds because bioremediation proce-dures are strongly dependent on process duration and biological efficiency of bacterial communities [11]. Thus, the isolation, characterization and profile of specific bacteria for hydrocarbon degradation are important when deciding the correct bioremediation strategy.

Turbidity and colorimetric measurements are low-cost and rapid procedures to detect the occur-rence of microbial metabolism, in both aerobic and


46

anaerobic conditions. The main advantages of these methods are rapid screening of large numbers of bacterial isolates, simplicity and the fact that they can be used with hydrocarbon-degrading nonculturable bacteria. The use of these measurements in the form of microtiter plates contributes to faster screening of a large number of microorganisms, and better selection of the most potent strains.

CONCLUSIONS

In this work, a set of three simple screening methods have been used to evaluate the biorem-ediation potential for PAH of bacterial isolates obtained from a soil contaminated with petroleum. In all three assays, Rhodococcus sp. strain RNP05 showed better potential than Planomicrobium sp. strain RNP01 for utilizing all the examined hydro-carbon substrates (phenanthrene, pyrene and DBT), which was confirmed by GC-FID analysis. Even

though these results indicate that the assays could be used separately, our recommendation is to combine all three assays for determination and comparison of bacterial bioremediation potential for PAHs, because they correspond to different parameters of bacterial metabolic activity.

To the best of our knowledge, this is the first report demonstrating that a combination of these simple screening methods is effective for determining the bioremediation potential for PAHs. Using all three methods to assess bioremediation potential for PAHs appears to be suitable for practical work and assures that the best microbial candidates for soil biorem-ediation are chosen.

Acknowledgements

The authors would like to thank the Ministry of Education, Science and Technological Development, Republic of Serbia for support of this study in the frame of National Project III43004.

Figure 3. GC-FID chromatograms of PAH extracts before (0 day) and after (30 days) biodegradation by Planomicrobium sp. strain RNP01.

Figure 4. GC-FID chromatograms of PAH extracts before (0 day) and after (30 days) biodegradation by Rhodococcus sp. strain RNP05.


47

REFERENCES

[1] K. Kim, S.A. Jahan, E. Kabir, R.J.C. Brown, Environ. Int. 60 (2013) 71–80

[2] A. Nzila, Environ. Pollut. 239 (2018) 788-802

[3] B.K. Behera, A. Das, D.J. Sarkar, P. Weerathunge, P.K. Parida, B.K. Das, P. Thavamani, R. Ramanathan, V. Bansal, Environ. Pollut. 241 (2018) 212-233

[4] United States Environmental Protection Agency, Toxic Release Inventory Public Data Office of Environmental Information, https://www.epa.gov/sites/production/files/ /documents/2000_national_analysis_executive_summary.pdf [accessed 10 January 2019].

[5] European Parliament and of the Council No 1272/2013 amending Annex XVII to Regulation (EC) No 1907/2006 on the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) as regards polycyclic aromatic hydrocarbon. Official Journal of the European Union L (2013) 328/69

[6] European Commision, European Waste Catalogue, https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/ /?uri=CELEX:02000D0532-20150601&from=EN [accessed 10 January 2019].

[7] K. Ramadass, M. Megharaj, K.R. Venkateswarlu Naidu, Int. J. Environ. Sci. Technol. 13 (2016) 2863–2874

[8] J. Avdalović, A. Đurić, S. Miletić, M. Ilić, J. Milić, M.M. Vrvić, Waste. Manage. Res. 34 (2016) 734–739

[9] T. Jednak, J. Avdalović, S. Miletić, L. Slavković-Beškoski, D. Stanković, J. Milić, M. Ilić, V. Beškoski, G. Gojgić-Cvijović, M.M. Vrvić, Int. Biodeterior. Biodegrad. 122 (2017) 47-52

[10] L.A. Juhasz, S. Aleer, E.M. Adetutu, Int. Biodeter. Biodegr. 95 (2014) 320-329.

[11] M. Wu, L. Chen, Y. Tian, Y. Ding, W. Dick, Environ. Pollut. 178 (2013) 152-158

[12] M. Crampon, F. Bureau, M. Akpa-Vinceslas, J. Bodilis, N. Machour, F. Le Derf, F. Portet-Koltalo, Environ. Sci. Pollut. Res. 21 (2014) 8133–8145

[13] J. Milic, J. Avdalovic, T. Solevic-Kundsen, G. Gojgic-Cvijovic, T. Jednak, M.M. Vrvic, Chem. Ind. Chem. Eng. Q. 22 (2016) 293−299

[14] K. Kubota, D. Koma. Y. Matsumiya, S.Y. Chung, M. Kubo, Biodegradation 19 (2008) 749-757

[15] X. Zhang, L. Chen, X. Liu, C. Wang, X. Chen, G. Xu, K. Deng, Environ. Sci. Pollut. Res. 21 (2014) 8198–8205

[16] F. Abbondanzi, A. Cachada, T. Campisi, R. Guerra, M. Raccagni, A. Iacondini, Chemosphere 53 (2003) 889-897

[17] J.J.V. van der Waarde, E.J. Dijkhuis, M.J.C. Henssen, S. Keuning, Enzyme assays as indicators of biodegradation. In W.J. van den Brick, R. Bosman, F. Arendt (Eds.), Contaminated Soil ’95, Kluwer Academic Publishers, Dordrecht, 1995, pp. 1377-1378

[18] H. Al-Nasrawi, J. Bioremed. Biodegrad. 3 (2012) 147-152

[19] S. Roy, D. Hens, D. Biswas, R. Kumar, World J. Microb. Biot. 18 (2002) 575-581

[20] A.P. Mariano, D.M. Bonotto, D.F. Angelis M.P.S. Pirollo, J. Contiero, Brazil. J. Chem. Eng. 25 (2008) 269-274

[21] E.D. Bidoia, R.N. Montagnolli, P.R.M. Lopes, Microbial biodegradation potential of hydrocarbons evaluated by colorimetric technique: a case study, In A. Mendez-Vilas (Ed.), Current Research Technology and Education Topics. Applied Microbiology and Microbial Biotech-nology A. Vol 2, Microbiology Book Series 2, Formatex, Extramadura, 2010, pp. 1277-1288

[22] M. Mathew, J.P. Obbard, J. Biotechnol. Lett. 23 (2001) 227.


48

MARIJA LJEŠEVIĆ1 JELENA MILIĆ1

GORDANA GOJGIĆ-CVIJOVIĆ1 TATJANA ŠOLEVIĆ KNUDSEN1

MILA ILIĆ1 JELENA AVDALOVIC1

MIROSLAV M. VRVIĆ1,2 1Institut za hemiju, tehnologiju i

metalurgiju, Univerzitet u Beogradu, Institut od nacionalnog značaja,

Njegoševa 12, 11000 Beograd, Srbija 2Brem Group, Ulica Oslobođenja 39b,

11090 Beograd, Srbija l

NAUČNI RAD

PROCENA SKRINING TESTOVA ZA ODREĐIVANJE BAKTERIJSKOG POTENCIJALA ZA DEGRADACIJU POLICIKLIČNIH AROMATIČNIH UGLJOVODONIKA

Degradacija policikličnih aromatičnih ugljovodonika (polycyclic aromatic hydrocarbons - PAH) ispitivana je testovima u mikrotitar pločama pomoću dva bakterijska soja Plano-microbium sp. RNP01 i Rhodococcus sp. RNP05, u periodu od 30 dana. Rezultati ispi-tivanja su ukazali da se kombinacijom određenih bioloških i biohemijskih karakteristika mogu kreirati dobri indikatori u proceni bakterijskog degradacionog potencijala. Tri testa, 2,6-DCPIP test, test dehidrogenazne aktivnosti i test rasta na ugljovodonicima pokazali su da Rhodococcus sp. RNP05 ima veći potencijal za degradaciju PAH jednjenja u odnosu na Planomicrobium sp. RNP01. Razlike u početnoj i krajnjoj optičkoj gustini i specifične konstante rasta bile su značajno više (r = 0,995, P < 0,05) u testu sa Rhodo-coccus sp. RNP05 na svim testiranim supstratima, u poređenju sa Planomicrobium sp. RNP01, a dobijeni rezultati su potvrđeni gasno-hromatogafskom-FID analizom. Dehidro-genazna aktivnost soja Rhodococcus sp. RNP05 bila je viša u odnosu na Planomic-robium sp. RNP01 (r = 0,9995, P < 0,05) i u pozitivnoj korelaciji sa testom rasta na ugljo-vodonicima (r = 0,999, P < 0,05, za Rhodococcus sp. RNP05, r = 0,986, P < 0,05 za Planomicrobium sp. RNP01). Rezultati prikazanog istraživanja ukazuju na to da se kom-binacija ovih testova može koristiti za određivanje bioremedijacionog potencijala za raz-gradnju PAH jedinjenja u zemljištu zagađenim naftom, pri čemu se dobija i mogućnost testiranja velikog broja bakterijskih sojeva.

Ključne reči: rast na ugljovodonicima, PAH, skrining testovi, dehidrogenazna aktivnost, Planomicrobium, Rhodococcus.




49

MOHD AZAHAR MOHD ARIFF

NORFAZILAH ABDULLAH

Faculty of Chemical Engineering, Universiti Teknologi MARA (UiTM)

Cawangan Pulau Pinang, Permatang Pauh, Pulau Pinang,

Malaysia

SCIENTIFIC PAPER

UDC 66.063.4:582.929.4:004.4’22

OPTIMIZATION OF REFLUX EXTRACTION FOR CAT’S WHISKERS LEAVES EXTRACT USING RESPONSE SURFACE METHODOLOGY

Article Highlights • Optimizing antioxidants from Cat’s whiskers leaves using reflux extraction is still

uncertain • Our findings provide optimum conditions for extracting antioxidants from cat’s whis-

kers leaves • The study on optimization relates to all parameters and not one factor at one time

(OFAT) Abstract

Cat’s whiskers (Orthosiphon stamineus) is a medicinal plant which comprises several dynamic pharmacological properties such as anti-inflammatory, anti-oxidant and antibacterial. In small and medium-scale industries, conventional reflux extraction method is favored as compared to other non-conventional ext-raction methods due to cost effectiveness and simple operating procedures. In this study, response surface methodology (RSM) was applied to optimize the reflux conditions for extraction of cat’s whiskers leaves in order to achieve a high content of antioxidant activity in the extracts. Central composite experi-mental design (CCD) with three factors and three levels was employed to consider the effects of the operation conditions. Antioxidant activity of the ext-racts were based on free radical scavenging activity (DPPH assay) and were analyzed using a UV-Vis spectrophotometer. Based on RSM, the antioxidant activity could be maximized when the operation conditions were 125 µm for particle size, 1.5:20 for sample-to-solvent ratio, and 2 h for extraction time. Under these optimal conditions, the predicted value of the antioxidant activity was compared with the actual, and the mean error was 0.46%. This indicates the suitability of the model for optimizing the conditions for the reflux extraction of cat’s whiskers leaves.

Keywords: reflux extraction, cat’s whiskers leaves, Orthosiphon stami-neus, antioxidant activity, response surface methodology, optimization.

The use of plants for healing purposes has been practiced by people from all over the world since ancient times. The World Health Organization stated that 80% of the population in emerging countries relies on herbal medicines as a remedy for basic health treatment and 85% of this herbal medicine Correspondence: M.A.M. Ariff, Faculty of Chemical Engineering, Universiti Teknologi MARA (UiTM) Cawangan Pulau Pinang, 13500 Permatang Pauh, Pulau Pinang, Malaysia. E-mail: [email protected] Paper received: 28 February, 2019 Paper revised: 10 July, 2019 Paper accepted: 21 August, 2019


originates from plant extracts [1]. In Malaysia itself, it has been estimated that the market for herbal and natural products is worth USD 1.4 million per year [2].

One of such herbs that is being widely used and receiving greater attention among people is cat’s whiskers, or scientifically known as Orthosiphon sta-mineus. Based on established scientific studies, this herb is associated with dynamic pharmacological pro-perties such as anti-inflammatory, antioxidant, and antibacterial [3]. Cat’s whiskers consist of numerous bioactive compounds such as rosmarinic acid, sinen-setin, eupatorin and 3′-hydroxy-5,6,7,4′-tetrameth-oxyflavone [3]. In cat’s whiskers, phenolic compounds

M.A.M. ARIFF, N. ABDULLAH: OPTIMIZATION OF REFLUX EXTRACTION... Chem. Ind. Chem. Eng. Q. 26 (1) 49−57 (2020)

50

are among the most substantial bioactive compounds which determine the antioxidant activity of the crude extract [4]. The leaves have the highest antioxidant properties compared to the other part of the plant due to their greater phenolic fractions [2]. According to Asif (2015), plants that contain phenolic compounds are reported to have strong antioxidant properties [5].

In Asian and European countries, cat’s whiskers leaves are often transformed into herbal drink, namely Java tea or “teh misai kucing” [3]. According to Doleč-ková (2012), the consumption of Java tea can facil-itate body detoxification [6]. It is very popular for its diuretic effect, which is more tenacious than most other natural diuretics [7]. Traditionally, the leaves have been used to treat various ailments including cancer, arthritis, hypertension and renal stones [3].

These properties made the herb to be con-sumed as an energy drink for promoting and main-taining general body health and as a remedy to treat various ailments such as kidney related diseases, bladder inflammation, gout and diabetes [8,9].

Extraction and analysis involving plant matrices are placed under scrutiny, which are crucial for the quality control, modernization and advancement of herbal formulations [1]. According to Blicharski and Oniszczuk (2017), the ideal extraction process has to be safe, simple, reproducible, inexpensive and appro-priate for industrial applications [10]. Usually, the con-ventional extraction method is mostly used on a small manufacturing enterprise (SME) level [11]. The con-ventional method, reflux extraction, has several advantages, being efficient, cost effective and has simple operating procedures [12].

In order to achieve higher antioxidant extraction, it is necessary to conduct optimization for the most relevant operational parameters [1]. In this study, the optimization of the extraction process was achieved through response surface methodology (RSM) based on a central composite design (CCD). RSM can be a convenient tool for investigating the effect of several variables that influence the responses by varying them simultaneously and carrying out a limited num-ber of experiments [13]. In the case of not using an optimization, it will cause unnecessary waste to be generated from the extraction process, thus increas-ing the energy consumption and production cost [14].

This study was conducted with the aim to opti-mize the reflux extraction conditions by studying the effect of particle size, sample-to-solvent ratio and dur-ation of extraction, to allow for the maximum extract-ion of antioxidants from cat’s whiskers leaves using RSM.


Preparation of plant materials

Fresh plant samples of cat’s whiskers leaves were bought from a local supplier in Pulau Pinang. The leaves were cleaned and washed with water and dried in an oven for 48 hours at 40 °C. The moisture content of the leaves were calculated to ensure that it is below 10% [15]. Eq. (1) shows the formula to cal-culate the moisture content of plant material. The dried leaves were then grinded and sieved into dif-ferent sizes:

( ) −=−

1 2

1 0

Moisture content % 100M MM M

(1)

where M1 is the mass of sample before drying, M2 is

the mass of sample after drying and M0 is the mass of the tray.

Extraction process

The extraction was performed by dissolving different ratios of plant sample to solvent (1.0:20, 1.5:20, 2.0:20, 2.5:20 and 3.0:20) with different par-ticle sizes (63, 125, 250, 500 and 1000 µm) and at different extraction time (1, 2, 3, 4 and 5 h). The ext-raction occurred at a constant temperature of 85 °C by using pure water as solvent. The extraction was done below the boiling point of water to avoid rapid evaporation of solvent when extracting at a longer time (5 h). After extraction, the extracts were cooled at room temperature and filtered using a suction pump and Whatman filter paper.

Determination of antioxidant activity

The free radical scavenging activity of the fluid extracts prepared by reflux extraction were evaluated using 1,1-diphenyl-2-picrylhydrazyl (DPPH) assay. 0.6 mM solution of DPPH in ethanol was prepared by diluting 0.1183 g of DPPH into ethanol in a 500 mL volumetric flask. The DPPH solution was shaken and incubated in a dark place for 30 min to stabilize the reaction. About 9 mL of this solution was then added to 1 mL of each extract. The mixture was shaken and incubated at room temperature for 30 min. As for the control, 2.5 mL of distilled water was diluted with ethanol in a 25 mL volumetric flask. The mixture was left incubated in a dark room for 30 min. After 30 min, 9 mL of DPPH solution was added to 1 mL of control solution. The mixture was shaken and incubated again for another 30 min. In this experiment, distilled water was used as blank. The absorbance of the samples was measured using UV-Vis spectrophoto-meter at 517 nm of wavelength and the antioxidant


51

activity of the extract was calculated using Eq. (2) [16]:

−= C S

C

Antioxidant activity (%) 100A A

A (2)

where, AS is the absorbance of the sample and AC is the absorbance of the control.

Experimental design

RSM was used to determine the optimal ext-raction conditions for maximizing the antioxidant act-ivity of cat’s whiskers leaves extract. The CCD expe-rimental design with three levels and three factors was employed to examine the extraction variables by generating 20 running sets of extraction. Table 1 shows the range of parameters that were used in the experiment [17].

Table 1. The range of parameters [23]

No. Factor Level of Range (coded)

-1.68179 -1.00000 0.00000 +1.00000 +1.68179

1 Particle size, µm

63 125 250 500 1000

2 Ratio of sample to

solvent

1.0:20 1.5:20 2.0:20 2.5:20 3.0:20

3 Duration of extraction,

hour

1 2 3 4 5

Statistical analysis

The Design Expert software, version 11 (DE11) [1], was used for data analysis. Analysis of variance (ANOVA) and response surface analysis were used to determine the regression coefficients and statistical significance of the model terms and to fit the mathe-matical models of the experimental data that aimed to optimize the overall region for response variables.


Antioxidant analysis using DPPH radical scavenging method

Table 2 shows the results for antioxidant activity of cat’s whiskers leaves extract for 20 samples of runs. The experiment set-up was carried out as per experimental design stated previously and it was found that the lowest and highest antioxidant activity was achieved at 50.8 and 87%, respectively.

Model fitting and statistical significance analysis

Table 3 shows the regression score of the anti-oxidant activity. Coefficient of determination (R2) and the significance of lack of fit were used as a guideline to determine the fitness and adequacy of the model. The closer the R2 value to one, the better the empi-rical model fits the actual data [18]. Based on Table 3, R2 of the model is 0.9721 which is close to one. R2 of 0.9721 means that 97.21% of the variance in pre-

Table 2. Antioxidant activity of cat’s whiskers leaves extract for each run

Run A: Particle size (um) B: Ratio of sample to solvent (g) C: Duration of extraction (h) Antioxidant activity (%)

1 250 2.0:20 5 86.2

2 125 2.5:20 4 61.1

3 250 1.0:20 3 59.7

4 500 2.5:20 4 74.3

5 125 1.5:20 2 87.0

6 63 2.0:20 3 70.0

7 250 2.0:20 1 84.8

8 250 2.0:20 3 58.8

9 125 1.5:20 4 78.8

10 250 2.0:20 3 54.3

11 500 1.5:20 4 56.4

12 250 3.0:20 3 50.8

13 500 1.5:20 2 68.4

14 125 2.5:20 2 59.8

15 1000 2.0:20 3 57.4

16 250 2.0:20 3 59.3

17 250 2.0:20 3 60.9

18 250 2.0:20 3 58.3

19 250 2.0:20 3 57.1

20 500 2.5:20 2 73.4


52

dicted data is determined by the actual (experimental) data. The predicted and adjusted R2 obtained is 0.8514 and 0.9469 respectively. These values are in reasonable agreement as the difference between them is less than 0.2.

Table 3. The regression score

Std. Dev. Mean C.V., % R2 Adjusted R2 Predicted R2

2.59 65.84 3.94 0.9721 0.9469 0.8514

Table 4 shows the ANOVA results for antioxid-ant activity of cat’s whiskers leaves extract. The P- -values were utilized as an indicator for determining the significance of each coefficient, which described the interaction patterns between independent vari-ables [18]. P-values less than 0.05 indicate model terms are significant and P-values greater than 0.1000 indicate the model terms are not significant [13,18].

Table 4. ANOVA results for the response surface quadratic model of cat’s whiskers leaves antioxidant activity

Source Sum of squares

df Mean

square F-value p-value

Signific-ance

Model 2341.33 9 260.15 38.64 < 0.0001 Signific-ant

A-Particle size

91.71 1 91.71 13.62 0.0042

B-Ratio sample to solvent

100.07 1 100.07 14.86 0.0032

C-Durat-ion of ext-raction

17.92 1 17.92 2.66 0.1338

AB 574.6 1 574.6 85.36 < 0.0001

AC 2.21 1 2.21 0.33 0.5797

BC 62.72 1 62.72 9.32 0.0122

A² 70.7 1 70.7 10.5 0.0089

B² 8.6 1 8.6 1.28 0.2847

C² 1418.83 1 1418.83 210.76 < 0.0001

Residual 67.32 10 6.73

Lack of fit 42.07 5 8.41 1.67 0.2944 Not sig-nificant

Pure error

25.25 5 5.05

Cor total 2408.65 19

Based on Table 4, particle size (A) and sample- -to-solvent ratio (B) had significant (p < 0.05) impact on the antioxidant activity of the extract. In addition, the interaction between particle size and sample-to- -solvent ratio (AB) as well as interaction between sample-to-solvent ratio and extraction time (BC) were also significant. Of all the three parameters studied in

this experiment, sample-to-solvent ratio played the most important role in the reflux extraction of cat’s whiskers leaves followed by particle size and extract-ion time.

The F-value of 38.64 implies the model is sig-nificant. There is only a 0.01% chance that an F-value this large could occur due to noise. The lack of fit F-value of 1.67 implies the lack of fit is not significant relative to the pure error. There is a 29.44% chance that a lack of fit F-value this large could occur due to noise. Non-significant lack of fit is good and desirable to ensure the fitness of the model. The predicted model generated by RSM was as in Eq. (3):

Antioxidant activity (%) = +58.08 – 2.59A – 2.71B – -1.15C + 8.47AB - 0.5250AC + 2.80BC + 2.21A2 – (3) -0.7727B2 + 9.92C2

Response surface of antioxidant activity of reflux extraction

Figure 1 shows the effect particle size and sample-to-solvent ratio on the antioxidant activity of cat’s whiskers leaves extract. Based on Figure 1, it was found that as particle size of plant sample inc-reased, the antioxidant activity decreased. This trend is consistent with those reported in the literature. This is because a smaller particle size has a large total surface area that provide a great mass transfer rate between the plant matrix and the solvent [19]. How-ever, the efficiency of the extraction could be disturbed if the particle used is too small in size [20]. During reflux extraction, very small particle size has limited time contact between solute and solvent as they tend to float on the surface of solvent, thus resulting in the lower antioxidant activity of the extract [20].

Similarly, it was found that the antioxidant act-ivity increased when the sample-to-solvent ratio inc-reased. The highest antioxidant activity achieved for the extract was 87%. This result is consistent with the previous study which stated that increasing the amount of water as solvent will increase the bioactive component of the plant extract [21]. This is because higher water content during extraction will cause the plant material to swell which leads to longer contact time between the solid matrix of the plant and the solvent, thus resulting in an increase of bioactive compounds [21,22].

Figure 2 shows the relationship of particle size and duration of extraction on the antioxidant activity of cat’s whiskers leaves extract. Based on Figure 2, maximum antioxidant activity (87%) was achieved at smaller particle size and shorter extraction time. Although prolong exposure of solute in the solvent provides enough time contact of solid matrix with


53

water, it was believed that extensive extraction time under the high boiling point of water could contribute to the degradation of bioactive compounds of the plant sample, thus decreasing its antioxidant activity [1,23]. Moreover, prolonged extraction time contri-butes to high cost and large energy consumption [1].

Figure 3 shows the relationship between the sample-to-solvent ratio and duration of extraction on the antioxidant activity of cat’s whiskers leaves ext-ract. Based on the response surface plot in Figure 3,

maximum antioxidant activity (87%) was achieved at a lower sample-to-solvent ratio and shorter extraction time. Lower sample-to-solvent ratio means more water was used as solvent during the extraction pro-cess. A lower sample-to-solvent ratio enhances the rate of diffusion due to the increase in concentration gradient which accelerates the extraction of the plant sample with water [24]. However, the antioxidant act-ivity of the plant extract will not increase any further once equilibrium is reached [24].

Design-Expert® SoftwareTrial VersionFactor Coding: Actual

Antioxidant (%)Design points above predicted value

50.8 87

X1 = A: Particle sizeX2 = B: Ratio sample to solvent

Actual FactorC: Duration of extraction = -1

-1

-0.5

0

0.5

1

-1

-0.5

0

0.5

1

50

60

70

80

90

Antio

xidan

t (%)

A: Particle sizeB: Ratio sample to solvent

Figure 1. Response surface plot for the effect of particle size and sample-to-solvent ratio on antioxidant activity.


Antioxidant (%)Design points above predicted valueDesign points below predicted value

50.8 87

X1 = A: Particle sizeX2 = C: Duration of extraction

Actual FactorB: Ratio sample to solvent = -1

-1 -0.5

0 0.5

1

-1 -0.5

0 0.5

1

50

60

70

80

90

Antio

xidan

t (%)

A: Particle size

C: Duration of extraction Figure 2. Response surface plot for the effect of particle size and duration of extraction.


54

Optimization of reflux extraction conditions for antioxidant activity

Numerical optimization in DE11 software was performed to identify the optimal conditions of inde-pendent variables with a desirable response goal, which is the antioxidant activity. Table 5 shows the experimental range and level of independent vari-ables involved in this study. Each coded value in RSM represents the actual parameter value for the extraction of cat’s whiskers leaves.

Table 5. Experimental range and levels of independent variables

No. Factor Level of range (coded)

-1.6817 -1 0 1 1.6817

1 A / μm 63 125 250 500 1000

2 B 1:20 1.5:20 2:20 2.5:20 3:20

3 C / h 1 2 3 4 5

Table 6 shows the actual and predicted values for antioxidant activity of cat’s whiskers leaves ext-ract. The predicted and actual values generated by RSM allow the researchers to ensure the expe-rimental data obtained from the analysis are in the range of the predicted values.

Figure 4 shows the graph of predicted versus actual value for the percentage of antioxidant activity of cat’s whiskers leaves extract. Figure 4 depicts that there are tendencies in the linear regression fit and the model adequately explains the experimental range studied. The actual percentage of antioxidant activity value is the measured result for a specific run and the predicted value is evaluated from the independent variables in the CCD model [13].

Table 6. Actual and predicted values for antioxidant activity of cat’s whiskers leaves extract

Run Independent variable Antioxidant activity, %

Particle size, μm Mass ratio of sample to solvent Duration of extraction, h Actual Predicted

1 250 2.0:20 5 86.2 84.2

2 125 2.5:20 4 61.1 63.0

3 250 1.0:20 3 59.7 60.5

4 500 2.5:20 4 74.3 73.8

5 125 1.5:20 2 87.0 86.6

6 63 2.0:20 3 70.0 68.7

7 250 2.0:20 1 84.8 88.1

8 250 2.0:20 3 58.8 58.1

9 125 1.5:20 4 78.8 79.8

10 250 2.0:20 3 54.3 58.1


Antioxidant (%)Design points above predicted valueDesign points below predicted value

50.8 87

X1 = B: Ratio sample to solventX2 = C: Duration of extraction

Actual FactorA: Particle size = -1

-1 -0.5

0 0.5

1

-1 -0.5

0 0.5

1

50

60

70

80

90

Antio

xidan

t (%)

B: Ratio sample to solventC: Duration of extraction

Figure 3. Response surface plot for the effect of sample-to-solvent ratio and duration of extraction.


55

Table 6. Continued

Run Independent variable Antioxidant activity, %

Particle size, μm Mass ratio of sample to solvent Duration of extraction, h Actual Predicted

11 500 1.5:20 4 56.4 56.6

12 250 3.0:20 3 50.8 51.3

13 500 1.5:20 2 68.4 65.6

14 125 2.5:20 2 59.8 58.7

15 1000 2.0:20 3 57.4 60.0

16 250 2.0:20 3 59.3 58.1

17 250 2.0:20 3 60.9 58.1

18 250 2.0:20 3 58.3 58.1

19 250 2.0:20 3 57.1 58.1

20 500 2.5:20 2 73.4 71.5

Design-Expert® SoftwareTrial Version

Antioxidant

Color points by value ofAntioxidant:50.8 87

Actual

Pred

icted

Predicted vs. Actual

50

60

70

80

90

50 60 70 80 90

Figure 4. Graph of predicted versus actual values for antioxidant activity of cat’s whiskers leaves extract.

Figure 5 shows the predicted value of particle size, ratio of sample to solvent and duration of ext-raction from the optimization of reflux extraction con-dition using RSM. Based on Figure 5, the optimal conditions developed by the software for maximizing

the antioxidant activity of the extract are: particle size of 250 μm, 1.5:20 (15 g) sample-to-solvent ratio and 2 h of extraction time. Under these optimal conditions, the predicted antioxidant activity is 86.64%.

Figure 5. Predicted value of particle size, sample-to-solvent ratio and duration of extraction from optimized

reflux extraction conditions using RSM.


56

Validation of the model

The experiment was repeated twice using the recommended optimal conditions to achieve the high-est antioxidant activity of the extract. Based on the recommended reflux extraction optimal conditions, the predicted value of the antioxidant activity is 86.64%. As shown in Table 7, the experimental values obtained were 87.1 and 86.3%, respectively. The inconsistent percentage errors for both runs were caused by the noise during the extraction process and the analysis of the extract using a UV-Vis spec-trophotometer. However, the mean percentage error calculated is 0.46% which is well below 5%, thus ensuring the reliability and accuracy of the model.

Table 7. Verification of the optimum conditions suggested by RSM

Run A

μm B g

C h

Value, %

Experimental Predicted Error

1 125 20 g 2 87.1 86.64 0.53

2 125 20 g 2 86.3 86.64 0.39

Mean error 0.46%

CONCLUSION

This study has been successfully performed to determine the optimum conditions for extracting the highest antioxidant activity of cat’s whiskers leaves extract using the reflux extraction method. From the results obtained, particle size and sample-to-solvent ratio significantly influenced (p < 0.05) the reflux ext-raction of antioxidant activity of cat’s whiskers leaves while extraction time did not show any significant impact (p > 0.05) on the antioxidant activity obtained. Among the three parameters studied, sample-to-sol-vent ratio played the most important role in the reflux extraction of cat’s whiskers leaves followed by particle size and extraction time. The optimal conditions gen-erated by RSM were: particle size of 250 μm, 1.5:20 sample-to-solvent ratio and 2 h of extraction time. At these conditions, the expected antioxidant activity was 86.64%. Two experiments were conducted using the optimal conditions as per generated for verific-ation. It was found that the highest antioxidant activity of the extract for both experiments were 87.1 and 86.3%, respectively, compared to the unoptimized conditions which was 86.2%. Furthermore, the sample-to-solvent ratio has been reduced from 2.0:20 to 1.5:20 and the extraction time was reduced as well from 5 to 2 h, compared to the unoptimized con-ditions.

Acknowledgement

This work was financially supported by Universiti Teknologi MARA, Malaysia.

REFERENCES

[1] A. Ghasemzadeh, H.Z.E. Jaafar, Sci. World J. 2014 (2014)

[2] N.A. Amir Hamzah, N.A. Morad, M.F.M. Nordin, A.N. Ilia Anisa, Y.A.M. Yusof, Aust. J. Basic Appl. Sci. 11 (2017) 15–21

[3] F.F.S.R. Al-suede, M.B.K. Ahamed, A.S.A. Majid, H.M. Baharetha, L.E.A. Hassan, M.O.A. Kadir, Z.D. Nassar, A.M.S.A. Majid, J. Evidence-Based Complementary Altern. Med. 2014 (2014) Article ID 396016

[4] W. Aida, Sci. York 578(4) (2011) 571–578

[5] M. Asif, Chem. Int. 1 (2015) 35–52

[6] I. Dolečková, L. Rárová, J. Grúz, M. Vondrusová, M. Strnad, V. Kryštof, Fitoterapia 83 (2012) 1000–1007

[7] O.Z. Ameer, I.M. Salman, M.Z. Asmawi, Z.O. Ibraheem, M.F. Yam, J. Med. Food 15 (2012) 678-690

[8] S.K. Ho, C.P. Tan, Y.Y. Thoo, F. Abas, C.W. Ho, Mole-cules 19 (2014) 12640–12659

[9] E.A. Engku Hasmah, S. Ahmad Tarmizi, N. Noor Isma-waty, O. Abdul Ghani, Acta Hortic. 1012 (2013) 837–842

[10] T. Blicharski, A. Oniszczuk, Open Chem. 15 (2017) 34–45

[11] A. Nn, Med. Aromat. Plants 4 (2015) 3–8

[12] L.S. Chua, N.A. Latiff, M. Mohamad, J. Appl. Res. Med. Aromat. Plants 3 (2016) 64–70

[13] M. Demirel, B. Kayan, Int. J. Ind. Chem. 3 (2012) 1–10

[14] L.S. Badwaik, K. Prasad, S.C. Deka, Int. Food Res. J. 19 (2012) 341–346

[15] J. Reeb, M. Milota, in Proceedings of Western Dry Kiln Association Conference, 50th Proceedings, Portland, OR, USA, 1999, pp. 66–74

[16] T.C. Shekhar, G. Anju. Am. J. Ethnomed. 1 (2014) 244– -249 (http://www.ajethno.com)

[17] P. Mag, J.H. Kim, H.K. Shin, C.S. Seo, Phcog Mag. 10 (2014) 606-613

[18] R. Kumar, S. Singh, M. Singh, Resour. Technol. 2 (2016) 148–157

[19] S.A. Makanjuola, Food Sci. Nutr. 5 (2017) 1179–1185

[20] A. Yeop, J. Sandanasam, S.F. Pan, S. Abdulla, M.M. Yusoff, J. Gimbun, FluidsChE 2017, in Proceedings of MATEC Web Conf. 111, Kota Kinabalu, Sabah, Malaysia, 2017, pp. 1–5

[21] I. Soraya, C. Sulaiman, M. Basri, H. Reza, F. Masoumi, W.J. Chee, S.E. Ashari, Chem. Cent. J. (2017) 1–11

[22] S. Wang, J. Gao, Y. Chen, J. Zhou, X. Yong, X. Liu, Y. Zhang, X. Liu, Y. Sun, T. Zheng, Adv. Mater. Res. 955- -959 (2014) 180-186

[23] L. Nor Afiqah, M.N. Noorzalila, ESTEEM Acad. J. 14 (2018) 32–41

[24] S. Norshazila, C.N. Koy, O. Rashidi, L.H. Ho, I. Azrina, R.A. Nurul Zaizuliana, Z. Zarinah, Sains Malays. 46 (2017) 231–237.


57

MOHD AZAHAR MOHD ARIFF NORFAZILAH ABDULLAH

Faculty of Chemical Engineering, Universiti Teknologi MARA (UiTM)

Cawangan Pulau Pinang, Permatang Pauh, Pulau Pinang, Malaysia

NAUČNI RAD

OPTIMIZACIJA REFLUKSNE EKSTRAKCIJE JAVANSKOG ČAJA POMOĆU METODOLOGIJE POVRŠINE ODZIVA

Javanski čaj ili mačja trava (Orthosiphon stamineus) je lekovita biljka koja sadrži neko-liko dinamičkih farmakoloških svojstava, kao što su protivupalna, antioksidativna i anti-bakterijska. U malim i srednjim proizvodnim kapacitetima radije se primenjuje konven-cionalna metoda refluksne ekstrakcije nego li druge nekonvencionalne metode zbog ekonomičnosti i jednostavnih operativnih procedura. U ovom istraživanju, primenjena je metodologija površine odziva (RSM) da bi se optimizovali refluksni uslovi za ekstrakciju iz lišća javanskog čaja kako bi se postigao visok sadržaj antioksidativne aktivnosti u ekstraktima. Centralni kompozitni eksperimentalni plan (CCD) sa tri faktora na tri nivoa je korišćen za razmatranje efekata radnih uslova. Antioksidativna aktivnost ekstrakata zasnovana je na aktivnosti uklanjanja slobodnih radikala (DPPH test) i analizirana je koristeći UV-Vis spektrofotometar. Na osnovu RSM, maksimalna antioksidativna aktiv-nost se može postići pri veličini čestica 125 μm, odnosu biljni material/rastvarač 1,5:20 i vremenu ekstrakcije 2 h. Pod ovim optimalnim uslovima, predviđena vrednost antioksi-dativne aktivnosti je upoređena sa stvarnom i srednja greška je bila 0,46%. Ovo ukazuje na pogodnost modela za optimizaciju uslova refluksne ekstrakcije lišća javanskog čaja.

Ključne reči: Ekstrakcija refluksa, lišće javanskog čaja, Orthosiphon stamineus, antioksidativna aktivnost, metodologija površine odziva.




59

EDUARDO RAMOS BRAGA

GEORGE DE SOUZA MUSTAFA

DANILO DE AGUIAR PONTES LUIZ ANTÔNIO MAGALHÃES

PONTES

Federal University of Bahia, Postgraduate Program of Chemical

Engineering, Federação, Salvador - BA, Brazil

SCIENTIFIC PAPER

UDC 547.391.1+66.012:547.426.1

ECONOMIC ANALYSIS AND TECHNICALITIES OF ACRYLIC ACID PRODUCTION FROM CRUDE GLYCEROL

Article Highlights • Crude glycerol was used as raw material for the production of acrylic acid, sub-

stituting the propene • The study was carried out with the oxidative dehydration reaction in a single step • The process was economically feasible and produced glacial acrylic acid 98 mass% Abstract

Glycerol is the main byproduct of industrial biodiesel plants, and new techno-logical routes using it as feedstock have been studied, due to the increase in world biofuel production. One of the possible applications is in the production of acrylic acid, a product with several industrial applications. This study ana-lyzed a new process of converting crude glycerol, through purification for the removal of impurities, followed by oxidative dehydration reaction in a single step and purification until glacial acrylic acid specification standards are met. It was attested that the process is economically viable, with a payback period of 5 years for an NPV/Investment greater or equal to 2 and an IRR greater or equal to 10% per annum, or 4 years in case tax incentives offered for the dev-elopment of green technologies.

Keywords: acrylic acid, glycerol, purification, biodiesel, time.

The current concern of human society towards environmental protection has led to the development of sustainable technologies and products, encompas-sing waste management in industries. In this context, biofuels are highlighted as a more sustainable alter-native when compared to fossil fuels. Biofuels com-prise a source of renewable energy and release less greenhouse gases into the atmosphere. Amongst bio-fuels, biodiesel has drawn attention for it is produced from simple and abundant feedstock, such as veget-able oils, and animal fat.

When produced from biomass, biodiesel pre-sents a neutral carbon balance, which significantly contributes to the reduction of the emission of green-house gases. In transesterification, animal fat or veg-

Correspondence: E.R. Braga, Federal University of Bahia, Post-graduate Program of Chemical Engineering, R. Prof. Aristides Novis, 02, 2º andar, Federação, CEP 40210-630, Salvador - BA, Brazil. E-mail: [email protected] Paper received: 11 January, 2018 Paper revised: 10 June, 2019 Paper accepted: 25 August, 2019

https://doi.org/10.2298/CICEQ180111025B

etable oil is used as feedstock which reacts with a short chain alcohol, in the presence of a homogene-ous alkaline catalyst, generating, as the main product, biodiesel and glycerol as a low-value byproduct. In this process, 10 mass% glycerol is produced when accounted for the total biodiesel mass produced [1].

Besides, biodiesel can be produced from waste, which makes biodiesel an attractive and versatile con-tender for the renewable replacement of mineral diesel [2]

Yearly biodiesel production growth in the Amer-ican market has been steady, having recently reached 6.02 million m3 in 2017, against 5.93 million m3 in 2016 and 4.78 million m3 in 2015 [3]. State and federal incentives have pushed this rise in biodiesel production by a large extent, and although federal biodiesel tax credit expired in late 2016, it returned in 2017 after Congress voted for it [4]. The European Union is the world’s largest producer of biodiesel, having production reaching 15.29 million m3 in 2017, representing an increase of circa 3.88% compared to 2016 and circa 5.95%, compared to 2015 [5]. Ger-many is the largest biodiesel producer in the EU, fol-lowed by France, and Spain [5]. Since 2008, there

E.R. BRAGA et al.: ECONOMIC ANALYSIS AND TECHNICALITIES... Chem. Ind. Chem. Eng. Q. 26 (1) 59−69 (2020)

60

have been no records of biodiesel production on an industrial scale in Serbia, with the largest plant, Vic-toria Oil, operating only in 2007, producing biodiesel from rapeseed oil [6]. The decision to stop biodiesel production was motivated by the lack of government incentives and the soaring prices of oilseeds and edible oils on the global market [6].

Although glycerol has several industrial applic-ations in cosmetics, pharmaceuticals, and food, there is a need for the development of new markets follow-ing the increase in the world supply of this product. The upside of this greater supply is its decreased market value, making glycerol attractive for possible use as feedstock for the generation of higher added value products. Raw glycerol produced in a biodiesel plant undergoes an acidic treatment and neutralizat-ion in order to precipitate impurities, followed by ther-mal treatment in order to evaporate excess water and residual alcohol. The result from this treatment is crude glycerol 80 mass%, where parts of the impur-ities have been removed in the biodiesel plant itself [7]. Crude glycerol can be purified and used as feed-stock in glacial acrylic acid production, used in the production of superabsorbents, acrylonitrile, deter-gents, dispersants, flocculants, and thickener agents.

The German company BASF is the largest world producer of acrylic acid, having inaugurated their most recent investment in the area with an acrylic acid complex in Brazil in 2015. This complex is cap-able of producing 160 thousand metric tons of acrylic acid per annum, and supplies the whole Latin Amer-ican market [8].Other large producers of acrylic acid are French Arkema, American Dow Chemical, and Japanese Nippon Shokubai [9]. An annual growth of 1.5-2% is expected for acrylic acid demand in North America and Europe in the next 4 years [10].

Commercially, glacial acrylic acid is that which possesses a purity higher than 98 mass% [11]. The consolidated production route is from the gas phase oxidation of propene, an oil derivative, under an oxide catalyst based on Bi/Mo [12]. However, propene prices in the world market oscillate greatly, and might reach US$ 1100 per ton, while the market value of crude glycerol has been dropping steadily, reaching US$205 per ton in 2018 in the US [13]. This opens a major opportunity for acrylic acid production from gly-cerol.

The proposed route for acrylic acid production needs purified glycerol. The existing purification pro-cess consists of vacuum distillation, followed by ads-orption in activated charcoal for the removal of odors and impurities.

The acrylic acid production process can occur in two steps. The first one involves dehydrating glycerol 99.5 mass% to acrolein using gas phase acidic cat-alysts at temperatures ranging from 250 to 350 °C, at an absolute pressure of up to 500 kPa abs [14]. Tem-perature is an important variable for improving cat-alyst selectivity to acrolein. For this reaction, com-monly used catalysts include zeolites, heteropoly-acids, and metallic oxides [15]. For the second stage, acrolein oxidation, catalysts containing mixtures of molybdenum, vanadium, chromium, copper, tantalum, and niobium oxides, which may or may not be sup-ported in silica, have been demonstrated to be highly active and selective, with an acrylic acid yield close to 90% in gas phase at 300 °C [16].

New studies for acrylic acid production through the “one pot” reaction of glycerol to acrylic acid, called oxidative dehydration of glycerol in a single step have been performed. In this route, glycerol reacts with oxygen forming acrylic acid in a single step, as shown in Eq. (1):

+ → + Δ = −3 8 3 2 3 4 2 21 kJ

C H O O C H O 2H O, 322.82 mol

H (1)

In this process, bifunctional catalysts such as vanadium silicates are used, which are capable of converting 93.6% of glycerol into acrylic acid, at a temperature of 320 °C, in a single stage with 85.4% selectivity [17]

The purification of acrylic acid produced from purified glycerol in a single step reaction is a radical innovation, still unprecedented, which implies evalu-ating the possible contaminants present in the feed-stock, and obtaining a product that meets specific-ation standards. In this context, it was also verified that there is no technical-economical proposal in the literature of a process encompassing a crude glycerol purification step, followed by an oxidative dehydration reaction in a single step of purified glycerol, and puri-fication of crude acrylic acid into glacial acrylic acid.

Considering the increasing biodiesel production, and the excess glycerol on the market, this paper aims at studying a new acrylic acid production pro-cess, which meets the 98 mass% specification, using crude glycerol as feedstock, as well as the economic analysis including the purification process of crude glycerol, and purification of the resulting acrylic acid. Aspen HYSYS v8.0 was used for the simulation. The economic viability study took into account two main economical criteria: NPV/Total investment ≥2, and internal rate of return (IRR) ≥ discount rate used in the project.


61


Reaction characterization

For reaction characterization, several kinds of catalysts and their respective yields for the formation of acrylic acid by the oxidative dehydration reaction of glycerol in a single step were verified in literature. The main papers are presented in Table 1.

Table 1. Catalysts and reactional data for the oxidative dehyd-ration reaction of glycerol in a single step

Tempe-rature, °C

Catalyst Conversion to

glycerol, % Selectivity for acrylic acid, %

Ref.

350 SiW/Al2O3 e Mo3VOx

100 19.8 [18]

90 SiW/Al2O3 83.78 25.11 [19]

300 V-W oxides 90 26 [20]

275 V/Beta zeolite 75 26 [16]

250 MoVW-5 100 30.5 [21]

450 SiW/Al2O3 e Mo3Vox

100 46.2 [18]

250 HZSM-5 + MoVW-5

100 47.2 [20]

265 W–V–Nb 81 50.5 [22]

285 2.5 mass% PO3/ W2.2V0.4Nb2.4O14

100 59.2 [23]

300 FeVO4 100 76 [14]

320 Vanadium silicates

93.6 85.4 [17]

Amongst these papers, the one from Possato et al. [17] was selected for presenting a good glycerol conversion (93.6%) and high acrylic acid selectivity (85.4%). The results obtained by Paula et al. [17] using vanadium silicates as catalysts are presented in Table 2.

Table 2. Reactional data obtained by Paula et al. (2016) [17]

Parameter Selectivity of formed products, mol%

% Oxygen 20 100

Acrolein 6.2 3.8

Acrylic acid 18.7 85.4

Acetaldehyde 5.4 3.2

Propanal 4.8 2.3

Acetic acid 2.2 2.5

3-Hydroxy-propanal 1.5 0

Propanoic acid 0 1.4

Carbon balance 0.5 0.9

Glycerol conversion, % 63 93.6

Temperature, °C 320

Pressure, kPa abs 101.32

In Table 2, the carbon balance signifies the per-centage of carbon from glycerol, which was converted into liquid products, with the remaining carbon con-verted into carbon monoxide and dioxide. The per-centage of oxygen represents the amount of oxygen in the feed gas. At 20%, the feed accounts for 80% nitrogen (synthetic air). At 100%, a pure oxygen feed is considered. The vanadium silicate catalyst acting under a feed stream of pure oxygen presented higher selectivity to acrylic acid, so the reactor mass balance was calculated based on data at these conditions (WHSV = 2.05 h-1).

For this study, a production capacity of 160,000 ton per annum was considered. New plants around the world have been built for this capacity [8]. The assumed project factor is 0.98, thus considering that the plant will maintain normal operation for 98% of days in a year, resulting in a 258.6 kmol/h output of glacial acrylic acid.

Process simulation and dimensioning and main equipment cost

For process simulation, the commercial simul-ator Aspen HYSYS v8.0, by AspenTech was used, using the UNIQUAC (UNIversal QUAsiChemical) thermodynamic model. UNIQUAC was chosen based on the methodology proposed by Carlson [24], where this model is indicated for non-electrolyte systems (electrolytes which may be present in this step of crude glycerol purification are very diluted, which makes them irrelevant for system properties), for sys-tems working with polar molecules, for low pressure systems (lower than 1000 kPa), for systems working with liquid-liquid equilibrium, and also for systems which present some degree of polymerization in the vapor phase, as is the case of glycerol.

The equipment had its dimensioning following the methodology proposed by Seider et al. [25], with purchase prices from the American market in USD. Equipment prices were updated according to the average American inflation rate from December 2010 to December 2017, being 1.67% per annum [26].

Besides this, the equipment had a premium of 5% relative to logistics.

Total investment calculation

For total plant investment calculation, the indir-ect investment value was estimated, as well as direct investment and working capital. Direct investment corresponds to investments applied directly to the productive process, such as equipment, piping, assembly, instrumentation, etc. The criteria used for the direct investment estimate in accordance with Peters et al. [27] are presented in Table 3.


62

Table 3. Criteria for direct investment calculation [27]; M.P.E.C. – main pieces of equipment cost. Direct investment is the sum of all items

Parameter Value, % of M.P.E.C.

Purchased-equipment installation 47

Instrumentation and controls (installed) 18

Piping (installed) 66

Electrical (installed) 11

Buildings (including services) 18

Yard improvements 10

Service facilities (installed) 70

Land (if purchase is required) 6

Indirect investment corresponds to costs with construction, engineering, and supervision, including a reserve fund for inevitable contingencies. Indirect investment calculation criteria are also in accordance to Peters et al. [27], and are presented in Table 4.

Table 4. Criteria for indirect investment calculation [27]; I.I. – indirect investment; D.I. - direct investment. Indirect investment is the sum of all items

Parameter Value

Engineering and supervision 33% of M.P.E.C.

Construction expenses 41% of M.P.E.C.

Contractor’s fee 5% of (D.I. + I.I.)

Contingency 10% of (D.I. + I.I.)

Working capital is estimated to be 15% of total investment, Peters et al. [27]. At last, total investment is the sum of indirect and direct investment, and working capital.

Fixed cost calculation

Fixed costs are considered independently of volume and production quantity. For calculating this cost, the proposed Peters et al. [27] methodology is again used, where the parameters and criteria in Table 5 stand as follows, and whose operation man-power value was considered to be US$ 4,500/month per worker.

Net present value and internal rate of return

Net present value (NPV) is the sum, from start- -up day, of cash flow balance, discounting interest rate, according to Eq. (2):

= −+

0(1 )

nj

n

CPV I

i (2)

where Cj = cash flow; I = discount rate; n = number of time periods used for economic analysis; I = total investment.

Table 5. Criteria for fixed cost calculation [27]; O.M. - operation manpower; T.I. - total investment. A total of 35 plant operators was considered. This value is again in accordance to the Peters et al. [27] methodology for a process with good auto-mation. Fixed cost is the sum of all items

Item Criteria

Operation manpower US$ 4,500/month per worker

Direct supervisory and clerical labor

17.5% O.M.

Maintenance and repairs 6% T.I./year

Operating supplies 0.75% T.I./year

Laboratory charges 15% O.M.

Insurance 0.7% T.I./year

In order to determine an enterprise’s viability, one may consider an NPV/total investment ratio greater or equal to 2. The discount rate considered was 10% per annum [28]. In order to meet the second criterium, the internal rate of return (IRR) must be equal or greater than the discount rate considered, or 10% per annum, for demonstrating that the project should present a rate of return greater or equal to capital cost. The IRR corresponds to the discount rate value which nullifies Eq. (2). The sum of taxes was considered to be 35% of gross profit. This value cor-responds to the average tax rate for investments in the US [29]. Plant annual depreciation was con-sidered as the difference between total investment and residual investment divided by economic analysis time (time for economic criteria to be met). This term refers to the technological lag of the process over the years, making it necessary to consider it in the cash flow in order to decrease tax costs.

Crude glycerol characterization

The specification for crude glycerol used as feedstock for acrylic acid production in this study can be found in Table 6.

Table 6. Crude glycerol specifications [30]

Characteristic Specification, mass%

Glycerol Min. 80

Ashes Max. 7

Organic Residue Max. 2.5

Methanol Max. 0.5

Water Max. 13

Ashes defined in Table 6 refer to sodium sulfate from raw glycerol neutralization. The organic residue is the residual oil not reacted in the process, for a biodiesel plant using soy oil as feedstock, the com-position of such residue is indicated in Table 7, con-sidering only the main constituents of soy oil.


63

Table 7. Composition (mass%) of soy oil considered [31]

Component Value

Linoleic acid 55.00

Oleic acid 18.00

Palmitic acid 10.00

Linolenic acid 13.00

Stearic acid 4.00


Material balance

Table 8 displays the performed material balance from data obtained by Paula et al. [17], for determin-ing mole fractions of each component at the reactor outlet.

The mole fractions in Table 8 were inputted to Aspen HYSYS v8.0 simulator as a basis for simul-ating the process.

Table 8. Reactor material balance

Input data

Glycerol flow rate used as a basis at the reactor inlet (kmol/h) 100

Seletivity for the acrylic acid reaction (mol%) 0.854

Selectivity to the acrolein reaction (mol%) 0.038

Selectivity to the acetaldehyde reaction (mol%) 0.032

Selectivity to the propanal reaction (mol%) 0.023

Selectivity to the acetic acid reaction (mol%) 0.025

Selectivity to the propanoic acid reaction (mol%) 0.014

Carbon balance 0.9

Conversion of glycerol (mol%) 0.936

Flow rates calculated from input data

Oxygen flow rate required for the reaction (kmol/h) 67.462

Reacted glycerol flow rate (kmol/h) 93.6

Unreacted glycerol flow rate (at the reactor outlet) (kmol/h) 6.4

Generated acrylic acid flow rate (at the reactor outlet) (kmol/h) 79.934

Generated acrolein flow rate (at the reactor outlet) (kmol/h) 3.556

Generated acetaldehyde flow rate (at the reactor outlet) (kmol/h) 4.492

Generated propanal flow rate (at the reactor outlet) (kmol/h) 2.152

Generated acetic acid flow rate (at the reactor outlet) (kmol/h) 3.51

Generated propanoic acid flow rate (at the reactor outlet) (kmol/h) 1.310

Generated water flow rate (kmol/h) 218.041

Flow of water present in glycerol at the reactor inlet (kmol/h) 66.9

Flow of water added in the form of steam (kmol/h) 537

Total water flow in the reactor outlet (kmol/h) 821.941

Generated carbon dioxide flow rate (at the reactor outlet) (kmol/h) 14.04

Generated carbon monoxide flow rate (at the reactor outlet) (kmol/h) 14.04

Total mole flow at the reactor outlet (kmol/h) 951.378

Mole Fractions calculated from Flow Rates at the Reactor Outlet (for the reactor outlet)

Mole fraction of glycerol 0.0067

Mole fraction of water 0.8639

Mole fraction of acrylic acid 0.0840

Molecular fraction of acrolein 0.0037

Mole fraction of acetaldehyde 0.0047

Mole fraction of propanal 0.0022

Mole fraction of acetic acid 0.0036

Mole fraction of propanoic acid 0.0013

Mole fraction of carbon dioxide 0.0147

Mole fraction of carbon monoxide 0.0147


64

Process simulation description

Crude glycerol, coming from the biodiesel plant (stream 1.0), is first pre-heated in the TC-01 heat exchanger through the PFR-100 reactor effluent stream (stream 22.0), followed by a vacuum distil-lation column of 0.9 kPa abs (CD-01), in which resi-dual heavy oils, ashes and alkali are removed, result-ing in high-purity glycerol of 88.2% in mass at the top of the column (stream 7.0). The resulting glycerol is then mixed to a vapor stream (stream 9.0). The addit-ion of steam is fundamental to avoid the risk of an explosion, because a mixture of glycerol and pure oxygen is dangerous. The glycerol/steam mix then follows to the heat exchanger TC-03. In this heat exchanger, the mixture is pre-heated by the PFR-100 reactor effluent stream (stream 21.0), before entering the FH-100 furnace, and reaching reaction tempera-ture. The furnace is heated by burning natural gas. This resulting stream from the FH-100 furnace is mixed to a pure oxygen stream (stream 17.0) at the PFR- -100 reactor inlet, as to avoid any explosion related risk. The oxidative dehydration reaction occurs in the PFR-100 fixed bed reactor at room pressure (101.3 kPa abs) and 320 °C, according to Paula et al. [17]. The reactor effluent (stream 19.0) is cooled in the TC- -04 heat exchanger, follows to compressor C-01 and immediately to the heat exchanger TC-03, where it heats up the reaction mixture. The heat still contained in the reactor effluent is recovered in heat exchanger TC-01 for pre-heating crude glycerol, which follows to distillation column CD-01. After passing through the two above-mentioned heat exchangers, the effluent containing acrylic acid (stream 23.0) is cooled with cooling water (CW) in heat exchanger TC-05.

The whole process step is described in Figure 1. The cooled stream from the TC-5 heat exchan-

ger (stream 24.0) proceeds to flash vessel V-01, where light reaction gases are removed (carbon mon-oxide and dioxide) in stream 25.0. The liquid product stream from the flash (stream 26.0) is sent to the extractor column CA-01, in which excess water and residual glycerol are removed from the acrylic acid stream (stream 46.0). 2-Ethyl-1-hexanol is used as the extractor liquid (stream 29.0). The light phase from CA-01 (stream 30.0) follows to distillation column CD-02, where extractor liquid recovery occurs, which is then reused in column CA-01 (stream 33.0). The top product from CD-02 (stream 37.0) is directed to distil-lation column CD-03, where acrylic acid 98,33 mass% is obtained as the column bottom product (stream 43.0), while impurities such as water and acetic acid are removed at the top (stream 40.0). This second step is described in Figure 2.

Process main streams compositions and pro-perties are described in Table 9.

Proposed process characteristics

For distillation columns using glycerol, the use of a vacuum is fundamental, for glycerol starts to poly-merize and decay at temperatures above 200 °C [32].

The vacuum specified in the distillation column, CD-01, where ashes and residual oils are removed from glycerol, is 0.9 kPa, which is in accordance with values indicated in literature, from 0.66 to 1.33 kPa [32].

The liquid extractor used was 2-ethyl-1-hexanol, which presented the highest recovery percentage of acrylic acid in the organic phase (circa 99.99%) and a

Figure 1. Process simulation (first part).


65

larger biphasic region. As a disadvantage, the organic phase contains a significant amount of water, accord-ing to Alvarez et al. [33]. It was verified that, despite this downside, the use of this solvent results in lower energy expenditure, in steam terms, in the following distillations, besides requiring smaller columns and lower reflux ratios. According to Schultz et al. [34], acrylic acid and propanoic acid form an azeotrope,

which removal demands special separation methods, such as chromatography, in order to achieve a purity greater than 99 mass%.

Economic evaluation

Table 10 refers to feedstock, acrylic acid, and main production input prices.

Figure 2. Process simulation (second part).

Table 9. Process main stream compositions and properties

Stream 1.0 8.0 9.0 15.0 17.0 18.0 19.0 26.0 30.0 34.0 28.0 37.0 45.0

Temperature (°C) 31 28.2 250 320 320 320 320 55 55 55 31 64.84 36

Pressure (kPa) 101.3 201.3 201.3 151.3 151.3 151.2 151.3 101.3 101.3 151.3 151.3 101.3 101.3

Mass flow (kg/h) 39,820 34,490 31,960 66,450 7,139 73,590 73,590 71,590 62,430 36,780 7,163 25,600 18,640

Steam fraction 0 0 1 1 1 1 0 0 0 0 0 0

Composition (mass fraction)

Glycerol 0.766 0.882 - 0.4578 - 0.4134 0.0257 0.0272 0.0027 0.0046 - - -

Water 0.1 0.118 1 0.5422 - 0.4896 0.6436 0.6776 0.0757 - - 0.1846 -

Acrylic acid - - - - - - 0.2512 0.2651 0.3083 - - 0.741 0.9833

Acetic acid - - - - - - 0.0091 0.0096 0.0072 - - 0.0175 -

Propanoic acid - - - - - - 0.0042 0.0044 0.005 - - 0.0123 0.0167

Acrolein - - - - - - 0.0085 0.0065 0.0074 - - 0.0181 -

Acetaldehyde - - - - - - 0.0086 0.0062 0.007 - - 0.0171 -

Propanal - - - - - - 0.0054 0.0034 0.0039 - - 0.0094 -

Carbon dioxide - - - - - - 0.0267 - - - - - -

Carbon monoxide - - - - - - 0.017 - - - - - -

Linoleic acid 0.016 - - - - - - - - - - - -

Oleic acid 0.0069 - - - - - - - - - - - -

Palmitic acid 0.0035 - - - - - - - - - - - -

Linolenic acid 0.0023 - - - - - - - - - - - -

Stearic acid 0.0013 - - - - - - - - - - - -

Oxygen - - - 1 0.097 - - - - - - -

Sodium sulfate 0.1 - - - - - - - - - - - -

Sodium hydroxide 0.004 - - - - - - - - - - - -

2-Ethyl-1-hexanol - - - - - - - 0.5828 0.9954 1 - -


66

Table 10. Prices for feedstock, acrylic acid and main pro-duction inputs

Feedstock/product/input Price (US$/unit) Ref.

Acrylic acid 1,750/t [35]

Oxygen 280.8/t [36]

Crude glycerol 205/t [13]

Cooling water 0.0148/t [37]

Electricity 0.06/kWh [37]

Natural gas 142.6/t [38]

2-Ethyl-1-hexanol 1,100/t [39]

Steam was considered to be produced on site, and the price of cooling water corresponds to pro-duction costs. Moreover, the used oxygen was consi-dered to be purchased from an industrial utilities pro-ducing company, since such companies are very common.

As for economic indexes (NPV/total Investment, and IRR), which were calculated using the economic criteria previously presented, one can verify the obtained values in Table 11.

Table 11. Results from the economic analysis, including taxes

Operation time span 5 years

Total investment (thousand US$) 1.42×108

Variable cost (thousand US$/year) 9.8×107

Fixed cost (thousand US$/year) 1.1×107

Revenue (thousand US$/year) 2.8×108

NPV (thousand US$) 3.4×108

NPV/total investment 2.37

IRR (% per annum) 57

With the results of Table 11, it is possible to verify that the project meets economic criteria in a period of operation of 5 years, which indicates pos-sible viability for the process.

Considering the country’s strategic interest in producing acrylic acid from biomass, a project ana-lysis with a 100% tax exemption was performed. The results are presented accordingly in Table 12.

Table 12. Results of economic analysis with tax exemption


NPV (thousand US$) 3.95.108



Through the results of Table 12, it is possible to verify that, considering tax exemption, the time to meet economic criteria drops from 5 to 4 years, favor-ing even more the viability of the productive process.

Comparison to the glacial acrylic acid production process using propene

The acrylic acid production industrial process uses propene as feedstock, and was developed many years ago. The process flow chart, as well as the eco-nomic analysis is made available by Turton et al. [40].

Luyben [41] has recently presented an eco-nomic analysis considering trade-offs based, parti-cularly, on variations in reactor temperature and pres-sure. The author informs that no studies in the open literature have been found that discuss the optimum economic design of the acrylic acid process.

The process has been constantly modified, with new options offered in patents by the main producers. US patent 7,294,741 B2 by Stockhausen GmbH [42] describes that the acrylic acid production process from propene. In this process, the reaction is per-formed in two steps; in the first one, dehydration of propene occurs, and, in the second, formed acrolein in the first stage is oxidized into acrylic acid. The puri-fication of acrylic acid if performed using a non-vol-atile aromatic solvent, or by the absorption of acrylic acid in water, therefore separating acrylic acid from vent gases from the oxidation step. The purification step following the absorption process is not described in the patent. The patent reitetates the importance of propene dilution in steam to avoid the risk of exp-losion due to propene’s high flammability in the pre-sence of oxygen.

EP patent 1,116,709 A1 by Nippon Shokubai [43] also described acrylic acid production from pro-pene. The difference lies in that in this process the reaction occurs in a single stage. The purification pro-cess consists primarily in the separation of acrylic acid from other reactor vent gases through an abs-orption process. Acrylic acid then follows to a distil-lation column to be separated from the solvent and other remaining contaminants.The acid then follows to a crystallization process in order to increase formed acrylic acid purity, being converted from raw acrylic acid into glacial acrylic acid.

In order to verify the competitiveness of the gla-cial acrylic acid production from crude glycerin pro-cess, a comparison was performed between the latter and the production process using propene, since this is currently the consolidated route in industry for the production of glacial acrylic acid. For the economic analysis, the studied process has the acrylic acid puri-fication step performed according to EP patent 1,116,709 A1 by Nippon Shokubai [43], with the reaction occurring in two steps, in a similar manner to US patent 7,294,741 B2 by Stockhausen GmbH [42]. The prices for propene and acrylic acid were obtained


67

from ICIS [44] (US$ 900/t and 1750/t, respectively). With the information from the propene based process, economic analysis results were obtained according to Table 13.

Table 13. Results from the economic analysis for the propene-based acylic acid production process


Total investment (thousand US$) 1.62×108

Variable cost (thousand US$/year) 1.02×108

Fixed cost (thousand US$/year) 1.51×107

Revenue (thousand US$/year) 2.8×108

NPV (thousand US$) 3.38×108



From results in Table 13, notice that the acrylic acid production process from crude glycerin is com-petitive, when compared to the propene based pro-cess, having the crude glycerin based process pre-sented a lower total investment, shorter operation time for achieving economic criteria, and a higher IRR, according to Table 13.

The shorter operation time to meet economic criteria of NPV and IRR using glycerol as feedstock (5 years) versus propene (6 years) suggests an inter-esting competitiveness for the glycerol-based acrylic acid production process. It should be highlighted that glycerol comes from the biodiesel production process, which uses renewable feedstock, and is therefore more environmentally sustainable.

Sensitivity analysis

Among the considered variable costs, the one which causes the greatest impact on the productive process and its feasibility is the price of crude gly-cerol, because its cost corresponds to circa 70.9% of total variable costs. For a deeper analysis, the feasib-ility of the process under crude glycerol price vari-ations was verified. The results are presented in the graph in Figure 3.

As noticed in Figure 3, even if crude glycerol price increases by 2.6 times (US$ 540/t×US$ 205/t), the operation time necessary to meet economic crit-eria is exactly 20 years, which is a period of time commonly used as a basis to verify process viability [45]. Besides, it can be observed that small variations in the price of crude glycerol keep the time to meet economic criteria constant or with a small variation, which demonstrates that the process provides a good amortization. This analysis demonstrated that even if a scenario where crude glycerol prices increase

nearly threefold is considered, the process still pre-sents a satisfactory economic potential.

Figure 3. Time to reach economic criteria as a function of the

price of crude glycerol.

An analysis considering possible price variations in the end product (glacial acrylic acid) was also per-formed. The results of such analysis are shown in Figure 4.

Figure 4. Time to reach economic criteria as a function of the

price of acrylic acid.

As noticed in Figure 4, for a market price of gla-cial acrylic acid equivalent to 59% of the price used as a basis in this study (US$ 1.033/t×US$ 1.750/t), the time to meet economic criteria was kept lower than 20 years (18 in this case). Therefore, as in the previous analysis, for small variations in the price of acrylic acid, the time necessary to meet economic criteria presented a small variation or remained cons-tant, which demonstrates that the process also pro-vides for a good amortization when varying the price of glacial acrylic acid. The analysis showed that even with this possible drop in the price of the end product (59% of the base price considered in this study), the process would remain viable.


68

CONCLUSIONS

An acrylic acid production unit using purified gly-cerol is an interesting economic alternative for the commissioning of new industrial plants. The high price gap between the evaluated process feedstock (crude glycerol: US$ 205/t), and feedstock from the conventional process (propene: US$ 900/t) justifies a new process, considering the continuous growth in biodiesel production, and, as a consequence, of gly-cerol, keeping feedstock prices attractive in the inter-national market. An important positive factor is the use of renewable feedstock, considering the main end product use of acrylic acid, which is SAP used in dis-posable diapers. The simulation has shown that it is possible to produce glacial acidic acid in an econ-omically feasible way, since the economic indexes obtained lead to enterprise feasibility: NPV/total inv-estment greater than 2 and an IRR greater than capital cost. It was verified that, in 5 years, consider-ing taxes, the industrial plant would obtain a financial return 2.37 times greater than the invested capital. The analysis demonstrated that the process using crude glycerin is highly competitive, when compared to the consolidated industrial technology, as it needs a shorter operation time to reach economic criteria. The main technological bottlenecks for commercial production dwell in reactional catalytic process opti-mization.

Nomenclature

Abs - Absolute CW – Cooling Water D.I - Direct Investment D.I - Indirect Investment IRR - Internal Rate of Return L.M. - Laboratory Manpower M.P.E.C. - Main Pieces of Equipment Cost NPV – Net Present Value O.M. - Operation Manpower T.I. - Total Investment WHSV - Weight Hourly Space Velocity, h-1

REFERENCES

[1] C.L. Gargalo, P. Cheali, J.A. Posada, K.V. Gernaey, Ind. Eng. Chem. Res. 55 (2016) 6801-6814

[2] C.J. Schaschke, Chem. Ind. Chem. Eng. Q. 22 (2016) 453-460

[3] Statista, The statistics portal, https://www.statista.com/ /statistics/509875/production-volume-of-biodiesel-in-the-us/ (accessed 9 October 2018)

[4] Green Fleet, U.S. biodiesel production increased in 2017, https://www.greenfleetmagazine.com/143787/u-s-

biodiesel-production-continues-increasing-after-tax-credit (accessed 8 October 2018)

[5] Global Agricultural Information Network, EU biofuels annual 2017, https://gain.fas.usda.gov/Rec-ent%20GAIN%20Publications/Biofuels%20Annual_The%20Hague_EU-28_6-19-2017.pdf (accessed 6 November 2018)

[6] N.D. Mladenović, F. Kiss, B. Škrbić, B.Tomić, R. Mićić, Z. Predojević, Renew. Sustain. Energy Rev. 81 (2018) 280- -291

[7] C.R. Coronado, J.A. Carvalho, C.A. Quispe, C.R. Sotomonte, Appl. Therm. Eng. 63 (2014) 97-104

[8] BASF, BASF inaugurates world-scale acrylic acid com-plex in Camaçari, Brazil, https://www.basf.com/en/com-pany/news-and-media/news-releases/2015/06/p-15-260.html (accessed 12 October 2017)

[9] ICIS, European chemical profile: Acrylic acid, https:// //www.icis.com/explore/resources/news/2010/06/14/9367243/european-chemical-profile-acrylic-acid/ (accessed 19 November 2018)

[10] ICIS, Chemical profile: European acrylic acid, https:// //www.icis.com/explore/resources/news/2016/02/04/9967187/chemical-profile-europe-acrylic-acid/ (accessed 19 November 2018)

[11] Niosh, Niosh pocket guide to chemical hazards, http:// //www.asok.org/npg/nioshdbs/npg/npgd0013.html (accessed 21 February 2018)

[12] B. Katryniok, S. Paul, M. Capron, F. Dumeignil, ChemSusChem 2 (2009) 719-730

[13] ICIS, US crude glycerole prices could dip as spring nears, https://www.icis.com/resources/news/2018/02/14/10193613/us-crude-glycerole-prices-could-dip-as-spring-nears/ (accessed 13 June 2018)

[14] F. Wang, F.J. Xu, J.L. Dubois, W. Ueda, ChemSusChem 3 (2010) 1383-1389

[15] Y.T. Kim, K. Jung, E.D. A. Park, Appl. Catal., A 393 (2011) 275-287

[16] C.F. M. Pestana, A.C. Guerra, G.B. Ferreira, C.C. Turci, C.J. Mota, J. Braz. Chem. Soc. 24 (2013) 100-105

[17] A.S. Paula, L.G. Possato, D.R. Ratero, J. Contro, K.K. Adamsky, R.R. Soares, G. Goobes, L. Martins, J.G. Nery, Micropor. Mesopor. Mater. 232 (2016) 151-160

[18] L. Liu, B. Wang, Y. Du, Z. Zhong, A. Borgna, Appl. Catal., B 174 (2015) 1-12

[19] S.T hanasilp, J.W. Schwank, V. Meeyoo, S. Pengpanich, M. Hunsom, Chem. Eng. J. 275 (2015) 113-124

[20] R.M. Yeates, M.J. Murdoch, P.D. Southon, A.C. McLaughlin, R.F. Howe, F. Bonino, S. Bordiga, A. Damim, Dalton Trans. 38 (2009) 8025-8032

[21] Y.S. Yun, K.R. Lee, H. Park, T.Y. Kim, D. Yun, J.W. Han, J. Yi, ACS Catal. 5 (2014) 82-94

[22] A. Chieregato, M.D. Soriano, F. Basile, G. liosia, S. Zamora, P. Concepción, F. Cavani, J.M.L. Nieto, Appl. Catal., B 150 (2014) 37-46

[23] K. Omata, K. Matsumoto, T. Murayama, W. Ueda, Catal. Today 259 (2016) 205-212

[24] E.C. Carlson, Chem. Eng. Prog. 92 (1996) 35-46


69

[25] W.D. Seider, J.D. Seader, D.R. Lewin, S. Widagdo, Pro¬duct and Process Design Principles: Synthesis, Ana-lysis and Design, John Wiley & Sons, Inc, Hoboken, NJ, 2009, pp. 558-594

[26] Global Rates, Inflation United States – Consumer price index (CPI), https://www.global-rates.com/economic-indicators/inflation/consumer-prices/cpi/united-states.aspx (accessed Feb 10, 2018)

[27] M.S. Peters, K.D. Timmerhaus, R.E. West, Plant Design and Economics for Chemical Engineers, McGraw-Hill, New York, 2003, pp. 183-198

[28] H.J. Cho, J.K. Kim, H.J. Cho, Y.K. Yeo, Ind. Eng. Chem. Res. 52 (2013) 462-468

[29] Heritage, 2018 index of economic freedom, https:// //www.heritage.org/index/country/unitedstates (accessed 17 June 2018)

[30] Bunge, Glycerol, https://www.bungeargentina.com/en/ /productos-y-servicios/alimentos-e-ingredientes/glicerol (accessed 14 October 2017)

[31] T.E. Clemente, E.B. Cahoon, Plant Physiol. 151 (2009) 1030-1040

[32] F. Shahidi, Bailey's Industrial Oil and Fat Products, John Wiley & Sons, Inc, Hoboken, NJ, 2005, pp. 30-36

[33] M.E. Alvarez, E.B. Moraes, A.B. Machado, R. Maciel Filho, M.R. Wolf-Maciel, Appl. Biochem. Biotechnol. 137 (2007) 451-461

[34] M.J. Schultz, T.P. Binder, N. Sudharsan, A. Hilaly, (Archer Daniels Midland Company), US patent 9493393B2 (2013)

[35] Tecnon OrbiChem, Chem-netfacts chemical market insight and foresight - on a single page - acrylic acid & acrylates, https://mdchemicals.net/wp-content/uploads/

/2018/05/May-2018-Acrylates-Report.pdf (accessed 21 June 2018)

[36] Ycharts, US producer price index: industrial gas manu-facturing: Oxygen:280.80 for 2017, https://ycharts.com/ /indicators/us_producer_price_index_industrial_gas_manufacturing_oxygen_yearly (accessed 21 June 2018)

[37] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, Analysis, Synthesis and Desing of Chemical Processes, Prentice Hall, Boston, MA, 2012, p. 231

[38] Forbes, The unexpected jump in U.S. Natural gas prices, https://www.forbes.com/sites/judeclemente/2018/09/20/the-unexpected-jump-in-u-s-natural-gas-prices/#774ee4a6556d (accessed 2 October 2018)

[39] Alibaba, 2 -ethyl hexanol/2-EH CAS 104-76-7, https:// //www.alibaba.com/product-detail/2-Ethyl-Hexanol-2-EH-CAS_60259944021.html?spm=a2700.7724857.normalList.45.7d1a7134Vatw3P (accessed 16 April 2018)

[40] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaelwitz, Ana-lysis, Synthesis and Design of Chemical Processes, Prentice Hall, 2nd ed., Upper Saddle River, NJ, 2003, Appendix B.9

[41] W.L. Luyben, Comput. Chem. Eng. 93 (2016) 118-127

[42] G.B. Marl, M.C. Rauxel, M. Recklinghausen, A.S. Moers, R.F. Dulmen, M.P. Hunxe, (Stockhausen GmbH) US patent 7,294,741 B2 (2007)

[43] U. Kouji, U. Masatoshi, N. Sei, (Nippon Shokubai Co., Ltd.), EP patent 1,116,709 A1 (2001)

[44] ICIS, US December propylene contracts settle down 8 cents/lb, https://www.icis.com/explore/resources/news/ /2018/12/19/10297466/us-december-propylene-con-tracts-settle-down-8-cents-lb/ (accessed 01 May 2019)

[45] Z. Lin, V. Nikolakis, M. Ierapetritou, Ind. Eng. Chem. Res. 53 (2014) 10688-10699.

EDUARDO RAMOS BRAGA GEORGE DE SOUZA MUSTAFA

DANILO DE AGUIAR PONTES LUIZ ANTÔNIO MAGALHÃES

PONTES

Federal University of Bahia, Postgraduate Program of Chemical

Engineering, Federação, Salvador - BA, Brazil

NAUČNI RAD

EKONOMSKA ANALIZA I TEHNIČKE KARAKTERISTIKE PROIZVODNJE AKRILNE KISELINE IZ SIROVOG GLICEROLA

Glicerol je glavni nusproizvod industrijskih postrojenja za dobijanje biodizela, a prouča-vani su i novi tehnološki postupci koji ga koriste kao sirovinu, zbog porasta svetske proizvodnje biodizela. Jedna od mogućih primena glicerola je u proizvodnji akrilne kise-line, proizvoda sa nekoliko industrijskih primena. Ovaj rad analizira novi proces kon-verzije sirovog glicerola koji uključuje prečišćavanje radi uklanjanja nečistoće, jednoste-penu reakciju oksidativne dehidracije i dalje prečišćavanje proizvoda sve dok se ne ispune standardi kvalieta za glacijalnu akrilnu kiselinu. Utvrđeno je da je proces eko-nomski održiv, sa vremenom otplate od 5 godina sa odnosom NPV (neto sadašnja vred-nost)/investicije ≥ 2 i IRR (interna stopa prinosa) ≥ 10% godišnje, odnosno 4 godine u slučaju poreskih podsticaja za razvoj zelenih tehnologija.

Ključne reči: akrilna kiselina, glicerol, prečišćavanje, biodizel, vreme.




71

ERHAN SULEJMANI

MUHAMET DEMIRI

Department of Food Technology, University of Tetova, Tetovo,

North Macedonia

SCIENTIFIC PAPER

UDC 637.524:66.014

VOLATILE COMPOUNDS OF MACEDONIAN FERMENTED SAUSAGE AS AFFECTED BY RIPENING PROCESS USING SPME/GC-MS

Article Highlights • Smoking induces numerous changes in meat that is relevant to sausages • Fermentation process enhanced the volatiles in sausages • Ketones and acids were the main volatiles in sausages during ripening Abstract

The profiles of volatile compounds of Macedonian dry fermented sausage were determined by gas chromatography-mass spectrometry (GC–MS) using a solid--phase microextraction (SPME). A total of 103 volatile compounds were iden-tified and consisted of 12 acids, 16 ketones, 21 terpenes, 20 alcohols, 9 esters, 13 hydrocarbons and 12 miscellaneous. The analysis of the main volatile aro-matic components proved that 62% of the compounds that were detected on the first days are terpenes and their content increased to 69 and 80% on days eight and fifteen, respectively. The 50% of the total concentration of volatile components results to be concentrated on the last day (15th) of ripening. They have the highest amounts of concentration of acids (57%), ketones (68%) and terpenes (56%) from the total concentration. There was a significant difference between the various periods of fermentation and there can be differences in the manufacture production stages. Macedonian dry sausages contain high levels of spicy terpenes, which could play an important role in the general fla-vour notes of this meat product.

Keywords: acids, terpenes, ketones, hydrocarbons, sausage.

A variety of fermented meat products are pro-duced nowadays and these products vary in ripening (slow, medium, fast), depending on the sugars used during drying. The final pH, flavour and texture are also different between fermented meat products [1]. Smoking and fermentation are some of the oldest conservations methods and sausages are one of the oldest meat products [2,3]. Fermented dry sausages are high-quality products in the meat industries, valued and demanded by consumers. The traditional sausage originated in the Mediterranean area is mainly preserved by drying only, while sausages ori-ginating from central and northern Europe are dried

Correspondence: E. Sulejmani, Department of Food Technology, University of Tetova, 1200 Tetovo, North Macedonia. E-mail: [email protected] Paper received: 3 February, 2019 Paper revised: 22 August, 2019 Paper accepted: 25 August, 2019

https://doi.org/10.2298/CICEQ190203026S

and smoked [4,5]. Meat product preservations are some of the oldest technologies since ancient times.

Macedonian sausage is dry fermented sausage and is produced in large quantities in all parts of the country in a traditional and commercial way. Tradit-ionally, it is produced during the winter period by fil-ling the bovine fillet meat and loins of elderly animals with the addition of kitchen salt, black pepper and garlic in thin beef wraps, and then drying in a classic smoke without control of the atmospheric conditions. However, commercial sausage is produced from beef and tallow with the addition of salts, spices, antioxid-ants, nitrites, and starter cultures. Such a sausage is filled with artificial, usually collagen, wrappers and subjected to controlled atmospheric conditions of dry-ing and ripening. The tradition of production and con-sumption of sausage in the territory of Macedonia and wider was known in the former Ottoman Empire under the name "soudjuk" or "sucuk", which in the past was produced only from beef meat, and today from sheep and buffalo meat also [6].

E. SULEJMANI, M. DEMIRI: VOLATILE COMPOUNDS OF MACEDONIAN... Chem. Ind. Chem. Eng. Q. 26 (1) 71−78 (2020)

72

Only few things are known about the manufact-uring technology and the gross compositional charac-teristics of dry fermented sausage produced in Mace-donia [7]. There is no published information on the volatile composition of Macedonian dry fermented sausage; therefore, this work aims at providing an impression of the volatiles of fermented sausage during ripening under industrial conditions and better understanding of these sausages due to the lack inf-ormation about them, which will lead to a protected designation of origin.

EXPERIMENTAL

Sausage dough was prepared from cattle meat trimmings (80%) mixed with fat from ribs (20%) as raw material. Then salt (2.0%), sugar with GDL (1%), potassium nitrate (E 252) (0.02 %) and spices (0.1- -0.5%) were added. The meat was minced to 2 cm then decreased to 3 mm at 0 °C at a speed of 1300 rpm and the spices were added and mixed. No starter culture was added. The meat mixture was stuffed into artificial casings (Kutezin, Czech Republik, 40 mm diameter) and the final weight of each sausage was around 700 g. The filling speed was 10 pieces per min. After conditioning (12 h at 18-20 °C and air hum-idity 58-60%), the products were subjected to ferm-entation, drying, smoking and ripening under the fol-lowing regime: drying and smoking at 18-19 °C, 86- -92% RH, air circulation 0.2-0.5 m/s, for the duration of 5 days; ripening at 17 °C, 78% RH, air circulation 0.5-0.8 m/s, for the duration of 10 days. The smoke was poured during 5 days for 3 h at a microclimate of 19 °C and 80-85% RH.

Sampling and sample preparation

During the production of the sausage, samples for physicochemical analysis were taken on the first day (0), second, fourth, sixth, eighth and fifteenth. Volatile components analyses were performed on the first, eight and fifteenth day of the ripening. All ana-lyses were carried out in duplicate.

Chemical analyses

Ten grams of sample were homogenized in 100 ml of distilled water and the pH of this mixture was determined by a pH meter (Metler Toledo). Moisture, salt, ash, protein, fat and residual nitrite contents measurements were determined according to the methods described by AOAC [8].

SPME–GC analysis

Analyses of the volatiles were performed by a solid-phase microextraction (SPME) method at three

stages of maturation (Day 0, 8 and 15) using a gas chromatography-mass spectrometry (GC-MS) system (Shimadzu GC-2010/QP-2010 mass spectrometry system, Shimadzu Corp.). A 3 g portion of each sample was placed in a 15 mL vial, and then 10 µL of internal standard containing 100 ppm 2-methyl-3-hep-tanone, 300 ppm 2-methyl-1-pentanoic acid and 278 ppm ethyl heptanoate in methanol (Sigma-Aldrich Co., St. Louis, MO, USA) was added to the vial and the mixture allowed to equilibrate at 40 °C for 30 min. The fiber was positioned at 3.0 scale units in each run. Desorption of the extracted volatiles was carried out and run in split mode (ratio was 1:20). During desorption, the fiber remained in the injector for 2 min at 250 °C, with helium as the carrier gas at a flow rate of 1.0 mL/min. The volatile compounds were separ-ated in a DB-Wax column (60 m×0.25 mm, 0.25 µm; J&W Scientific, Folsom, CA, USA). The identifications were based on comparing mass spectra of unknown compounds with those in Wiley 7 (7th ed., John Wiley & Sons Inc., 2005) and NIST/EPA/NIH 02 (http:// //www.nist.gov) mass spectral library. Identifications were also confirmed by comparing retention times with reference standards when it was available. Each compound was expressed as μg/100 g of sausage. The analyses were performed in triplicate

Statistical analyses

One way (ANOVA), post-hoc (Duncan test) was performed using the software package SPSS pro-gram for Windows, version 9.0 (SPSS Inc., Chicago, IL, USA). Differences were considered significant at P < 0.05.


Chemical compositions of sausages were in accordance with national regulations and showed similarities to those reported form other authors [9,10,11]. A total of 103 volatile compounds were identified and consisted of 12 acids, 16 ketones, 21 terpenes, 20 alcohols, 9 esters, 13 hydrocarbons and 12 miscellaneous compounds. The analysis of the main volatile aromatic components proved that that 62% of the compounds that were detected on the first days are terpenes and their content increased until 69 and 80% on days eight and fifteen, respectively. Trip-licate analyses for all samples were performed and the average values with standard deviation (SD) are listed in Tables 1–7. The volatiles of Macedonian fermented sausages have not been characterized and the results are discussed by comparison to data rep-orted by other researchers.


73

Carboxylic acids

Carboxylic acids were the second most abun-dant chemical family isolated from the headspace of the Macedonian fermented sausages from the differ-ent fermentation period (Table 1). A high concen-tration (713.8 μg (100 g)–1) of these acids was det-ected at the last day of fermentation. However, a lower concentration (about 268.2 μg (100 g)–1) was detected in the beginning and middle of the product-ion stage of the fermented sausage. Twelve different acids were identified in the Macedonian fermented sausage. The principal acid was acetic acid and its concentration was about 501.78±292.89 μg (100 g)–1 of sausage on day 15 of fermentation, while consider-ably higher levels of propiolic acid (73.68 μg (100 g)–1) and butanoic acid, 3-methyl (40.67 μg (100 g)–1) were detected on the first day of production. Throughout the fermentation period of the sausage, most of the carboxylic acids came from lipolysis of triglycerides followed by those produced from lactate metabolism; therefore, lipolysis was the main pathway responsible for the release of carboxylic acids. Acetic acid was probably the product of citrate or lactate fermentation of amino acid catabolism by bacteria. The differences between concentrations in the volatile acids are asso-ciated with the fermentation periods since the longer ripening period results in higher amount of carboxylic acids (Table 1).

Ketones

Most of the ketones in Macedonian fermented sausages were methyl ketones (Table 2). There were significant differences among the sausages (P < 0.05)

for different fermentation periods for 2,3-butanedione, 2,3-pentanedione, 2,3-octanedione and ethanone-1. The highest (358.5 μg (100 g)–1) and lowest (69.69 μg (100 g)–1) concentrations were identified at the end and on the eighth day of fermentation, respectively. Methyl ketones are produced from free fatty acids by an alternative pathway to β-oxidation. About 6 ket-ones were identified in sausages of the first day of fermentation and 12 identified ketones were found in sausages at the end of fermentation with 2-cyclo-hexen-1-one, 2-methyl being the most abundant. 2,5- -dimethyl-3(2H)-furanone and other furanones are very important aromas of heat-treated beef meat [12].

Terpenes

A total of 21 terpenes were identified in Mace-donian fermented sausages, being the most abundant chemical family (72%) isolated from the headspace of the sausages from the different fermentation period (Table 3). Terpenes presumably come from animal feed and therefore, they are important for determining the geographical origin of the sausage [13]. The maj-ority of terpenes were identified in sausages on the last day of fermentation and the most abundant was limonene, 1618.6 µg per 100 g of sausage. In sau-sages manufactured in the traditional way, terpenes are important volatile compounds with origins in plants that constitute the mixture of the product. A wide variety of terpenes and their derivatives have been isolated in Turkish sausages [9,10].

Alcohols

Twenty different alcohols consisting of primary, secondary and branched-chain alcohols were present

Table 1. The abundance of carboxylic acids in Macedonian dry fermented sausage during ripening. Results are given as mean values µg/(100 g) sausage with ±SD from at least triplicate analysis; RI: retention index; n.d: not detected; ns: not significant; P: probability;*P < 0.05

Acid RI Fermentation period, days

P 0 8 15

Propiolic acid <600 73.68±4.07 12.05±1.86 13.76±1.55 *

Propanoic acid, 2 methyl <600 n.d 6.91±1.97 49.27±1.31 *

Acetic acid 1326 n.d 200.15±35.18 501.78±92.89 *

Propanoic acid 1471 n.d 4.70±2.15 n.d ns

2,4-Hexadienedioic acid 1560 n.d n.d 2.29±1.97 ns

Butanoic acid 1610 74.22±4.64 11.23±1.45 73.67±4.65 ns

Butanoic acid,3-methyl 1673 40.67±2.78 11.58±20.05 n.d *

Butanoic acid, 2-methyl 1675 n.d 3.97±6.88 3.93±6.81 ns

Heptanoic acid 1916 n.d 10.20±17.66 n.d ns

Hexanoic acid 1918 79.60±5.46 n.d 61.60±5.10 ns

Octanoic acid 2194 n.d n.d 7.54±1.06 ns

2-Ethyl-butanoic acid 2352 n.d 6.27±1.85 n.d ns

Total 268.17 267.06 713.85


74

Table 2. The abundance of ketones in Macedonian dry fermented sausage during ripening. Results are given as mean values µg/(100 g) sausage with ±SD from at least triplicate analysis; RI: retention index; n.d: not detected; ns: not significant; P: probability;*P < 0.05

Ketone RI Fermentation period, days

P 0 8 15

2-Propanone 771 14.90±4.01 8.39±2.42 57.83±3.57 ns

2,3-Butanedione 1145 1.13±0.29 n.d n.d *

2-Butanone, 3-methyl 1149 n.d n.d 0.94±0.82 ns

2,3-Pentanedione 1419 1.90±0.58 n.d n.d *

Cyclopentanone 1945 n.d n.d 11.32±5.25 ns

2-Butanone, 3-hydroxy 2331 6.90±1.72 7.19±1.11 11.28±4.49 ns

2,3-Octanedione 2435 3.83±0.65 0.79±0.37 n.d *

3-Octanone, 2-methyl- 2440 n.d n.d 1.55±1.38 ns

2-Cyclopenten-1-one, 2-methyl- 2667 n.d 7.62±2.14 18.75±9.00 ns

trans.trans-3,5-Heptadien-2-one 3135 n.d n.d 24.10±8.59 ns

Ethanone, 1-(2-furanyl) 3200 n.d 7.47±1.37 n.d *

2-Cyclopenten-1-one, 2.3-dimethyl- 3299 n.d 1.32±1.28 0.63±0.09 ns

Butyrolactone 3590 n.d n.d 1.32±0.29 ns

2-Furanone, 2,5-dihydro-3,5-dimethyl 3643 n.d n.d 0.85±0.47 ns

2(5H)-Furanone, 3-methyl- 3868 n.d 0.61±0.06 17.76±2.20 ns

2-Cyclohexen-1-one, 2-methyl-5- 3934 69.89±5.81 36.31±10.11 212.14±10 ns

Total 98.54 69.69 358.46

Table 3. The abundance of terpenes in Macedonian dry fermented sausage during ripening. Results are given as mean values µg/100 g sausage with ±SD from at least triplicate analysis; RI: retention index; n.d: not detected; ns: not significant; P: probability;*P < 0.05

Terpene RI Fermentation period, days

P 0 8 15

Alpha-Pinene 1315 99.78±12.40 247.75±64.62 438.09±57.15 ns

Alpha-thujene 1327 62.96±13.31 147.51±76.90 302.57±39.13 ns

Camphene 1463 1.94±0.25 4.07±1.38 1.48±0.28 ns

l-Beta-pinene 1603 111.05±33.93 291.87±91.94 566.55±77.17 ns

Sabinene 1657 310.68±36.02 714.91±70.36 1411.78±185.51 ns

trans-Sabinene hydrate 2447 9.08±1.34 16.41±2.26 1.25±0.17 ns

cis-Sabinene hydrate 2694 n.d 8.26±1.77 13.44±1.64 ns

Delta 3-carene 1764 212.95±12.97 529.62±95.18 958.80±120.95 ns

Beta-myrcene 1803 92.45±2.14 237.68±81.04 463.59±66.81 ns

L-Phellandrene 1822 19.38±1.13 222.16±57.43 214.90±23.51 ns

Alpha-terpinene 1883 22.68±3.63 62.39±6.47 128.56±12.52 ns

Limonene 1973 359.28±5.28 910.45±50.64 1618.60±220.66 ns

Beta-phellandrene 1996 19.68±7.05 87.79±8.53 180.51±27.89 ns

1,4-Cyclohexadiene, 1-methyl-4- 2005 n.d 23.27±5.68 n.d ns

Gamma-terpinene 2150 50.13±5.39 95.44±5.31 235.48±37.97 ns

Para-cymene 2243 26.91±0.95 52.35±8.51 15.35±3.53 ns

Alpha-terpinolene 2260 29.08±1.84 64.34±8.19 140.14±11.22 ns

Delta-elemene 3019 0.67±1.15 5.34±8.11 0.30±0.53 ns

Trans-caryophyllene 3491 32.55±8.59 60.22±7.96 81.15±18.75 ns

Dihydrocarvone 3587 2.70±0.25 14.51±1.50 n.d *

Alpha-humulene 3731 1.27±0.35 3.64±5.07 n.d ns

Total 1465.22 3799.99 6772.53 0


75

in the sausages samples. The levels of five alcohols were significantly affected (P < 0.05) by the ferment-ation period of sausage manufacture. Generally, pri-mary alcohols originate from the corresponding alde-hydes produced from fatty acids and from amino acid metabolism. Among these, ethanol may be formed by lactose metabolism or by reduction of acetaldehyde. Secondary alcohols are obtained by the enzymatic reduction of methyl ketones. Alcohols were quantitat-ively the most abundant volatiles on the last day of sausage fermentation (Table 4) with 14 different alco-hols being identified. Ethanol was the most abundant alcohol in the sausage on day 15 of fermentation. The ethanol was responsible for the increase in alcohols caused by the microbial growth in beef meat during storage [14,15]. It is noted that ethanol is a contributor to a weak grilled and acetaldehyde-like aroma char-acteristics [16].

The concentration of alcohols in the sausages from another period may have been higher than the reported levels, as it was reported that the level of alcohols can fluctuate during fermentation. The sec-ond most abundant alcohol was cyclopentanemeth-anol, with the highest concentration in the sausages on day 15 of fermentation. Amounts of other alcohols showed decreased levels in the sausages, probably due to alcoholysis or esterification reactions.

Esters

Nine esters were found in the fermented sau-sages at different fermentation times. These com-pounds are produced by enzymatic or chemical react-ions of fatty acids with primary alcohols, so the alco-hol concentration is a limiting factor. The most fre-quently identified subgroups were seven methyl and two ethyl esters (Table 5). The amounts of two esters were significantly different (P < 0.05) in the sausages. Esters were the unique chemical family with a high presence on the first day of sausage fermentation.

Hydrocarbons

A total of 13 hydrocarbons were identified in Macedonian fermented sausages, being the least abundant chemical family (1%) isolated from the headspace of the sausages at different fermentation periods (Table 6). The majority of hydrocarbons were identified in sausages on the last day of fermentation and the most abundant was toluene with 39.5 µg per 100 g of sausage, which might be a consequence of the cyclization of unsaturated carbonylic chains pro-duced by lipid degradation [17].

Miscellaneous

Twelve miscellaneous compounds, including nine aldehydes, two sulfur, and one imine compounds

Table 4. The abundance of alcohols in Macedonian dry fermented sausage during ripening. Results are given as mean values µg/100 g sausage with ±SD from at least triplicate analysis; RI: retention index; n.d: not detected; ns: not significant; P: probability;*P < 0.05

Alcohol RI Fermentation period, days

P 0 8 15

1,2-Ethanediol 825 5.82±0.43 19.77±2.72 n.d ns

2-Propanol 892 2.54±0.60 2.73±0.61 0.70±0.21 ns

Ethanol 908 3.89±0.85 15.28±1.23 83.56±1.98 ns

1-Hexanol, 2-ethyl- 1104 0.60±0.05 n.d n.d ns

1-Butanol 1554 0.42±0.30 n.d n.d ns

1-Butanol, 3-methyl- 1765 3.36±0.81 11.50±1.17 6.24±1.12 ns

Eucalyptol 1777 10.89±3.91 17.60±1.76 2.09±0.62 ns

1-Pentanol 1899 3.42±0.34 3.36±0.60 6.89±1.93 ns

1-Hexanol 2237 1.85±0.17 n.d n.d *

3-Buten-2-ol 2540 4.29±3.73 n.d n.d ns

1-Heptanol 2562 1.18±0.09 n.d n.d *

Cyclopentanemethanol, alpha-methyl- 2825 n.d n.d 34.61±5.82 ns

L-Linalool 2833 5.71±0.59 10.12±2.83 0.89±0.54 ns

2,3-Butanediol 2929 n.d 5.35±7.02 18.69±2.82 ns

3-Cyclohexen-1-ol, 4-methyl-1- 3043 14.58±2.34 27.10±3.99 2.19±1.79 ns

2-Furanmethanol 3169 n.d 106.41±13.31 7.43±1.88 ns

Phenol, 2-methoxy- 3732 n.d 63.33±7.67 3.83±1.64 ns

2-Methoxy-4-methylphenol 3974 n.d 30.63±3.69 1.51±0.62 ns

Phenol 4073 n.d 17.00±2.42 1.12±0.93 ns

Phenol, 2-methoxy-4-(2-propenyl)- 4463 5.67±3.06 15.93±2.31 1.06±0.84 ns

Total 64.22 346.12 170.81


76

Table 5. The abundance of esters in Macedonian dry fermented sausage during ripening. Results are given as mean values µg/100 g sausage with ±SD from at least triplicate analysis; RI: retention index; n.d: not detected; ns: not significant; P: probability;*P < 0.05

Ester RI Fermentation period, days

P 0 8 15

Butanoic acid, methyl ester 967 2.14±1.03 1.11±0.97 1.83±0.44 ns

Butanoic acid, 2-methyl-ester 1033 0.65±0.61 0.24±0.21 0.79±0.10 ns

Pentanoic acid, methyl ester 1247 1.65±0.75 3.87±1.26 0.68±0.10 ns

Pentanoate, 2-methyl 1282 7.30±4.22 16.68±2.96 0.57±0.19 ns

Hexanoic acid, methyl ester 1570 14.77±4.87 22.84±2.37 4.66±1.41 ns

Hexanoic acid, ethyl ester 1708 0.16±0.28 n.d 1.03±0.22 *

Heptanoic acid, methyl ester 1891 18.00±4.62 85.16±12.87 n.d ns

Octanoic acid, methyl ester 2208 2.32±0.86 0.93±0.93 n.d *

1,2-Benzenedicarboxylic acid, diethyl ester 4621 18.93±1.56 1134.16±190.51 53.24±7.61 ns

Total 65.92 1265 62.8

Table 6. The abundance of hydrocarbons in Macedonian dry fermented sausage during ripening. Results are given as mean values µg/100 g sausage with ±SD from at least triplicate analysis; RI: retention index; n.d: not detected; ns: not significant; P: probability; *P < 0.05

Hydrocarbon RI Fermentation period, days

P 0 8 15

Undecane, 2,4-dimethyl- 760 1.00±0.74 n.d 5.87±2.56 ns

Octane, 4-methyl 849 1.31±0.82 n.d 0.29±0.10 *

1-Propene, 3-(methylthio) 1096 0.63±0.09 4.15±1.65 15.52±9.37 ns

Decane, 3,7-dimethyl- 1345 2.96±2.57 0.79±0.37 n.d ns

Benzene, methyl- (CAS) toluene 1371 7.69±1.41 30.03±7.19 39.52±2.67 ns

Ethylbenzene 1678 1.61±0.37 n.d n.d *

4-Undecene, (Z)- 1710 1.00±0.01 n.d n.d *

Cyclopropane, 1,1,2,3-tetramethyl- 1945 0.94±0.10 n.d n.d ns

Styrene 2199 7.23±2.21 5.82±4.2 7.75±2.5 ns

Benzene, 1-methyl-2- 2237 0.54±0.23 8.00±5.12 n.d ns

Dimethylstyrene, alpha-para 2331 n.d 2.56±1.11 n.d ns

1-Octene 2407 n.d 4.39±2.12 n.d ns

Copaene, alpha- 2554 3.28±1.11 5.36±2.22 0.29±0.11 ns

Total 28.18 61.1 69.24

were detected (Table 7). Diallyl disulfide was the most abundant miscellaneous compound that has been found to determine distinct aroma flavors of cooked meat, as well as being a key flavor compound of cooked Irish Angus beef [18]. Sulfur compounds such as diallyl disulfide and methyl disulfide have been derived from allicin, which is one of the major com-pounds in garlic [19].

CONCLUSIONS

The purpose of this study was to characterize the volatile profile of the sausages that are important for the food sector of North Macedonia. Volatile ter-penes were the most abundant compounds isolated in the headspace analyses of fermented sausage.

These terpenes are of the highest importance for the aromatic profile of this type of sausage. The pattern of volatile acids formation according to their most pro-bable origin could be associated with the different and typical characteristics in each sausage at different fer-mentation times. The concentration of volatile com-pounds varied greatly with high standard deviations, due to the lack of standard manufacturing protocols and age-related differences. In general, the highest concentration of acids, ketones and terpenes were observed on the last day, whereas the highest concentration of the esters and alcohols compounds were observed on the eighth day of fermentation of the sausage. In addition, the volatile profiles could be applied for quality control of the meat products. Com-plementary sensory and microbial analyses should be


77

performed in the future to further develop the relation-ship between manufacturing factors and the formation of volatiles in dry fermented sausages.

Acknowledgments

We thank Dr A.A. Hayaloglu from the Depart-ment of Food Engineering from the Inonu University (Malatya, Turkey) for the technical support during laboratory analyses. We cordially thank the Meat Pro-cessing Plant “Uka Komerc” in Skopje, North Mace-donia, for providing the sausages.

REFERENCES

[1] L. Kröckel, Fermented Meats. Blackie Academic & Professional Publishing, 1995, pp. 69-109

[2] E. Petäjä-Kanninen, E. Puolanne, Principles of Meat Fer-mentation, In: Handbook of Fermented Meat and Poultry, edited by F. Toldrá, Y.H. Hui, I. Astiasarán, Ë.K. Nip, Eds., J.G., 2007, pp. 17-30

[3] I. Vuković, Osnove tehnologije mesa, Veterinarska komora Srbije, Beograd, 2012, pp.151-210 (In Serbian)

[4] F. Toldrá, Improving the sensory quality of cured and fermented meat products: Processed Meats, J.P Kerry J.F. Kerry, Eds., Woodhead Publishing Series in Food Science, Technology and Nutrition, 2011, pp. 508-526

[5] M.L. Latorre-Moratalla, T. Veciana-Nogués, S. Bover-Cid, M. Garriga, T. Aymerich, E. Zanardi, A. Ianieri, M.J. Fra-

queza, L. Patarata, E.H. Drosinos, A. Lauková, R. Talon, M.C. Vidal-Carou, Food Chem. 107 (2008) 912–921

[6] J. Gasparik-Reichardt, S. Toth, L.G. Cocolin, E. Comi, H, Drosinos, Ž. Cvrtila, L. Kozačinski, A. Smajlović, S. Saičić, B. Borović, Meat Tech. 46 (2005) 143-153

[7] M. Stojanova, O. Najdenovska, Z. Pejkovski, M. Trajcev, Iosr J Eng. (2017) 44-49

[8] AOAC (1990), Official Methods for The Analysis (15th Ed.). Washington, DC, Association of Official Analytical Chemists

[9] K.T. Ozkara, A. Amanpour, G. Guclu, H. Kelebek, S. Selli, Food Anal. Methods 12 (2019) 729-741

[10] G. Kaban, Int. J. Food Prop. 13 (2010) 525-534

[11] G.C. Gürakan, T.F. Bozoglu, N. Weiss, LWT - Food Sci. Tech. 28 (1995) 139-144

[12] R. Kerscher, W. Grosch, Z. Lebensm. Unters. Forsch. A. 204 (1997) 3–6

[13] V. Vasta, A. Priolo, Meat Sci. 73 (2006) 218–228

[14] N. Kocsis, M. Amtmann, Z. Mednyánszky, K. Korány, J. Food Comp. Ana. 15 (2002) 195-203

[15] H.A. Ismail, E.J. Lee, K.Y. Ko, D.U. Ahn, Meat Sci. 80 (2008) 582-591

[16] D.S. Mottram, Food Chem. (1998) 415-424

[17] A. Meynier, E. Novelli, R. Chizzolini, E. Zanardi, G. Gan-demer, Meat Sci. 51 (1999) 175-183

[18] D. Machiels, S.M. Van Ruth, M.A. Posthumus, L. Istasse, Talanta (2003) 755-764

[19] M.S. Rahman, Int. J. Food. Prop. 10 (2007) 245-268.

Table 7. The abundance of miscellaneous in Macedonian dry fermented sausage during ripening. Results are given as mean values µg/100 g sausage with ±SD from at least triplicate analysis; RI: retention index; n.d: not detected; ns: not significant; P: probability;*P < 0.05

Miscellaneous Compounds RI Fermentation period, days

P 0 8 15

Acetaldehyde <600 n.d 2.09±1.63 n.d ns

Butanal, 3-methyl- 807 5.98±0.66 0.33±0.17 20.71±2.10 ns

Pentanal 952 4.10±1.26 7.72±1.45 21.05±2.04 ns

Hexanal 1238 7.05±2.35 n.d 37.96±6.56 ns

Octanal 1898 1.21±1.05 n.d n.d ns

Nonanal 2227 1.86±0.22 n.d n.d *

2-Heptenal, (E)- 2320 0.33±0.11 n.d n.d ns

2-Furancarboxaldehyde l 2412 n.d 2.34±1.06 n.d ns

Diallyl disulphide 2498 111.15±6.24 271.99±19.76 n.d ns

Benzaldehyde 2632 n.d 2.92±5.06 245.03±33.08 ns

Allylmethyl trisulfide 2830 1.21±0.14 n.d n.d *

Methoxy, phenyl- ,oxime 3158 4.24±0.22 6.54±0.59 n.d ns

Total 137.13 293.94 324.74


78

ERHAN SULEJMANI MUHAMET DEMIRI

Department of Food Technology, University of Tetova, Tetovo,

North Macedonia

NAUČNI RAD

ODREĐIVANJE ISPARLJIVIH JEDINJENJA U PROCESU ZRENJA MAKEDONSKE FERMENTISANE KOBASICE POMOĆU SPME/GC-MS

Profili isparljivih jedinjenja makedonske suvo fermentisane kobasice određeni su gas-nom hromatografijom-masenom spektrometrijom (GC–MS) korišćenjem mikroekstrakcije čvrstom fazom (SPME). Identifikovana su ukupno 103 isparljiva jedinjenja, i to: 12 kise-lina, 16 ketona, 21 terpena, 20 alkohola, 9 estera, 13 ugljovodonika i 12 ostalih. Analiza glavnih isparljivih aromatičnih komponenti pokazala je da 62% jedinjenja koja su otkri-vena prvih dana iz grupe terpena, pri čemu njihov sadržaj raste do 69 i 80% osmog i petnaestog dana, redom. Rezultat je da je 50% ukupnih isparljivih sastojaka koncentri-sano poslednjeg (15.) dana zrenja. Oni sadrže najviše kiselina (57%), ketona (68%) i terpena (56%). Uočene su značajne razlike između različitih perioda fermentacije. Makedonske suve kobasice sadrže visoke koncentracije pikantnih terpena, koje bi mogle igrati važnu ulogu u opštim notama ukusa ovog mesnog proizvoda.

Ključne reči: kiseline, terpeni, ketoni, ugljovodonici, kobasica.




79

MARIJA JOKANOVIĆ1

BOJANA IKONIĆ1

PREDRAG IKONIĆ2

VLADIMIR TOMOVIĆ1

TATJANA PEULIĆ2

BRANISLAV ŠOJIĆ1

SNEŽANA ŠKALJAC1 MAJA IVIĆ1

JELENA IVETIĆ3 1University of Novi Sad, Faculty of

Technology, Novi Sad, Serbia 2University of Novi Sad, Institute of

Food Technology, Novi Sad, Serbia

3University of Novi Sad, Faculty of Technical Sciences, Novi Sad,

Serbia

SCIENTIFIC PAPER

UDC

TOWARDS REPRODUCIBILITY OF TRADITIONAL FERMENTED SAUSAGES: TEXTURE PROFILE ANALYSES AND MODELLING

Article Highlights • Texture profiles of three traditional dry-fermented sausages were investigated • Regression analyses were used to correlate chemical composition and texture char-

acteristics • Differences were significant between sausages of the same type produced in different

facilities • Regression analyses showed that moisture content was the most effective for texture

profile • PCA revealed the best reproducibility of analysed characteristics for Petrovská klo-

basá Abstract

The aim of this study was to investigate textural characteristics of three tradit-ional dry fermented sausages (Sremski kulen, Lemeški kulen and Petrovská klobása) manufactured in different small-scale facilities in northern Serbia, and to correlate them with physicochemical and sensory characteristics. The sample sausages were supplied by different local traditional producers. The textural characteristics were correlated with physicochemical and sensory characteristics using multiple linear regression analysis and principal compo-nent analysis. Differences in physicochemical characteristics reflected even more notable differences in texture characteristics. Regarding regression equations, obtained results showed that moisture content was significant for hardness, springiness and cohesiveness. Hardness was also influenced by fat content, while chewiness was influenced by protein content. Principal compo-nent analysis separated samples of Petrovská klobása, as the group with the most reproducible analysed characteristics. Obtained results of statistical ana-lyses should provide knowledge for possible improvements of the traditional production, in a way that these sausages could be produced in different facil-ities with consistent textural characteristics.

Keywords: dry fermented sausage, MLR, PCA, texture analysis.

Almost all European countries have cultural traditions linked with specific traditional foods, which are recently attracting more interest from both res-earchers and consumers. Consumers are attracted by traditional products for their peculiar characteristics due to the use of specific raw materials, authentic ing-

Correspondence: B. Ikonić, University of Novi Sad, Faculty of Technology, Bulevar cara Lazara 1, 21000 Novi Sad, Serbia. E-mail: [email protected] Paper received: 24 December, 2018 Paper revised: 16 August, 2019 Paper accepted: 30 August, 2019

https://doi.org/10.2298/CICEQ181224027J

redients, application of some manufacturing methods passed on from generation to generation and region specific environmental and climatic conditions. In order to promote rural areas and support the local population, the European Union got involved in the protection of high-quality traditional foods from spe-cific regions or areas [1-4].

In the case of the Republic of Serbia, a number of traditional dry fermented sausages are produced in the northern part of the country (Autonomous Pro-vince of Vojvodina). Because of the specific and recognizable quality, three of them (Sremski kulen, Lemeški kulen and Petrovská klobása) have been

M. JOKANOVIĆ et al.: TOWARDS REPRODUCIBILITY OF TRADITIONAL... Chem. Ind. Chem. Eng. Q. 26 (1) 79−87 (2020)

80

granted a protected designation of origin (PDO) label, according to Serbian legislation [5,6]. They are hand-crafted, usually in small-scale facilities during winter, when atmospheric temperatures are around 0 °C or lower. In the traditional production of these sausages, curing salts and starter cultures are not used, and the required fermenting microorganisms originate from the meat itself or from the environment. Sausage bat-ter is stuffed in pig appendix (Sremski kulen, Lemeški kulen) or pig large intestines (Petrovská klobasá), and undergoes slow drying and ripening processes in traditional smoking/drying rooms [6-8].

Traditional practices in the small scale facilities lead to great variability in products' properties (hetero-geneous quality), as there is no strict uniformity in the product manufacturing by different homemade pro-cessors [1,3,9]. The texture is one of the most import-ant components of meat products’ quality. Many fac-tors affect the final texture of fermented sausages, including ingredients used, processing parameters, acidification method, drying/ripening conditions, as well as interactions among these factors over an ext-ended period of time [10-13]. As it is one of the most important components of the sausage quality, it is important to create a product of textural attributes accepted by the consumer, but it is also important to ensure the product's quality reproducibility, i.e., a low variability of product characteristics [14]. However, insufficient work has been conducted so far to assess the reproducibility of the texture quality of traditional dry fermented sausages [15].

Multiple linear regression (MLR) is a mathemat-ical tool that generates an equation to describe the statistical relationship between a dependent variable and one or more independent variables, and could be used to predict the textural parameters using the physicochemical parameters as independent vari-ables. Also, evaluation of products’ quality variability requires a collection of numerous different data [14]. Using principal component analysis (PCA), the total number of variables could be reduced, and original variables could be transformed into new factors or principal components [4].

Considering the high interest of consumers for traditional dry fermented sausages and the lack of information about the texture characteristics of these products, which could contribute to their character-ization, the aims of this work were:

− to determine and compare physicochemical, sensory and textural characteristics of three tradit-ional dry fermented sausage types (Sremski kulen, Lemeški kulen and Petrovská klobasá) produced in different small scale facilities;

− to explore the correlation of texture with physicochemical and sensory characteristics;

− to obtain a deeper insight into the intercor-relation of investigated characteristics in order to understand factors leading to quality reproducibility.

EXPERIMENTAL

Samples

Three Serbian traditional dry fermented sausage types were considered in this study: Sremski kulen - S (samples S1-S4), Lemeški kulen - L (samples L1- –L6), and Petrovská klobása - P (samples P1-P7). The manufacturing procedures for all three types of sausages are shown in Figure 1. The sausages of Sremski kulen and Lemeški kulen have a weight of 700-800 g and the diameter varies from 80 to 95 mm, whereas the weight of Petrovská klobása sausages is 500-600 g and the diameter ranges from 40 to 50 mm. The sample sausages were supplied in triplicate (different batches) by different local traditional pro-ducers and were stored at 4 °C before textural testing and sensory evaluation.

Figure 1. Manufacturing procedures of three Serbian traditional dry fermented sausages: S – Sremski kulen; L - Lemeški kulen;

P - Petrovská klobása.

Chemical composition analyses

The pH was measured using the portable pH meter Testo 205 (Testo AG, USA) equipped with a combined penetration tip with temperature probe [16]. The measurements were performed by direct penet-ration in different areas of the internal part of the sau-sage. The pH meter was calibrated before and during


81

the readings using standard phosphate buffers (pH value of calibration buffers was 7.00 and 4.01 at 25 °C. Measurements were performed in triplicate. Sau-sage samples were homogenized (Waring 8010ES Blender, USA; capacity 1 L, speed 18000 rpm, dur-ation of homogenization 10 s, temperature after homogenization <10 °C), packaged in polyethylene bags and stored at –20 °C until the determination of proximate chemical composition. Moisture content was determined by drying at 130 °C [17]; nitrogen concentration (N) was determined by the Kjeldahl method; protein content was calculated as total N×6.25 [18] and fat content was obtained by Soxhlet extraction using petroleum ether [19]. All analyses were performed in duplicate.

Texture profile analysis

Texture profile analysis (TPA) was performed as described by Bourne [20], at room temperature, using TA.HDplus texture analyser (Stable Micro Systems, Godalming, UK) equipped with a standard cylindrical plate of 75 mm in diameter. The samples (cylinders) 2 cm thick and 2.54 cm in diameter, after discarding the external layer of the sausage, were compressed twice to 50% of their original thickness at a constant test speed of 1 mm/s. The following parameters were det-ermined: hardness (kg), springiness, cohesiveness and chewiness (kg). Hardness was defined by peak force during the first compression cycle. Springiness was defined as the rate at which a deformed sample goes back to its undeformed condition after the def-orming force is removed. Cohesiveness was calcul-ated as the ratio of the area under the second curve to the area under the first curve. Finally, chewiness was obtained by multiplying hardness, cohesiveness and springiness.

Colour measurements

Colour measurements were performed on the fresh cut of the sausage at room temperature using MINOLTA Chroma Meter CR-400 (Minolta Co., Ltd., Osaka, Japan); D-65 lighting, 2° standard observer angle; 8-mm aperture in the measuring head. Before each set of measurements, the instrument was calib-rated using a white ceramic tile (CR-A43). Colour measurements were always performed in the center of the sausage sample. Sausage colour character-istics were expressed by CIE L*a*b* system (L* – lightness, a* – redness, b* – yellowness) [21].

Sensory evaluation

The sensory analyses were performed by the 9-member panel previously trained in descriptive ana-lysis for different meat products [22]. Panel tasters

were asked to score samples by using a 1–9 scale for each attribute to be evaluated, where 1 = very low and 9 = very high intensity. The descriptors con-sidered chewiness, cohesiveness, and juiciness. For the assessment casing was removed and the sau-sages were cut into slices of approximately 3 mm thickness and served at room temperature on white plastic dishes. Water and unsalted toasts were pro-vided to cleanse the palate between samples. Assess-ments were performed under natural light.

Statistical analysis

The statistical analysis was carried out using software packages Statistica version 12 and Minitab version 17. The analyses were conducted across all sausages types and the differences in investigated characteristics were analysed using ANOVA. Multiple linear regression models were fitted to estimate tex-tural parameters such as hardness, springiness, cohesiveness and chewiness using physicochemical parameters (moisture, protein, and fat content, as well as pH) as independent variables. The best regression model for each dependent variable was based on the value of the coefficient of determination (R2), F-value and p-value. Coefficient of determination is a mea-sure of the extent to which the total variation of the dependent variable is explained by the regression which means how close the data are to the fitted reg-ression line. The F-value and the p-value are used to decide whether the model as a whole has the statis-tically significant predictive capability (at the 95% confidence level, the upper critical values of the F-dis-tribution were given in the Statistica Handbook (NIST/ /SEMATECH e-Handbook of Statistical Methods)). Significance of each coefficient in the model was det-ermined using a t-test, at the 5% significance level. The larger the magnitude of the t-value the more sig-nificant is the corresponding coefficient. PCA calcul-ation was performed to obtain an insight into the rel-ationships of data obtained by TPA test, physico-chemical and sensory analysis and to reduce the number of relevant variables.


The results of physicochemical, textural and sensory analyses of all three groups of traditional dry fermented sausages with basic statistics for charac-terization of the variability of the analysed samples, as well as ANOVA results, are given in Table 1.

Physicochemical characteristics

Based on the average moisture content, all three groups of samples were within the recommended values


82

for dry fermented sausages [23], i.e., lower than 35%. Sausages from group P had, on average, the lowest (p < 0.05) moisture content (21.56%). Consequently, their average fat content was the highest (37.91%), followed by L and S sausages, with significant differences between all of them. The lower moisture content in P sausages could be due to the difference in diameter since ripening processes were in approximately the same environmental conditions for the same period of time. Determined fat contents were in accordance with the fact that higher fat con-tent is a dominant characteristic of dry fermented sau-sages [15,24,25]. Fat contributes to the sensory char-acteristics but also has a technological function in the manufacture of dry fermented sausages as it helps the continuous release of moisture from the inner layer of the sausage, a process necessary for undisturbed

fermentation and flavour development [24]. Average protein contents for all three groups of samples were within the recommended values for dry fermented sausages, i.e., minimally 24% [23]. Protein content od S sausages (40.42%) was higher (p < 0.05) than that of P sausages (32.39%). Further, samples of P group showed the lowest (p > 0.05) pH value, with an aver-age value of 5.36, whereas S and L sausages had average pH value of 5.69 and 5.60, respectively. These values are in accordance with the previous results for naturally fermented dry sausages [6,7,26], as well as with the fact that these sausages are gen-erally characterised by low acidity [15]. Low process (environmental) temperatures limit the intensity of fer-mentation, and thus the pH does not decrease by more than 0.2–0.4 units, as it was previously found [27-29]. During the drying and maturation phases, the

Table 1. Textural, physicochemical and sensory characteristics of traditional dry fermented sausages; HR - hardness [kg]; SP – springiness; CH – cohesiveness; CW - chewiness [kg]; W – water content (g/100 g); P – protein content (g/100 g); F – fat content (g/100 g); PH – pH value; D – diameter (mm); Chew - chewiness; Juic – juiciness; Cohe – cohesiveness; L* - lighteness; a* - redness, b* - yelowness; SD – standard deviation; * a–c: means within the same column with different superscript letters are different (p < 0.05)

Sample

Characteristics

Textural Physicochemical Sensory Colour

HD SP CH CW W P F pH D Chew Juic Cohe L* a* b*

S1 11.78 0.531 0.404 2.49 35.71 36.96 17.07 5.51 95.00 6.64 6.11 6.71 33.76 21.45 15.23

S2 8.02 0.583 0.469 2.19 37.21 35.85 18.95 5.70 80.67 5.71 8.74 6.68 36.08 20.17 14.95

S3 24.76 0.470 0.407 4.73 30.71 48.92 12.43 5.99 90.67 5.72 7.22 7.74 29.78 17.03 11.15

S4 13.24 0.525 0.471 3.27 35.41 39.95 14.98 5.55 93.33 5.00 9.00 8.91 30.04 16.26 11.27

Mean 14.45 0.53a* 0.44 3.1a 34.76 a 40.42 a 15.86 c 5.69 89.92 a 5.77 a 7.77 7.51 32.41 18.73 b 13.15 b

SD 7.21 0.05 0.04 1.13 2.81 5.93 2.80 0.22 6.42 0.67 1.36 1.05 3.05 2.48 2.24

CV 43.23 7.65 7.41 31.02 7.01 12.70 15.30 3.34 6.18 10.11 15.14 12.16 8.14 11.47 14.76

L1 13.71 0.424 0.300 1.78 31.98 37.29 18.19 5.38 81.00 4.64 7.52 7.32 33.41 23.33 20.24

L2 10.56 0.417 0.380 1.66 36.36 32.94 20.68 5.10 86.00 5.00 8.87 7.29 39.18 29.61 25.98

L3 4.73 0.409 0.318 0.61 30.68 34.39 28.80 5.53 90.00 4.28 6.73 8.55 31.60 24.27 16.52

L4 18.72 0.395 0.363 2.64 27.00 44.08 20.35 5.57 88.00 4.55 7.42 7.14 29.85 21.38 15.11

L5 5.75 0.490 0.556 1.58 39.84 36.67 15.75 6.02 86.33 4.64 6.73 6.55 28.73 21.80 15.21

L6 10.29 0.588 0.432 2.59 31.85 34.46 25.52 6.04 95.33 5.30 5.00 4.00 29.59 21.36 15.96

Mean 10.63 0.45b 0.39 1.81 b 32.95 a 36.64 ab 21.55 b 5.60 87.78 a 4.74 b 7.05 5.90 32.06 23.63 a 18.17a

SD 5.17 0.07 0.09 0.74 4.51 3.98 4.81 0.37 4.76 0.36 1.27 1.52 3.87 3.16 4.27

CV 44.43 14.75 21.75 37.73 12.50 9.92 20.36 6.00 4.95 6.94 16.47 20.41 11.01 12.20 21.45

P1 10.70 0.426 0.261 1.19 22.03 35.54 35.67 5.49 43.33 6.07 7.52 7.07 33.62 20.94 17.21

P2 5.48 0.410 0.536 1.23 24.02 34.65 34.39 5.17 47.33 6.36 6.88 3.00 33.74 24.01 18.63

P3 3.66 0.328 0.284 0.34 23.24 27.86 40.66 5.33 46.33 5.14 7.59 5.71 33.82 26.58 19.99

P4 2.93 0.377 0.272 0.30 22.70 28.54 38.95 5.52 43.67 4.96 9.19 5.79 31.47 24.21 17.56

P5 12.38 0.376 0.244 1.14 20.25 31.76 38.32 5.34 43.50 5.70 4.10 4.10 32.29 23.75 17.04

P6 9.82 0.399 0.265 1.03 18.78 34.81 38.68 5.48 44.25 6.62 6.08 7.40 32.70 24.87 18.20

P7 10.07 0.408 0.357 1.43 20.51 33.54 38.70 5.23 43.75 5.63 7.87 7.64 33.66 26.13 20.84

Mean 7.86 0.39b 0.32 0.96 b 21.65 b 32.39 b 37.91a 5.36 44.60 b 5.78 a 7.03 5.82 33.04 24.36 a 18.50a

SD 3.76 0.03 0.10 0.45 1.87 3.11 2.14 0.14 1.58 0.61 1.60 1.75 0.91 1.85 1.44

CV 44.26 7.71 30.14 43.88 7.98 8.88 5.22 2.37 3.28 9.75 21.11 27.80 2.55 7.04 7.22


83

pH may return to similar values to those of the rip-ened meat due to the liberation of peptides, amino acids and ammonia from proteolytic reactions [29].

Textural characteristics

The highest mean hardness value was deter-mined for sausages S (14.45 kg), followed by L (10.63 kg) and P (7.86 kg) sausages. The differences among mean hardness values were not significant (p > 0.05), probably because of high variability among measure-ments within each sausage type, with almost the same coefficients of variability for all three experimen-tal groups. Hardness values obtained in the present study for P sausages, were in agreement with results reported for sucuk [30], while the ones obtained for S and L sausages were higher than those for sucuk [30], chorizo de Pamplona [25], and Italian low-acid sausage [31]. Cohesiveness mean values, ranging from 0.32 to 0.44 did not differ significantly among samples. According to Spaziani et al. [31], the lack of variability in cohesiveness could be a consequence of pH value being close to or over the isoelectric point. Springiness was the highest (p < 0.05) for S (0.53) sausages followed by L (0.45) and P (0.39) sausages, which did not differ significantly. Chewiness value had the same trend as hardness and springiness, being highest (p < 0.05) for S sausages (3.17 kg), and fol-lowed by L (1.81 kg) and P (0.96 kg) sausages, with-out significant difference between the last two. The major changes in texture take place in the ferment-ation process when the pH declines and the solubil-ized miofibrillar proteins aggregate to form a gel, thus transforming the meat mixture into a ripened sau-sage. In addition to fermentation, drying is an impor-tant factor affecting texture properties [11,30,31].

Colour and sensory characteristics

The quality of meat and meat products is often estimated based on colour and appearance [30]. Average L* values of analysed three groups of sau-sages were in the range of 32.06-33.04, with no sig-nificant differences (p > 0.05) between them. Regard-ing a* and b* values S sausages were significantly the lowest (p > 0.05), while L and P did not differ sig-nificantly. Colour characteristics values were lower than determined for chorizo de Pamplona [25] but higher than the ones reported for sucuk [30]. Results of sensory analyses revealed no significant differ-ences among analysed groups of sausage samples for juiciness and cohesiveness, while chewiness was found to be the lowest (p < 0.05) for L sausages.

Multiple linear regression

In order to better understand the effect of phys-icochemical characteristics (independent variables) on the textural parameter of sausages (dependent variables), a series of MLR analyses were performed. Firstly, models for each of the four textural outcome variables included a linear combination of all physico-chemical predictor variables. Considering the obtained results (not reported), it was observed that regression models for springiness and chewiness were statis-tically significant (significant overall F-statistics), although each individual coefficient in the model was not (not-significant t-tests for the individual coef-ficients). This often indicates possible multicollinearity among the predictor variables [32]. Multicollinearity occurs when the model includes multiple factors that are correlated not just to the response variable, but also to each other.

The variance inflation factor (VIF), measures how much the variance of the estimated regression coefficient is “inflated” by the existence of correlation among the predictor variables in the model, and help to detect multicollinearity. The literature data indicates VIF exceeding 10 is a sign of serious multicollinearity requiring correction [33]. Multicollinearity does not affect the goodness of fit and the goodness of predict-ion, but it can be a problem if the purpose of the study is to estimate the contributions of individual pred-ictors. For example, the VIF values were 28.04, 13.78, 49.78 and 1.60 for moisture, protein and fat content and pH value, respectively, and indicate a serious multicollinearity problem. In order to over-come this issue, a stepwise regression was per-formed. Results of stepwise MLR analysis, using Minitab v17, are given in Table 2.

Regarding the regression equation for hardness, R2 indicated that the model explained 87.4% of the variability, and both predictors in the model (moisture and fat content) are statistically significant according to their t-values. Also, a negative correlation between dependent and independent variables (negative values of regression coefficients) was observed, meaning that hardness decreases when moisture and fat con-tents increase. Olivares et al. [24], analysing texture parameters based on fat content, found a significant increase in hardness due to fat reduction only at longer ripening times due to the loss of moisture. Contrary to our findings, authors [11,25] found a sig-nificant correlation between hardness and pH value. Additionally, the significant positive correlation between protein content and hardness was found (Pearson r = 0.86), but this predictor was omitted from


84

the stepwise MLR due to the high negative correlation with fat content (Pearson r = -0.72).

The regression analysis for the springiness showed that only moisture content played an import-ant role in the model (positive correlation), and the model explained 61.4% of the variability in the sprin-giness. The same correlation was determined by Boz-kurt and Bayram [30], while Gonzalez-Fernandez et al. [11] found a significant correlation between sprin-giness and pH value. Pearson correlation between fat content and springiness was significant (-0.65) in this study.

The third model positively correlated cohesive-ness with moisture content, but the value of the coef-ficient of determination was only 0.439, meaning that the model can explain less than 50% of the variability. Bozkurt and Bayram [30] also found a correlation between cohesiveness and moisture content, whereas other authors found a significant correlation between cohesiveness and fat content [24] and between coh-esiveness and pH value [11,25]. The significant neg-ative correlation between fat content and cohesive-ness was also found (Pearson r = -0.59) in this study, but this predictor was omitted from the stepwise MLR due to the high negative correlation with moisture content (Pearson r = -0.88).

Finally, highly significant MLR (R2 = 0.811) was found between chewiness and chemical composition (protein and fat content), but only protein content had a significant influence on chewiness, according to its t-value, and these two variables were positively cor-related. However, the Pearson correlation between fat content and chewiness was significant (-0.78). In dif-ferent studies, authors determined correlations between chewiness and fat content [24], between chewiness and moisture content [30], and between chewiness and pH value [11,25].

Comparing the four models obtained by MLR, it could be concluded that some textural characteristics, namely hardness and chewiness, were more influ-enced by physicochemical characteristics than others. This implies that special attention should be paid to raw materials used in the production. Particularly, the fat content should be tightly controlled, because it was significantly correlated with all analysed textural characteristics either by MLR or Pearson correlation.

Principal component analysis

For a global view, the principal component analysis was performed using the data of instrumental (textural and colour measurements), physicochem-ical, and sensory characteristics of the fermented sausages samples (Figures 2 and 3).

Three main components (F1, F2 and F3) were used, which accounted for 73.77% of the total vari-ance data, and the individual contributions of the components were 45.90, 15.87 and 12.00%, respect-ively. The fact that three dimensions were used, and explained only 73.77%, suggested the wide variability of the sample regarding the analysed parameters what was linked to different raw material/formulation and processing parameters. First PC (F1) was asso-ciated with most of the physicochemical parameters and redness. The second PC (F2) was mostly loaded by sensory and colour parameters, and the third one was mostly associated with textural characteristics.

Through first two components, it was only pos-sible to separate samples of P sausages, which were allocated on the negative quadrant of F1, highly influ-enced by the fat content of the samples, and negat-ively correlated with moisture content and diameter size. As it was previously shown these samples had the highest fat content, with the lowest coefficient of variation. Data for other samples were scattered mostly in the positive quadrant of F1.

Table 2. Results of stepwise MLR analysis; R2 - Coefficient of determination; *p < 0.05; VIF - variance inflation fact

Dependent variable R2 F-value p-value Independent variable Regression coefficient t-value VIF

Hardness 0.874 48.64 <0.0005 Constant 76074 10.39*

Moisture content, % -1298 -8.07* 4.44

Fat content, % -1054 -9.79* 4.44

Springiness 0.614 11.14 0.001 Constant -0.293 -1.17

Moisture content, % 0.0058 2.90* 1.23

pH 0.1032 2.11 1.23

Cohesiveness 0.439 11.75 0.004 Constant 0.1003 1.23

Moisture content, % 0.00945 3.43* 1.00

Chewiness 0.811 30.06 <0.0005 Constant -2603 -1.50

Protein content, % 147.1 3.93* 2.11

Fat content, % -32.6 -1.75 2.11


85

Regarding instrumental colour characteristics, F2 cooperated with separating sample L2, which had highest L*, a* and b* values, comparing with all other samples, i.e., it was the lightest and with the highest share of red colour. Lightness (L*) seems to be the most informative parameter for colour changes, but the importance of red (a*) should not be ignored [25]. Further, the third PC (F3) was associated with hard-ness and protein content what was important in the characterisation of S3, S4 and L4 samples, since these samples were the highest in these character-istics compared to the others. Also, springiness and cohesiveness position on the positive quadrant of F3 was important for the characterisation of samples S1, S2, L5 and L6 being the highest in these values com-pared to other samples, and also being the highest in moisture content which was positively correlated with these two textural parameters. Analysed sensory pro-perties of sausages (chewiness, cohesiveness and juiciness) did not make any difference for the charac-

terisation of samples, and no significant correlations (p > 0.05, data not shown) were determined between instrumentally and sensory determined texture para-meters.

The results of PCA lead to the following conclus-ions. Firstly, it could be observed that Petrovská klo-basá sausages are clearly distinct from the other two types of sausages. Secondly, all three types of sau-sages exhibited high variation which is evident from the cluster sizes (Figure 3). Considering the analysed parameters, a grouping of likewise variables into fac-tors was evident, to some extent, although with sev-eral exceptions. More comprehensive and informative results regarding the relationship between the type of sausage and their quality characteristics could be obtained by using canonical correlation analysis (CCA), however, due to the very limited sample size, it was not possible to perform this type of analysis in this study.

Figure 2. PCA loading plots in the plane of factors 2 and 3 vs. factor 1 – the labels correspond to variables listed in Table 1.

Figure 3. Results of the PCA – score plots in the plane of factors 2 and 3 vs. factor 1 – the labels correspond to samples given in Table 1.


86

CONCLUSION

Although the recipes and procedures in the trad-itional production of Sremski kulen, Lemeški kulen and Petrovská klobasá are well known, obtained results showed differences in physicochemical and sensory characteristics even between sausages of the same type produced in different small scale facil-ities. These differences reflected to even more not-able differences in texture characteristics. Obtained results of statistical analyses should provide know-ledge for possible improvements of the traditional pro-duction, in a way that these sausages could be pro-duced in different facilities with consistent textural characteristics.

Acknowledgements

Research was financially supported by the Min-istry of Education, Science and Technological Dev-elopment, the Republic of Serbia, projects TR31032 and III44006.

REFERENCES

[1] L. Aquilanti, S. Santarelli, G. Silvestri, A.Osimani, A. Petruzzelli, F.Clementi, Food Microbiol. 120 (2007) 136- –145

[2] J. Jordana, Food Res Int. 33 (2000) 147-152

[3] R. Talon, S. Leroy, I. Lebert, Meat Sci. 77 (2007) 55-62

[4] P. Wang, M.Y. Han, Q.L. Dong, X.L. Xu, G.H. Zhou, S.L. Lu, J. Texture Stud. 41 (2010) 757-773

[5] Lj. Petrović, N. Džinić, V. Tomović, P. Ikonić, T. Tasić, Republic of Serbia: Intellectual Property Office Decision No. 9652/06 G-03/06, 2007

[6] I. Vuković, S. Saičić, D.Vasilev, Tehnologija mesa 52 (2011) 134-140

[7] P. Ikonić, T.Tasić, Lj. Petrović, S. Škaljac, M. Jokanović, A. Mandić, B. Ikonić, Food Control 30 (2013) 69-75

[8] T. Tasić, P. Ikonić, A. Mandić, M. Jokanović, V. Tomović, S. Savatić, Lj. Petrović, Food Control 23 (2012) 107-112

[9] A. Salgado, M.C. Garcia Fontan, I. Franco, M. Lopez, J. Carballo, Food Control 17 (2006) 213-221

[10] A.S. Szczesniak, Food Qual. Prefer. 13 (2002) 215-225

[11] C. Gonzalez-Fernandez, E.M. Santos, J. Rovira, I. Jaime, Meat Sci. 74 (2006) 467-475

[12] S. Barbut, in Handbook of Fermented Meat and Poultry, F. Toldra, Y.H. Hui, I. Astiasarán, W.K. Nip, J.G.

Sebranek, E.T.F. Silveira, L.H. Stahnke, R. Talon (Eds.), Blackwell Publishing, Ames, IA, 2007, p. 217

[13] A. Bagdatli, A. Kundakci, J Food Sci Technol. 53 (2016) 1679-1689

[14] G. Probola, L. Zander, J Food Eng. 83 (2007) 93-98

[15] J. Rason, A. Laguet, P. Berge, E. Dufour, A. Lebecque, Meat Sci. 75 (2007) 359-370

[16] ISO 2917, International Organisation for Standardisation (1999)




[20] M.C. Bourne, Food Technol. 32 (1978) 62-66

[21] CIE, International Commission on Illumination, Color-imetry: Official Recommendation of the International Commission on Illumination, Publication CIE No. (E- -1.31), Bureau Central de la CIE Paris, 1976

[22] ISO 8586–1, International Organisation for Standard-isation (1993)

[23] Serbian Legislation 94- 2015, Official Gazette of the SCG (2015)

[24] A. Olivares, J. L. Navarro, A. Salvador, M. Flores, Meat Sci. 86 (2010) 251-257

[25] O. Gimeno, D. Ansorena, I. Astiasaran, J. Bello, Food Chem. 69 (2000) 195-200

[26] P. Ikonić, Lj. Petrović, T. Tasić, M. Jokanović, S. Savatić, B. Ikonić, N. Džinić, Chem. Ind. Chem. Eng. Q. 18 (2012) 163-169

[27] S. Bover-Cid, M. Izquierdo-Pulido, C. M. Vidal-Carou, Meat Sci. 57 (2001) 215-221

[28] P. Ikonić, M. Jokanović, Lj. Petrović, T. Tasić, S. Škaljac, B. Šojić, N. Džinić, V. Tomović, J. Tomić, B. Danilović, B. Ikonić, Int J Food Prop. 19 (2016) 1924-1937

[29] G. Kaban, M. Kaya, J Food Sci. 74 (2009) 58-63

[30] H. Bozkurt, M. Bayram, Meat Sci. 73 (2006) 344-350

[31] M. Spaziani, M. Del Torre, L. Stecchini, Meat Sci. 81 (2009) 77-85

[32] L. Simon, Penn State lectures, https://onlinecourses.sci-ence.psu.edu/stat501/node/347 (accessed 11 May 2018)

[33] Mathworks Support, http://www.mathworks.com/help/ stats/f-statistic-and-t-statistic.html (accessed 24 May 2018).


87

MARIJA JOKANOVIĆ1

BOJANA IKONIĆ1

PREDRAG IKONIĆ2

VLADIMIR TOMOVIĆ1

TATJANA PEULIĆ2

BRANISLAV ŠOJIĆ1

SNEŽANA ŠKALJAC1 MAJA IVIĆ1

JELENA IVETIĆ3 1Univerzitet u Novom Sadu, Tehnološki fakultet Novi Sad, Bulevar cara Lazara

1, 21000 Novi Sad, Srbija 2Univerzitet u Novom Sadu, Naučni

institut za prehrambene tehnologije u Novom Sadu, Bulevar cara Lazara 1,

21000 Novi Sad, Srbija 3Univerzitet u Novom Sadu, Fakultet

tehničkih nauka, Trg Dositeja Obradovića 6, 21000 Novi Sad, Srbija

NAUČNI RAD

POSTIZANJE REPRODUCIBILNOSTI U PROIZVOD-NJI TRADICIONALNIH FERMENTISANIH KOBASICA: ANALIZA TEKSTURE I MODELOVANJE

Cilj ovog rada bio je da se ispitaju karakteristike teksture tri tradicionalne fermentisane suve kobasice (Sremski kulen, Lemeški kulen i Petrovská klobása) proizvedene u razli-čitim malim objektima za preradu mesa na severu Srbije, i da se dobijeni rezultati povežu sa fizičko-hemijskim i senzorskim karakteristikama. Uzorci za analizu su nabav-ljeni od različitih lokalnih proizvođača tradicionalnih kobasica. Karakteristike teksture su korelirane sa fizičko-hemijskim i senzorskim karakteristikama korišćenjem višestruke linearne regresione analize i analize glavnih komponenti. Razlike u fizičko-hemijskim karakteristikama odražavale su se još značajnije na razlike u karakteristikama teksture. Rezultati regresione analize pokazali su da je sadržaj vlage značajan za tvrdoću, elas-tičnost i kohezivnost kobasica. Na tvrdoću je značajno uticao i sadržaj masti, dok je na žvakljivost uticaj imao i sadržaj proteina. Analiza glavnih komponenti izdvojila je uzorke Petrovská klobása kao grupe sa najvećom ponovljivošću analiziranih karakteristika. Dobijeni rezultati statističkih analiza treba da ukažu na mogućnost poboljšanja tradicio-nalne proizvodnje na način da kobasice proizvedene u različitim tradicionalnim objek-tima imaju konzistentne karakteristike teksture.

Ključne reči: fermentisane suve kobasice, MLR, PCA, analiza teksture.




89

EVGENY AKULININ

OLEG GOLUBYATNIKOV DMITRY DVORETSKY

STANISLAV DVORETSKY

Tambov State Technical University, Tambov, Russia

SCIENTIFIC PAPER

UDC 66.074.5.081.3:66.074.35:66.083

OPTIMIZATION AND ANALYSIS OF PRESSURE SWING ADSORPTION PROCESS FOR OXYGEN PRODUCTION FROM AIR UNDER UNCERTAINTY

Article Highlights • PSA oxygen production unit is optimized under uncertainty • Optimal modes of the PSA oxygen production unit are studied • A constraint on gas flow velocity in the frontal layer of adsorbent of the PSA unit is

introduced • Limiting gas flow velocity provides for resource saving of granulated adsorbent • The fulfilment of gas flow velocity constraint is ensured by controlling the PSA unit's

valves Abstract

Pressure swing adsorption (PSA) units are widely used for atmospheric air separation and oxygen concentration. However, the efficiency of such install-ations is reduced due to accidental changes in the characteristics of the atmo-spheric air to be separated. The article formulates and solves the problem of optimizing the regimes of operation of PSA units with zeolite adsorbent 13X, according to the criterion of oxygen recovery rate in the conditions of interval uncertainty of composition, temperature and pressure of atmospheric air. The optimization problem also takes into account the fulfillment of the requirements on purity of oxygen, productivity of the unit and resource saving of granulated adsorbent from granule abrasion. It is proposed to provide adsorbent saving by limiting the speed of incoming flow in the frontal layer of the adsorbent by means of "soft" stepwise change of the degree of opening of control inlet and outlet valves of the unit. The problem (including the search for time change programs for the degree of opening of control valves) was solved with the use of the developed mathematical model of cyclic heat- and mass exchange processes of adsorption-desorption in a PSA unit and a heuristic iterative algo-rithm. The comparative analysis of the results of the optimization problem sol-ution, with and without taking into account the constraint on the gas flow velo-city in the frontal layer of the adsorbent, is carried out. The influence of the specified requirements for the performance of the PSA unit and the purity of oxygen on the degree of its recovery has been studied.

Keywords: pressure swing adsorption, zeolite, mathematical modeling, optimization, numerical simulation, uncertainty.

The most common and effective method of ind-ustrial air separation on solid adsorbents is the method of pressure swing adsorption (PSA), devel-

Correspondence: E.I. Akulinin, Tambov State Technical Univer-sity, Sovetskaya str., 106, 392000 Tambov, Russia. E-mail: [email protected] Paper received: 14 April, 2019 Paper revised: 10 September, 2019 Paper accepted: 18 September, 2019


oped in 1960 in the USA by Charles W. Skarstrom. Concentration of oxygen from the atmospheric air by the PSA method is carried out practically in iso-thermal conditions and allows to obtain a gas mixture containing up to 95.5% of oxygen at the outlet.

The Skarstrom method (PSA) is widely used in chemical engineering [1–20]. In its implementation, adsorbents (mainly zeolites of types 13X, 5A, LiX) with a low adsorption heat in relation to the separated components of gas mixtures are used (typical values

E. AKULININ et al.: OPTIMIZATION AND ANALYSIS OF PRESSURE SWING... Chem. Ind. Chem. Eng. Q. 26 (1) 89−104 (2020)

90

of h/RT for PSA fall within the range of 1-10). In sys-tems where the target component is poorly sorbed (e.g., nitrogen and oxygen in comparison with water vapor in humid air on silica gel or hydrogen in syn-thesis gas on microporous sorbents), almost pure gases can be produced while ensuring their recovery rate up to 85-90%. When differences in adsorption heat for components are insignificant (e.g., nitrogen, oxygen and argon in the air on microporous sorbents), the purity of the product can be achieved only through a significant reduction in its recovery [1,21–24].

Annual growth in oxygen demand averages 4– -5% due to increased demand in ferrous metallurgy, chemical industry, aluminum production and other industries and social sphere. Thus, the essential share of consumers uses not so much pure oxygen, as the air enriched with oxygen from 40 to 90%. [25–28].

In modern adsorption plants for the separation of gas mixtures, Rapid PSA (RPSA) and ultra rapid PSA (URPSA) processes are implemented with the use of multi-adsorber flow diagrams [29–33]. The main idea behind the RPSA and URPSA processes is to reduce cycle times compared to conventional PSA. In RPSA and URPSA units, in comparison with tradit-ional PSA units, the pressure drop increases on the average by an order of magnitude and, accordingly, the length of the adsorbent layer is reduced, which makes it possible to reduce the size of the install-ations by several times and reduce the cycle time up to ten seconds. They are effective when using a fairly narrow range of adsorbent granule sizes (0.25–2 mm).

Increase of gas flow velocity in PSA processes inevitably leads to increase of aerodynamic resist-ance of an adsorbent layer, adsorbent granule abras-ion rate and reduction of its service life. Experimental and computational studies, as well as the analysis of the literature sources [18,24] showed that at the full jump opening of the inlet and outlet valves of the PSA unit, the velocity of the gas flow in the frontal layer of the adsorbent can reach the values of fluidization, when the granules of the sorbent begin to move rel-ative to each other, which leads to abrasion of the adsorbent and the appearance of significant amounts of dust in the product flow.

It is worth noting that even at gas flow rates lower than the rate of fluidization, the abrasion of ads-orbent granules can be quite significant. This is due to the impact on the granules of changing "lateral" forces or the so-called lateral von Karman forces, causing the oscillating displacement of the granules relative to each other. Both destructive effects are likely to occur when the technological stages of ads-orption and desorption change, when large pressure

gradients of the adsorbent layer appear. The pre-sence of dust in the product flow can lead to failure of solenoid valves, with the help of which the gas dis-tribution inside the unit is carried out, which substan-tially increases the operating costs of PSA units [24].

There are a number of design solutions to red-uce adsorbent abrasion in the layer: hardening and reduction of adsorbent granule roughness, direction of filtered flow from top to bottom, immobilization of adsorbent granules in the layer.

One of the effective ways of solving this prob-lem, as suggested in this work, is limiting the gas flow velocity in the frontal layer of the adsorbent, which virtually eliminates the abrasion of adsorbent gra-nules. For this purpose, it is necessary to use control solenoid inlet and outlet valves, the programs of open-ing of which, along with other regime variables, can be determined by solving an optimization problem.

Optimization problems of operation regimes of PSA units for oxygen enrichment were formulated and solved in the works of the authors: Jiang et al. [34], Cruz et al. [35], Santos et al. [21], Liu et al. [36], Hossein-zadeh Hejazi et al. [19] and Ding et al. [37]. These studies considered 2-bed RPSAs [21,34-36], VSA [19,34,35] and PVSA [37] units with a capacity of up to 4 l/min at NTP. Both industrial zeolites 13X [35], 5A [35,36], LiLSX [37] were used as adsorbents, which allow to obtain oxygen with concentration up to 95 vol.%, and promising experimental sorbents AgLiLSX [21] and Ag-ETS-10 [19], which have high argon selectivity and allow to obtain oxygen with con-centration up to 99.5 vol.%, but with a sufficiently low degree of recovery ∼27.3% and lower [19,21].

The works [21,34-36] used oxygen recovery rate as the goal function determining the efficiency of the PSA unit; the works [19,37] used energy consump-tion; the works [35,36] used profit from the use of the PSA unit; and in [37] the productivity of the unit was used. As optimized variables, the following were con-sidered: duration of adsorption and desorption cycle steps [19,21,34-37], valve capacity [21,34,35,37], pressure in the receiver [34], pressure at the steps of adsorption and desorption [19,35,36], rate of dis-charge [36] and inlet flow rate [19], backflow rate [36], air flow rate at the compressor and vacuum pump outlet [37], as well as the length of the adsorbent bulk layer [19]. As constraints in the optimization problem, the operating procedure requirements for oxygen purity [19,21,34,35,37] and productivity of the unit [21] were used.

The work [34] has applied a simultaneous tail-ored methods approach for the solution of the opti-mization problem, which is more economical in the


91

sense of reducing the time spent on obtaining a sol-ution. However, in the majority of other works [19,21,35-37] optimization problems were solved with the use of the so-called black-box approach, which is widely used both for solving practical and scientific problems [34,38].

An important aspect of the study of the effi-ciency of the PSA air separation units is to take into account the influence of uncertain parameters - the composition, temperature and pressure of the atmo-spheric air. During the operation of the PSA unit, uncertain parameters may take on random values in some intervals, depending on the operating condit-ions of the units and, in particular, on the climatic and geographical characteristics. For instance, when operating PSA medical concentrators indoors, the ambient temperature may vary from 293 to 303 K, the pressure may vary from 0.75×105 to 1.0×105 Pa at altitudes up to 2 km above sea level, and the concen-tration of oxygen in the air may vary from 18 to 21%.

A new approach is proposed in this paper to account for the interval uncertainty in the design of PSA air separation units. The idea of the approach is to define such values of the regime parameters, at which the optimal value of the efficiency criterion (for example, the rate of oxygen recovery) is achieved and the requirements for the PSA unit (for the purity of oxygen and the productivity of the unit) are met, regardless of the values of uncertain parameters from

the given intervals of their possible changes. The purpose of this work is to set and solve the

problem of optimization of the regime parameters of the two-adsorber PSA unit for air oxygenation. As an optimization criterion the rate of oxygen recovery, cal-culated in the cyclic steady state (CSS) regime, will be used, and the operating procedure requirements on the purity of oxygen, productivity of the unit and resource saving of the adsorbent will be taken into account.

Mathematical description of pressure swing adsorption processes of oxygen concentration

The technological process of separation of atmospheric air containing oxygen in the amount of 18-21 vol.%, nitrogen 78-80 vol.%, argon and other impurities 1-2 vol.% is carried out in the two-adsorber PSA unit with granulated synthetic zeolite adsorbent 13X [13,14] (Figure 1).

Atmospheric air is supplied to the unit after pre-drying with the overpressure inP from 2×105 to 6×105 Pa. Pressure rise in adsorbers is performed by “soft” opening of control inlet valves v1 (v2), through which air is supplied to the layer of granulated bulk ads-orbent. The bulk of the oxygen-enriched product air stream from the adsorber A1 (A2) is directed to the consumer through the check valve v5 (v7) and product storage-tank STP, while the other part of the stream is directed through the throttle v6 to the adsorber A2 (А1)

Figure 1. Structural diagram of the PSA unit as an object of optimization.


92

for washing the adsorbent layer and desorption of mainly nitrogen, after which through the control outlet valve v4 (v3) it is directed for further processing. The v8 valve is required for manual control of the amount of gas extracted from the product storage-tank STP. The pressure cyclogram in the adsorber A1 is shown in Figure 2a. As the adsorbers operate alternately, the pressure cyclogram for the adsorber A2 is similar to that of the adsorber A1, but with a shift of half a cycle of "adsorption-desorption". Figure 2b shows a cyclo-gram of operation of the valves in the unit. Compared to the standard Skarstrom cycle, which includes 4 steps, the pressure feed and adsorption steps, as well as blowdown and desorption, are combined in this unit as in RPSA units.

The analysis of the technological process of air separation and oxygen concentration by the PSA method [13,14,21-28,34-44] allowed to determine:

1) regime parameters u - duration adst ( = =ads des c / 2t t t ) of adsorption step (half cycle); pressure inP at compressor outlet; backflow coef-ficient θ (for adsorbent regeneration); programs ψ ψ

1 4v v( ), ( )t t , ∈ ads[0, ],t t ψ ψ2 3v v( ), ( ),t t ∈ ads c[ , ]t t t

of opening of control inlet v1, v2 and outlet v3, v4 valves; with ψ ψ ψ= =

1 21 v v( ) ( ),t t ψ ψ ψ= =3 42 v v( ) ( )t t ,

as adsorbers work alternately; 2) design parameters d - diameter of DA of ads-

orber, length L of the layer of granulated adsorbent 13X, granule diameter dgr, throughput capacity vK of inlet and outlet valves of the unit;

3) uncertain parameters ξ = env env env{ , , }y T P - concentration of yenv gases in the air, temperature of Tenv and ambient pressure of Penv;

4) output variables z, which determine the effi-ciency of the PSA unit - the rate of oxygen recovery η , the concentration outy of product flow compo-nents at the outlet of the unit and productivity outG of the PSA unit.

When air (adsorbate) components are ads-orbed, the following mass and heat exchange pro-

cesses take place in the adsorbent layer: 1) diffusion of adsorbate in the gas-air mixture flow; 2) mass and heat exchange between the gas phase and adsor-bent; 3) adsorption of predominantly N2 on the sur-face and in micropores of zeolite adsorbent granules with heat generation and desorption of predominantly N2 from micropores and from the surface of granules with heat absorption.

The analysis of the results of physical modeling of these processes has shown that diffusion of ads-orbate (N2, O2, Ar + impurities) and distribution of heat in gas and solid phases are carried out mainly in the axial direction relative to the movement of the gas mixture along the length of the adsorbent layer. At the same time, the process of enrichment of the gas-air mixture at adsorption (N2, O2, Ar + impurities) by gra-nulated zeolite adsorbent with granule diameter 0.25– –2 mm takes place in the external-diffusion region (determined by the coefficient of external mass trans-fer), as well as equilibrium ratios of adsorbate con-centrations in the phases, which is in line with [21,34,35].

The following assumptions were made in the mathematical description of the air oxygenation pro-cess:

1) initial gas-air mixture is 3-component (con-tains 1 – oxygen O2 with concentration of 18-21 vol.%, 2 – nitrogen N2 with concentration of 78-80 vol.%; 3 – argon Ar and impurities with concentration of 1-2 vol.%) and is considered as ideal gas, which is quite acceptable at pressure in the adsorber up to 200×105 Pa [45];

2) diffusion of adsorbate and heat propagation in gas and solid phases are carried out only in the axial direction of the gas mixture flow in the adsorber (along the length of the adsorbent layer) [1,21–28,46];

3) granulated zeolite 13X with a granule dia-meter of 1.6 mm is used as an adsorbent [14];

4) adsorption equilibrium (adsorption isotherm) is described by the Dubinin-Raduskevich equation [47];

Figure 2. Cyclograms of pressure in adsorber A1 (а) and operation of PSA unit’s valves (b).


93

5) desorption branches of adsorption isotherms (N2, O2, Ar + impurities) for zeolite 13X coincide with adsorption branches [14];

6) gas temperature in the receiver R is equal to the gas temperature at the adsorber outlet, heat losses to the environment are negligibly small.

According to the accepted assumptions, the mathematical model of the PSA process of atmo-spheric air separation and oxygen concentration has the following form (Table 1).

The mathematical model, Eqs. (1)–(12), represents a system of differential equations of the parabolic type in partial derivatives and is supple-mented by the corresponding initial (Table 2) and boundary conditions (Table 3) for the adsorption and desorption steps.

Formulas for calculating model coefficients and parameter values are given in the nomenclature sec-tion.

To solve Eqs. (1)–(12) with initial and boundary conditions we used the method of straight lines [48] in the MatLab software environment, and the solution of the system (1)–(12) was carried out before the onset of the CSS regime [21,34] of the PSA unit. Reaching the CSS regime was determined when the condition was met:

ε−− ≤out out1, 1, 1n ny y (13)

where ε – a small positive number, ε = 1.0×10-3 [21,35]; the time of the CSS onset tst was determined by = *

st ct n t , where n* is the number of the cycle at which the condition (13) is valid. Average comput-ational time (CPU time) of the CSS regime startup was ∼210 s (Win7x64, Intel Core i7-7700, DDRIV 16 Gb).

The analysis of the adequacy of the mathem-atical model, Eqs. (1)–(12), was carried out with the use of the factual RMS error:

Table 1. Mathematical model of the process of air separation and oxygen concentration in the PSA unit

Parameter Expression Eq. No.

Component mass balance ε ν

ε∂ ∂ ∂− ∂ ∂ + + = = ∂ ∂ ∂ ∂ ∂

g g( , ) ( , ) ( , )(1 )

( ) ( , ) , 1,2,3k k kk

y x t a x t y x t D x y x t kt t x x x

(1)

Linear driving force model (kinetics) ( )β∂= −

∂*

sp( , )

( , )kk k k

a x t S a a x tt

(2)

Dubinin-Radushkevich isotherm ρσ −

= − ×

22g S,* 0

a * 2 5exp lg1.013 10 0.0224

kk

k k k

T PWa B

V P y (3)

Heat propagation in the gas-air mixture ( )αρ ρ ν λε

∂ ∂ ∂+ − − =

∂ ∂ ∂

2g g g

g g g g g sp a g g 2

( , ) ( , )( , ) ( , )p p

T x t T x t Tc с S T x t T x t

t x x (4)

Heat propagation in the adsorbent ρ α λ∂ ∂∂ + − − = ∂ ∂ ∂a

2a a

a sp a g a 2

( , ) ( , )( , )( , ) ( , ) k

p kk

T x t T x ta x tс S T x t T x t ht t x

(5)

Flow continuity equation νν

∂ ∂ − =

∂ ∂

g

g 0k

kk

k

yy

x x

(6)

Gas phase momentum balance (Ergun equation)

( )( )

( )ε εμ ν ρ ν

ςες ε

− −∂ = − + ∂

2

2g g g g g2 3

grgr

150 1 11.75

P Mx dd

(7)

Valves equations ( )λψ λ= − = =in in out

v v v v v , 1,4, 1,i i i i i

G K P P i m

( )= −out out outads ads SG G P P

(8)

Throttler equation θ=out out

in ads adsdes in

des

( ) ( )( )

G t P tG

P t(9)

Set pressure equation in the adsorber ( )ε∂

= −∂ ×

in inads/des des v inA

A3 inA60 10

iK P GP P P

t P V (10)

Set pressure equation in the product storage-tank

( )∂= −

∂ ×

out outinS S ads

S3 inS60 10

P K P G P Pt P V (11)

Product storage-tank purging equation ( )∂= − −

∂

outS, out

S,S

( )( )k

k ky G t y y t

t V (12)


94

Table 2. Initial conditions of the mathematical model, Eqs. (1)–(12), for the steps of adsorption and desorption

Adsorption Adsorption Desorption

= = ≤ ≤0, 1, 0t n x L : = × = ≤ ≤ads , 2,3,..., 0 :t n t n x L = × = ≤ ≤des , 1,2,..., 0 :t n t n x L

( ) = 0,0 ( )k ky x y x ( ) =ads desdes, ( , )k ky x t y x t ( ) =des ads

ads, ( , )k ky x t y x t

( ) =,0 0ka x

( ) =ads desdes, ( , )k ka x t a x t ( ) =des ads

ads, ( , )k ka x t a x t

( ) ( )= 0g g,0T x T x ( ) ( )=ads des

g g des, ,T x t T x t ( ) ( )=des adsg g ads, ,T x t T x t

( ) ( )= 0a a,0T x T x ( ) ( )=ads des

a a des, ,T x t T x t ( ) ( )=des adsa a ads, ,T x t T x t

( ) ( )ν ν= 0g g,0x x

( ) ( )ν ν=ads des

g g des, ,x t x t ( ) ( )ν ν=des adsg g ads, ,x t x t

( ) ( )= 0,0P x P x =ads desdes( , ) ( , )P x t P x t =des ads

ads( , ) ( , )P x t P x t

( ) = 0A 0P P ( )=ads des

A A des( )P t P t ( )=des adsA A ads( )P t P t

( ) = 0S 0P P ( )=S S ads( )P t P t

( ) = 0S, S,0k ky y ( ) ( )= out

S, adsk ky t y t

Table 3. Boundary conditions of the mathematical model (1)–(12) for the steps of adsorption and desorption

Adsorption Desorption

x = 0: x = L: x = 0: x = L:

= in(0, ) ( )k ky t y t ∂

=∂

( , )0ky L t

x

∂=

∂(0, )

0ky tx

= ads( , ) ( , )k ky L t y L t

= ing g(0, ) ( )T t T t

∂

=∂g( , )

0T L t

x ∂

=∂

(0, )0gT t

x = ads

g g( , ) ( , )T L t T L t

α∂= −

∂ina

sp a g(0, )

( (0, ) ( ))T t S T t T t

x

∂=

∂a( , )

0T L t

x

∂=

∂a(0, )

0T t

x α∂

= −∂a

sp a g( , )

( ( , ) ( , ))T L t

S T L t T L tx

νε

=in

gA

( )(0, )

G ttS

ν∂

=∂g( , )

0L tx

ν∂

=∂g(0, )

0t

x ν

ε=

indes

gA

( )( , )

G tL t

S

= A(0, ) ( )P t P t = A(0, ) ( )P t P t

( )( )δ=

= −2

out out,e out,eММ 1, 1, 1,

1

1100 ( ) ( ) / ( )

N

i i ii

y t y t y tN

where i is the measurement number, N is the total number of measurements. RMS error δММ at N = 24 was 5.2%, which allows using the mathematical model for technological calculation, optimization of

cyclic regimes and design of PSA units for air separ-ation and oxygen concentration (Figure 3).

Optimization of cyclic processes of the adsorption separation of atmospheric air under uncertainty of parameters

The purpose of the PSA air separation unit is to produce oxygen with a given concentration of out

1,defy

Figure 3. Dependence of concentration out

1y of product oxygen on: a) the duration adst of adsorption step; b) the length of adsorbent layer L of the granulated zeolite adsorbent 13X: 1– grd =1.6 mm, 2– grd = 0.5 mm (, – experimental data, calculation by the model).


95

vol.% and in a given amount (the efficiency of the PSA unit - the flow rate of product air outG with a given concentration of oxygen out

1y , must be not lower than the specified one out

defG ). As a criterion of optimality that determines the

efficiency of oxygen PSA units with productivity up to 4 l/min at NTP, it is advisable to use the rate of oxygen recovery η in the CSS regime, which is noted in [21,34-36]. Oxygen recovery rate η characterizes the share of product flow in relation to the flow sup-plied for separation:

η =out out1

in in1

100y Gy G

We shall introduce constraints that take into account the requirements of the operating procedure of the PSA unit in the CSS regime:

- for the purity of oxygen out1y at the outlet of the

unit, - for the productivity outG of the unit and - gas flow rate gv in the frontal layer of ads-

orbent, which should not be lower than the specified values

out1,defy , out

defG and higher than +gv , respectively.

In addition, we introduce a limit on the pressure drop in the adsorbent layer, calculated using the Ergun Equation (7). The pressure in the closing (final) adsorbent layer at the absorption out

adsP and desorption out

desP steps should be greater than or equal to the atmospheric pressure Penv.

Problem statement. Let vector:

ξ = env env env{ , , }y T P

of uncertainty parameters belong to an area Ξ , ξ ξ ξ− +Ξ = ≤ ≤{ } , i.e., ξ ∈ Ξ . We shall introduce a

set of approximation points ξ i , ∈ 1i J , ξ ∈ 1i S , evenly

covering the uncertainty area ξ ξ ξ− +Ξ = ≤ ≤{ } , and a set of critical points ξ ξ= ∈ Ξ ∈2 2{ : , }l lS l J , in which the above introduced constraints of the optimization problem can be violated.

The problem of optimizing regime parameters of the PSA unit under the interval uncertainty of the parameters ξ = env env env{ , , }y T P is formulated as fol-lows: at fixed values of the vector of design para-meters = A gr v{ , , , }d D L d K it is necessary to deter-mine the regime variables:

λ λθ ψ ψ λ∗ ∗ ∗ ∗ ∗ ∗= =inads 1 2{ , , , , , 1,20}u t P

such that the average value of the rate of oxygen recovery in the CSS regime, Eq. (13), reached the maximum value, i.e.

ω η ξ∈

= 1

*( ) max ( , , )iiu

i J

I u d u (14)

in the form of mathematical model (1)–(12) equations and constraints:

- purity of product oxygen, out1y :

( )( ) ( )ξ ξ= − ≤out out1 st 1,def 1, , , , 0g z d u y y d u (15)

- PSA unit productivity, outG :

( )( ) ( )ξ ξ= − ≤out out2 st def, , , , 0g z d u G G d u (16)

- velocity of gas mixture gv in the frontal layer of adsorbent:

( )( ) ( )ξ ξ +

∈= − ≤

с3 st g g[0, ]

, , max ( , , ) 0t t

g z d u v d u v (17)

- pressure gradient in the adsorbent layer at the adsorption step:

( )( )ξ ξ= − ≤out4 st env ads, , ( , , ) 0g z d u P P d u (18)

- pressure gradient in the adsorbent layer at the desorption step:

( )( )ξ ξ= − ≤out5 st env des, , ( , , ) 0g z d u P P d u (19)

- ranges of change of optimized variables, − +≤ ≤u u u :

≤ ≤ads0.5 20t s, × ≤ ≤ ×5 in 52 10 6 10P Pа,

θ≤ ≤0 6, λψ≤ ≤10 1, λψ≤ ≤20 1, λ = 1,20 (20)

- ranges of change in uncertain variables ξ ξ ξ− +≤ ≤ :

≤ ≤env,118 21y vol.%, ≤ ≤env,278 80y vol.%,

≤ ≤env,31 2y vol.%, ≤ ≤env293 303T K,

× ≤ ≤ ×5 5env0.75 10 1 10P Pа (21)

Here, η= out outst { , , }z y G is the vector of output vari-

ables in the CSS regime, which is determined by solving Eqs. (1)-(12) of the mathematical model before the onset of the CSS regime (13); ω −i weight coefficients satisfying the conditions of ω ≥ 0i ,

ω∈ = 1ii J . Since it is not known with what proba-bility the uncertain parameters can take some values from the given ranges ξ ξ ξ− +≤ ≤ , it is assumed that they will be distributed in accordance with the equi-probable law. Then the coefficients ωi will be the same for all approximation points ξ ∈ 1,i i J , i.e., ω =

11/i JK , = ∈

1 11, ,Ji K i J . Changes in the timing of the opening of inlet

λλ λψ ∈1 ads( ), [0, ]t t t and outlet λ

λ λψ ∈2 ads c( ), [ , ]t t t t


96

valves were approximated by step-functions with a number of steps equal to 20.

Taking into account the constraint on the max-imum air velocity gv in the frontal layer of the ads-orbent ensures the protection of granulated adsorbent from destruction due to aerodynamic impact with the full jump-opening of control valves, which helps to increase the service life of the adsorbent.

In accordance with the calculations (Figure 4), as well as the data from literature [18,24], it can be concluded that the oscillating shifts of granules relat-ive to each other, causing the adsorbent abrasion during the operation of the PSA unit, arise at speeds of ∼0.15-0.25 m/s, which are 2-2.5 times less than the rate of fluidization of the layer. Calculation of the fluid-ization rate νf was carried out by the Todes formula [49], which allows to calculate with high accuracy the critical velocity of gas flow at homogeneous fluidiz-ation of the particle layer (the formula is given in the nomenclature).

Figure 4. Dependence of fluidization rate ν f and flow rate in the

adsorbent frontal layer ν g (0) on the pressure in the adsorber during the adsorption step in

adsP at the adsorbent particle diameter dgr: 1– 0.5 mm; 2– 1 mm; 3– 1.5 mm (∆ – experimental

data, full line – calculation by the model).

The formulated optimization problem, Eqs. (14)- –(21), belongs to the class of nonlinear programming problems, the solution of which was carried out by the method of sequential quadratic programming in the MatLab software environment (solver fmincon) [50] with the use of the optimized black-box approach [19,21,35-37] and the developed heuristic iterative algorithm.

The global maximum in the problem (14)-(21) can be guaranteed if the goal function and constraints

( )( )ξ =st , , , 1,5jg z d u j are convex functions with respect to the optimization variables, which is a necessary and sufficient Kuhn-Tucker condition [50].

It should be mentioned that the increase in pres-sure at the compressor outlet Pin simultaneously con-tributes to the increase in oxygen recovery (goal func-

tion) and energy consumption of the PSA unit. There-fore, it is expedient to search for the minimum value of the pressure Pin, at which: all the constraints of the problem ( )( )ξ =st , , , 1,5jg z d u j would be fulfilled; the maximum degree of oxygen recovery and the minimum energy consumption would be provided. Therefore, it is proposed to determine the pressure at the compressor outlet Pin by an iterative procedure.

Algorithm of solving the optimization problem, Eqs. (14)–(21)

Step 1. Set the initial numbers of iterations k = 1, ν = 1 and the initial value of pressure = ×in( ) 52 10kP Pa at the compressor outlet.

Step 2. Set the approximation points ξ ξ∈ ∈1 1, ,i ii J S , the initial set of critical points

ν νξ ξ− −= ∈ Ξ ∈( 1) ( 1)2 2{ : , }l lS l J and the initial approx-

imations of the regime variables (0)u . The initial set of critical points (0)

2S will be formed from the angular points ξ − , ξ + of uncertainty area Ξ in the assumption of convexity of functions describing the constraints ( )( )ξ =st , , , 1,5jg z d u j .

Step 3. Using the sequential quadratic program-ming method, we find a solution to the auxiliary problem:

ν ω η ξ∈

= 1

( , )( ) max ( , , )k iiu

i J

I u d u (А)

at links (1)-(12), ranges of change of optimized (20) and uncertain (21) variables, constraints (15)-(19), which are calculated in approximation points ξ ∈ ∈1 1,i S i J and critical points ξ ∈ ∈2 2,l S l J , and we define values of ν( , )( )kI u and of vector

ν ν ν ν λ ν λ νθ ψ ψ λ= =( , ) ( ) in( ) ( ) ( ) ( )ads 1 2{ , , , , , 1,20}k k, k, k, k, k,u t P .

If the solution of the auxiliary problem (A) is obtained, i.e.,

ν ν ν ν λ ν λ νθ ψ ψ λ= =( , ) ( ) in( ) ( ) ( ) ( )ads 1 2{ , , , , , 1,20}k k, k, k, k, k,u t P ,

we move to the next step. If the solution of the auxiliary task (A) has not

been obtained, i.e., at least one constraint ( )( )ξ ≤st , , 0i

jg z d u , ( )( )ξ ≤st , , 0ljg z d u , = 1,5j ,

∈ ∈1 1ξ ,i S i J , ξ ∈ ∈2 2,l S l J has not been fulfilled, then we check the fulfillment of the conditions:

1) if < ×in( ) 56 10kP Pa, we accept that = + Δin( ) in(k)kP P P ( ΔP =0.1×105 Pа), k = k+1 and

pass to step 3; 2) if = ×in( ) 56 10kP Pa, the algorithm finishes its

work and the solution of the problem (14)-(21) cannot be obtained for the specified requirements to the unit

out1,defy , out

defG , ν +g .

Step 4. To determine the new critical points where the constraints (15)-(19) ( )( )ν ξ ≤( , )

st , , 0kjg z d u ,


97

= 1,5j , ξ ∈ Ξ are violated, we solve five extremum problems:

( )( )ν

ξξ( , )

stmax , ,kjg z d u , ξ ξ ξ− +≤ ≤ , = 1,5j

and define five points νξ ( )q , = 1,5q which deliver the

maximum of functions ( )( )ν ξ =( , )st , , , 1,5k

jg z d u j , respectively. In case the constraint functions

( )( )ν ξ =( , )st , , , 1,5k

jg z d u j are non-convex with res-pect of uncertain parameters ξ, it is necessary to use global search methods, for example, the Global-Search in MatLab method [50], to solve extremum problems.

Step 5. Check the fulfillment of j×q constraints (15)-(19) ( )( )ν νξ ≤( , ) ( )

st , , 0kj qg z d u , = 1,5j , = 1,5q ,

and form a new set of critical points: ν ν ν−= ( ) ( 1) ( )2 2S S R ,

( )( ){ }ν ν ν νξ= > = =( ) ( ) ( , ) ( )stξ : , , 0, 1,5, 1,5k

q j qR g z d u j q . If the set ν( )R is empty, then the solution of the

problem at ν-th iteration is obtained ν∗ = ( , )ku u and the algorithm finishes its work; otherwise, we accept ν ν= +1 and go to step 3.

If the solution of the auxiliary problem (A) at step 3 of the algorithm was not obtained, the following variants are possible:

1) to “soften” the requirements to the unit, i.e., to reduce the required oxygen purity out

1,defy or unit pro-ductivity out

defG , to increase the value of the velocity constraint

+νg in the frontal layer of the adsorbent,

and to solve the problem (14)-(21) again with the help of the given algorithm;

2) to reduce the range of changes in the values of uncertain parameters ξ ξ ξ− +≤ ≤ and solve the problem (14)-(21) anew, using the above algorithm.

It should be noted that the number of iterations necessary to obtain the optimal solution of the prob-lem (14)-(21) using the algorithm described above, as a rule, does not exceed 2-3. A certain disadvantage of the developed algorithm is an increase in the number of critical points at each iteration and, accordingly, the number of constraints to be considered.

Results of the optimization problem solution under uncertainty for cyclic processes of atmospheric air adsorption separation

The optimization problem (14)-(21) was solved for the different values of oxygen purity out

1,defy (15) and productivity of the PSA unit out

defG (16) taking into account the constraint (17) on the gas flow velocity ν g and without the constraint (17) for the following cases:

1) outdefG = 2 l/min, out

1,defy = 40, 50, 60, 70, 80, 90 vol.% without the constraint (17) on velocity ν g (Fig-ure 5a and b) and taking into account the constraint (17) on velocity ν g (Figure 6a and b);

2) out1,defy = 90 vol.%, out

defG = 0.5, 1, 1.5, 2 l/min without constraint (17) on velocity ν g (Figure 5c and

Figure 5. Results of the optimization problem (14)–(21) solution without constraint (17) on the velocity of gas flow gv : at different values

of product oxygen purity out1,defy and out

defG = 2 l/min (a, b); and at different values of unit productivity outdefG and out

1,defy = 90 vol.% (c, d).


98

d) and taking into account constraint (17) on velocity ν g (Figure 6c and d);

3) out1,defy = 90 vol.%, out

defG =2 l/min, ν g = 0.15, 0.2, 0.3, 0.4, 0.5 m/s (Figure 7).

The initial data for the solution of the problem (14)-(21) are given in Table 4.

Tables 5 and 6 show the obtained programs of time changes in valve opening and CPU time of the optimization problem solution.

Analysis of graphs in Figure 8 testifies to the fact that at full jump opening of the inlet valve (Figure 8a, curve 1) the flow rate in the frontal layer of the ads-orbent (Figure 8b, curve 1) exceeds the rate of fluid-ization (Figure 4, curve 3). Oscillating motion of the particles relative to each other and abrasion of the

adsorbent (ν g = 0.65> fv ≈ 0.4 m/s) is observed. Opti-mal programs of “soft” stepwise opening of valves allow to provide the velocity in frontal adsorbent layer, not exceeding maximum admissible value of ν g .

The comparative analysis of the obtained opti-mal regimes without the constraint (17) on the gas flow velocity ν g (Figure 5a) and taking into account the constraint (17) (Figure 6a) allows to draw a con-clusion that the fulfillment of the requirements on resource saving of adsorbent leads to the increase of energy consumption (pressure inP increases), but at the same time the increased degree of oxygen rec-overy η is achieved. For example, at out

1,defy = 90 vol.% the account of constraint (17) provides the higher rec-overy rate η on ∼17 %, and the pressure increases

Figure 6. Results of the optimization problem (14)–(21) solution under constraint (17) on the velocity of gas flow +

gv = 0.2 m/s at different values of product oxygen purity out

1,defy and outdefG = 2 l/min (a, b); and at different values of unit productivity out

defG and out1,defy = 90 vol.% (c, d).

Figure 7. Results of the optimization problem (14)–(21) solution at different values of +

gv , out1,defy = 90 vol.% и out

defG = 2 l/min.


99

Table 4. Initial data for solving the optimization problem

Initial data Value

Composition of gas-air mixture, k 1- O2, 2- N2, 3- Ar

Zeolite adsorbent 13X

Isotherm parameters

Maximum adsorption volume W0, cm3/g 0.17

Parameter B of Dubinin-Radushkevich equation, ×10-6 1/К2 6.55

Affinity ratio, σ1 1.1

σ2 1

σ3 1.1

Design parameters

Number of adsorbers in the PSA unit 2

Internal diameter of adsorber DA, m 0.04

Height (length) of adsorbent layer L, m 0.2

Diameter of adsorbent granules dgr, mm 1.6

Valve throughput capacity Kv, l/min 15

Volume of product storage-tank VS , l 2

Set values of constraints

Oxygen concentration out1,defy , vol. % 40, 50, 60, 70, 80, 90

Productivity of the unit outdefG , l/min at NTP 0.5, 1, 1.5, 2

Velocity at the frontal layer of the adsorbent ν+g , m/s 0.08, 0.15, 0.2, 0.3, 0.4, 0.5

Regime variable

Duration of adsorption step (half-cycle) adst , s 0.5–30

Pressure at the compressor outlet inP , 510× Pa 2–6

Backflow coefficient θ , rel. units 0–6

Degree of valve opening λψ j , rel. units 0–1

Uncertain variable

Concentrations of components in the initial mixture, = −in 2env 4y y b ac , vol. %

oxygen in1y 18–21

nitrogen in2y 78–80

argon and impurities in3y 1–2

Temperature of the initial mixture =ing envT T , К 293–303

Pressure at the unit outlet =indes envP P , × 510 Pa 0.75–1

Table 5. Valve opening programs obtained as a result of solving the optimization problem (14)–(21) without constraint on ν+g in (17)

Requirements ψ1 ( )t , rel. unit ψ 2 ( )t , rel. unit CPU time,min out1,defy =40 vol. %, out

defG =2 l/min 1 1 145 out1,defy =50 vol. %, out





defG =2 l/min 1 1 158 outdefG =0.5 l/min, out

1,defy =90 vol. % 1 1 115 outdefG =1.0 l/min, out

1,defy =90 vol. % 1 1 142 outdefG =1.5 l/min, out

1,defy =90 vol. % 1 1 141

Table 6. Valve opening programs obtained as a result of solving the optimization problem (14)–(21) with constraint on ν+g = 0.2 m/s


defG =2 l/min 0.26, 0.83, 1 0.45, 1 318 out1,defy =50 vol. %, out

defG =2 l/min 0.24, 0.71, 1 0.41, 1 327


100

Table 6. Continued


defG =2 l/min 0.21, 0.54, 1 0.35, 0.95, 1 331 out1,defy =70 vol. %, out

defG =2 l/min 0.19, 0.47, 1 0.32, 0.91, 1 352 out1,defy =80 vol. %, out

defG =2 l/min 0.17, 0.39, 0.95, 1 0.29, 0.88, 1 345 out1,defy =90 vol. %, out

defG =2 l/min 0.11, 0.19,0.33, 0.56, 1 0.25, 0. 34, 0.72, 0.97, 1 404 outdefG =0.5 l/min, out

1,defy =90 vol. % 0.13, 0.27, 0.56, 1 0.29, 0.51, 0.91, 1 377 outdefG =1.0 l/min, out

1,defy =90 vol. % 0.13, 0.26, 0.52, 1 0.28, 0.51, 0.89, 1 372 outdefG =1.5 l/min, out

1,defy =90 vol. % 0.11, 0.22, 0.42, 0.81, 1 0.25, 0.41, 0.82, 1 388

Figure 8. Programs (degrees) of opening of inlet valves (a) and velocity of gas-air flow in the frontal layer of the adsorbent (b): 1 – no constraint (17) on the maximum permissible velocity ν+

g ; and under constraint (17): 2– ν+g = 0.15 m/s; 3– ν+

g = 0.2 m/s; 4– ν +g = 0.3 m/s.

inP ≈1×105 Pa. Accounting for the constraint (17) on the velocity of the gas flow gv leads to an increase in the duration of the absorption and desorption steps, which reduces the number of operations of the control solenoid valves of the unit and, accordingly, prolongs their service life. It is especially evident at out

1,defy = 90 vol.%, when the difference in duration of tads of the adsorption step is 40% (Figures 5a and b).

The analysis of graphs in Figures 5c–d and Figures 6c–d shows that with the increase in the unit productivity out

defG , there is also an increase in the oxygen recovery rate of η. This regularity can be exp-lained by the increase of the inlet gas flow rate due to the increase of the compressor outlet pressure inP and the decrease of the flow tapped for adsorbent regeneration. For example, at the increase of out

defG by 4 times (from 0.5 to 2 l/min) a significant increase of oxygen recovery rate η (taking into account the cons-traint (17) - by 3 times (Figure 6c), without taking into account the constraint (17) - by 4 times (Figure5c)) is observed. This is several times higher than the inc-rease in energy consumption of the compressor due to pressure rise inP (taking into account the constraint (17) - by 1.3 times (Figure 6d), without taking into account the constraint (17) - by 1.1 times (Figure 5d)).

Decrease in maximum admissible value of velo-city +

gv from 0.5 to 0.2 m/s allows to raise oxygen recovery rate η on average by 15%, simultaneously increasing pressure inP at the compressor outlet by 12% (Figure 9), and decrease of +

gv from 0.2 to 0.15

m/s is inexpedient, as at pressure inP increase by

33% the oxygen recovery rate η grows only by ∼20% (Figure 9).

Figure 9. Dependence of PSA unit productivity outG on the set

values of product purity out1,defy and out

defG = 2 l/min: 1 – without constraint (17), 2 – under constraint (17).

Comparative analysis of the graphs in Figure 9 shows that, at the required purity of oxygen out

1,defy in the range from 40 to 70 vol.%, the productivity of the PSA unit can be increased by 13% when the cons-traint (17) on the velocity gv of gas flow is taken into account.

CONCLUSION

The problem of optimization of regime variables of PSA process of air separation and oxygen concen-


101

tration under interval uncertainty of initial composition of atmospheric air, temperature and ambient pressure was formulated and solved. A heuristic iterative algo-rithm for solving the optimization problem using the criterion of the oxygen recovery rate and taking into account operating procedure requirements on oxygen purity, unit productivity and gas flow velocity in the frontal layer of the adsorbent was developed. Gas flow velocity constraint in the adsorbent layer was provided by a stepwise change in the degree of open-ing of inlet and outlet valves in time, which made it possible to increase the adsorbent service life along with the achievement of the maximum degree of oxy-gen recovery and purity.

Statement of the optimization problem under partial uncertainty of the initial data and the dev-eloped algorithm of its solution can be used in the modernization of existing and design of new res-ource-saving PSA units for air oxygenation and other adsorption plants of separation and purification of multi-component gas mixtures, in which it is neces-sary to use expensive or unique granular adsorbents (silver-containing types of zeolites, sodium forms, per-ovskite adsorbents, etc.).

Funding. This work was supported by the Min-istry of Education and Science of the Russian Feder-ation (grant number 10.3533.2017).

Nomenclature

English symbols А1, А2 – adsorbers Ar – Archimedes number a – concentration in the adsorbent, mole/m3

a* – equilibrium adsorption value, mole/m3 B – parameter of the Dubinin thermal equation, К-2 b – molar volume constant, l/mole С – compressor

apс – specific heat capacity of adsorbent, J/(kg×K) pgс – specific heat capacity of gas-air mixture, J/(mol K)

AD – internal diameter of the adsorber, m gD – diffusion coefficient in gas phase, m2/s

d – vector of design parameters grd – diameter of adsorbent granules, m

E – characteristic adsorption energy, J/mole G – flow rate, l/min

outG – unit productivity, l/min out

1G – gas-air mixture flow rate at the unit outlet, l/min g – constant of free fall acceleration, m/s2 h – sorption heat, J/mole K – kinetic coefficient of the rate of pressure rise and fall in the adsorber, s-1

vK – throughput capacity of inlet and outlet valves, l/min

L – height (length) of the adsorbent layer, m M – mathematical expectation

gM – molar mass of gas-air mixture, kg/mole NTP – normal temperature and pressure n – number of adsorption-desorption cycles P – pressure, Pa

inP – pressure at the compressor outlet, Pa out

1P – pressure at the unit outlet, Pа Pr – Prandtl number Ps – saturation pressure, Pa PSА – pressure swing adsorption R – universal gas constant, J/(mole×К) Re – Reynolds number Recr – critical Reynolds number S – adsorber cross-section area, m2

spS – area of specific surface of pores (meso- and macropores) of the adsorbent, m2/m3 STС – compressor storage-tank STP – product storage-tank T – temperature, К

= inenv gT T – temperature of the initial gas-air mixture at

the unit inlet, К out

1T – temperature of the gas-air mixture at the unit outlet, К t – time, s

= =ads des c / 2t t t – half-cycle time (adsorption, desorp-tion), s u – vector of regime variables V – volume, m3

*V – molar volume, sm3/mmole v1, v2, v3, v4 – control valves v6 – throttle v5, v7 – backflow valves v8 – manual valve W0 – adsorption capacity, cm3/g x – spatial coordinate, m y – concentration in gas phase, mole/m3

= inenvy y – composition of atmospheric air, vol.% inky – inlet concentration of k-th component in gas-air

mixture, vol.% outky – outlet concentration of k-th component in gas-

air mixture, vol.% out,1y – concentration of gas-air mixture components

at the outlet of the unit, vol.%

Greek symbols α – heat transfer coefficient from adsorbent to gas-air mixture, Wt/(K×m2) ρ – mass transfer coefficient attributable to the concentration of adsorbate, m/s ε – adsorbent porosity coefficient, m3/m3 φ – thermal coefficient of limiting adsorption, rel. unit η – recovery rate, rel. unit


102

θ – backflow coefficient; defines the amount of outlet flow returning for adsorbent regeneration, rel. unit λa – heat transfer coefficient of adsorbent, Wt/(m×K)

gλ – heat transfer coefficient of gas mixture,

Wt/(m×K) μ g – dynamic viscosity of gas-air mixture, Pa×s ν – velocity of gas-air mixture, m/s νf – fluidization rate, m/s χ – latent condensation heat for volumetric liquid phase, J/mole ξ – vector of uncertain variables ρ – density, kg/m3

ρa – adsorbent density, kg/m3 ρ g

– molar density of gas, mole/m3 ρ * – molar density, g/sm3 ρ*

cr – adsorbate density at critical temperature, g/sm3 σ – affinity ratio ς – sphericity coefficient of adsorbent granules τ – index of the Dubinin thermal equation υ * – ratio of the components of atomic diffusion volumes ψ – degree of opening of inlet and outlet valves, rel. unit ω – weight coefficient

Subscripts + – maximum admissible value of a parameter – – minimum admissible value of a parameter A – adsorber a – adsorbent ads – adsorption step b – boiling с – cycle cr – critical C – compressor def – given des – desorption step e – experimental value env – environment, ambient g – gas phase i – countable index in – inlet j – countable index k – number of a component of the gas-air mixture nc – under normal conditions out – outlet P – product S – product storage-tank st – steady state v –valve λ – step number for a valve opening step-function

REFERENCES

[1] E. Shokroo, D. Farsani, H. Meymandi, N. Yadoliahi, Korean J. Chem. Eng. 33 (4) (2016) 1391–1401

[2] S. Effendy, C. Xu, S. Farooq, Ind. Eng. Chem. Res. 56 (18) (2017) 5417–5431

[3] Q. Fu, H.Y. Yan, Y.H. Shen, Y.J. Qin, D.H. Zhang, Z. Zhou, J. Cleaner Prod. 170 (2018) 704–714

[4] D. Ko, R. Siriwardane, L. Biegler, Ind. Eng. Chem. Res, 44 (21) (2005) 8084-8094

[5] T.M. Hu, H.Y. Zhou, H. Peng, H.Q. Jiang, Front. Chem. (Lausanne, Switz.) 6 (2018) 329

[6] S. Cavenati, C.A. Grande, A.E. Rodrigues, Chem. Eng. Sci. 61 (12) (2006) 3893–3906

[7] M.T. Ho, G.W. Allinson, D.E. Wiley, Ind. Eng. Chem. Res. 47 (14) (2008) 4883–4890

[8] J.H. Park, S.S. Han, J.N. Kim, S.H. Cho, Korean J. Chem. Eng. 21 (2004) 236–245

[9] S.U. Rege, R.T. Yang, Chem. Eng. Sci. 57 (7) (2002) 1139–1149

[10] H. Wang, M. Goto, M. Sasaki, T. Hirose, Ind. Eng. Chem. Res. 43 (11) (2004) 2753–2758

[11] S.I. Yang, D.Y. Choi, S.C. Jang, S.H. Kim, D.K. Choi, Adsorption 14 (2008) 583–590

[12] P. Brea, J.A. Delgado, V.I. Agueda, M.A. Uguina, Chem. Eng. Sci. 355 (2019) 279–289

[13] D.M. Ruthven, S. Farooq, K.S. Knaebel, Pressure swing adsorption, New York, 1994, p. 352

[14] N.V. Keltsev, Basics of adsorption technology, Moscow, 1984, p. 592

[15] E.S. Kikkinides, V.I. Sikavitsas, R.T. Yang, Ind. Eng. Chem. Res., 34 (1) (1995) 255–262

[16] A. Malek, S. Farooq, AIChE J. 44 (9) (1998) 1985–1992

[17] J.A. Ritter, R.T. Yang, Ind. Eng. Chem. Res. 30 (5) (1991) 1023–1032

[18] L.I. Kheifets, V.L. Zelenko, Y.V. Pavlov, Elements of the theory of processes adsorption separation gas mixture, Moscow, 2004, p. 69

[19] S. Hosseinzadeh Hejazi, L.E. Perez, A. Rajendran, S.M. Kuznicki, Ind. Eng. Chem. Res. 56 (19) (2017) 5679–5691

[20] J.Y. Rumbo-Morales, G. López-López, V.M. Alvarado- -Martínez.; J.S. Valdez-Martínez, F.J. Sorcia-Vázquez, J.A. Brizuela-Mendoza, Rev. Mex. Ing. Quim. 17 (2018) 1051-1081

[21] J.C. Santos, P. Cruz, T. Regala, F. D. Magalhaes, A. Mendes, Ind. Eng. Chem. Res. 46 (2) (2007) 591–599

[22] J.C. Santos, A.F. Portugal, F.D. Magalhaes, A. Mendes, Ind. Eng. Chem. Res. 43 (2004) 8328–8338

[23] J.C. Santos, A.F. Portugal, F.D. Magalhaes, A. Mendes, Ind. Eng. Chem. Res. 45 (2006) 1085–1096

[24] A.K. Akulov, Simulation of separation of binary gas mix-tures by pressure swing adsorption, Ph. D. Saint-Peters-burg, St PbSIT (Tech. Univ.), 1996, p. 304

[25] A. Moran, O. Talu, Ind. Eng. Chem. Res. 57 (35) (2018) 11981–11987

[26] J.H. Li, Biomed. Mater. Eng. 24 (5) (2014) 1771–1779


103

[27] C. Wu, R. Vemula, M. Kothare, S. Sircar, Ind. Eng. Chem. Res. 55(16) (2016) 4676–4681

[28] V.R. Rao, S. Farooq, W.B. Krantz, AIChE J. 56 (2) (2010) 354–370

[29] S. Kulish, R.P. Swank, (Impact MST, Incorporation), Pat. USA 5827358 (1998)

[30] M.W. Ackley, A.B. Stewart, G.W. Henzler, F.W. Leavitt, F. Notaro, M.S. Kane, (Praxair Technology, Inc.), Pat. USA 6027548 (2000)

[31] S. Kulish, R.P. Swank, (Impact MST, Incorporation), Pat. USA 6068680 (2000)

[32] S. Sircar, Gas Separation by Rapid Pressure Swing Ads-orption, Pat. USA 5071449 (1991)

[33] J. Labasque, S. Moreau, C. Marot (L'Air Liquide, Societe Anonyme pour l'Etude et l'Exploitation des), Pat. USA 6068678 (2000)

[34] L. Jiang, L.T. Biegler, V.G. Fox, AIChE J. 49 (5) (2003) 1140–1157

[35] P. Cruz, F.D. Magalhaes, A. Mendes, AIChE J. 51 (5) (2005) 1377–1395

[36] Y. Liu, F. Sun, Exp. Sys. Applic. 40 (11) (2013) 4496– -4502

[37] Z.Y. Ding, Z.Y. Han, Q. Fu, Y.H. Shen, C.X. Tian, D.H. Zhang, Adsorption 24 (5) 499–516

[38] F. Capra, M. Gazzani, L. Joss, M. Mazzotti, E. Martelli, Ind. Eng. Chem. Res. 57 (30) (2018) 9977–9993

[39] S.J. Lee, J.H., Jung, J.H. Moon, J.G. Jee, C.H. Lee, Ind. Eng. Chem. Res. 46 (11) (2007) 3720–3728

[40] A.K. Beeyani, K. Singh, R.K. Vyas, Sh. Kumar, Su. Kumar, Pol. J. Chem. Technol 12 (2) (2010) 18–28

[41] S. Farooq, D.M. Ruthven, H.A. Boniface, Chem. Eng. Sci. 44 (12) (1989) 2809–2816

[42] R. Rota, P.C. Wankat, AIChE J. 36 (9) (1990) 1299–1312

[43] M. Xu, H.C Wu, Y.S. Lin, S.G. Deng, Chem. Eng. Sci. 354 (2018) 62–74

[44] P. Cruz, J. C. Santos, F. D. Magalhaes, A. Mendes, Chem. Eng. Sci. 58 (2003) 3143–3158

[45] V.А. Kirillin, V.V. Sychev, А.Е. Sheindlin, Technical thermodynamics, Moscow, 2016, p. 496

[46] S.W. Chai, M.V. Kothare, S. Sircar, Ind. Eng. Chem. Res. 50(14) (2011) 8703–8710

[47] M.M. Dubinin, Adsorption and porosity, Moscow, 1972, p. 128

[48] R.G. Rice, D.D. Do, Applied Mathematics and Modeling for Chemical Engineers, 2 ed., Wiley, Hoboken, NJ, 2012, p. 397

[49] M.E. Aerov, O.M. Todes, D.A. Narinkiy, Devices with a stationary granular layer. Hydraulic and thermal bases of work, Saint-Petersburg, 1979, p. 176

[50] A.L. Goldstein, Optimization in MATLAB, Perm’, 2015, p. 192

[51] R. Reid, J. Prausnitz, T. Sherwood, Properties of gases and liquids, New York, 1977, p. 592

[52] E.I. Akulinin, O.O. Golubyatnikov, S.A. Skvortsov, D.S. Dvoretsky, S.I. Dvoretsky, International Multidisciplinary Scientific GeoConference Surveying Geology and Mining Ecology Management, SGEM, 18(4.2) (2018), pp. 539–546

[53] E.I. Akulinin, O.O. Golubyatnikov, S.I. Dvoretsky, D.S. Dvoretsky, International Multidisciplinary Scientific GeoConference Surveying Geology and Mining Ecology Management, SGEM, 18(4.2) (2018), pp. 599–606

[54] A.G. Kasatkin, Basic processes and apparatus of chem-ical technology, Moscow, 2004, p. 753

[55] S.A. Skvortsov, E.I. Akulinin, O.O. Golubyatnikov, D.S. Dvoretsky, S.I. Dvoretsky, Journal of Physics: Confe-rence Series, 1015(3) (2018) 032002

[56] Y.I. Sumyatsky, Industrial adsorption processes, Moscow, 2009, p. 182

[57] D.N. Zhedyaevskiy, V.D. Kosmin, V.A. Lukyanov, S.S. Kruglov, Hydro-mechanical processes. Guidelines for practical training in the laboratory of processes and apparatus of oil and gas processing, Moscow, 2012, p. 76.

APPENDIX

( )υ

−

=

∏

3g g, g,

g 21/3*

10 /k kk k

kk

T M MD

P, calculated by the Fuller, Schettler, and Gidding's method [51];

*1υ =16.6,

*2υ =17.9,

*3υ =16.1 [51]; k = 1,2,3: 1 – oxygen, 2 – nitrogen, 3 – argon and impurities;

τ τϕχτ

− = + +

1 1 1* *

aads, ln lnk k

k kk k

Ta ah Ea a

[47]; τ τϕχ

τ

− = + + +

1 1 1* *

ades, 1 ln lnk k

k kk k

Ta ah Ea a

[47];

= × 14.187 4.754E

B [47]; χ1=6824.5 J/mole, χ2=5581 J/mole, χ3=5694 J/mole [51]; τ=2 [47];


104

μν

ρ= cr g

fg gr

Red

[49]; ( )( )ε ε ε ε

=− + ×

cr3 3 3

ArRe

150 1 / 1.75 / Ar / 1.75 [49];

ρμ

=3gr g

2g

gAr

d [49];

ρρ = g,nc g,nc

gg nc

T PT P

;g = 9.81

m/s2;

( ) = − −

Sg

exp / 760kk

k

FP AT C

, Ak, Fk, Ck – constants of the Antoine equation for the k-th component of gas-air

mixture [51];

α =0,83

g

gr

0.24Re λd

[49]; βε

⋅=

0.64 1/3g7.9 Re Pr

4kD

[54]; μ

ρ= g

g g

PrD

[54]; ρ

μ ε ε−= g eq g

g

2Re

3 (1 )w d

; ε=eqsp

4S

d – equivalent

diameter of adsorbent pore ducts, m [54]; =sp gr6 /S d [49];

ρ= g,*

*k

kk

MV [47];

ϕρ ρ − −= g b,0.434 ( )*b, 10 k kT T

k k [47]; ( )

ρρ

ϕ

=

−

b,*cr,

cr, b,

lg

0.434

k

kk

k kT T [47]; ρ =*

cr, 1000k

kk

Mb [47];

= cr,

cr,

18

k

k

RTb

P [47];

θ = ads des

des ads

( ) ( )( ) ( )

P t G tP t G t

, rel. unit, θ≤ ≤ ads

des

( )0

( )P tP t

[56];

B = 6.55×10-6 K-2 [14, 55]; apс = 1000 J/(kg K); pgс = 830 J/(mol K); =ads 1K s-1, =des 4K s-1 [52, 53]; W0 = 0.17

sm3/g [14, 55]; ε = 0.39 [14, 55]; λ =a 0.139 Wt/(m K) [14];

λg =0.0259 Wt/(m K) [49]; μ g =1.81×10-5 Pa s [49]; ρa =2140 kg/m3; σ1 = 1.1, σ 2 = 1, σ 3 = 1.1 [14]; ς =1

for round granules [57]; λψ≤ ≤0 1j .

EVGENY AKULININ

OLEG GOLUBYATNIKOV DMITRY DVORETSKY

STANISLAV DVORETSKY

Tambov State Technical University, Tambov, Russia

NAUČNI RAD

OPTIMIZACIJA I ANALIZA PROCESA ADSORPCIJE SA PROMENLJIVIM PRITISKOM ZA PROIZVODNJU KISEONIKA IZ VAZDUHA

Postrojenja za adsorpciju sa promenljivim pritiskom (PSA) se široko koriste za razdva-janje atmosferskog vazduha i koncentrisanje kiseonika. Međutim, efikasnost takvih pos-trojenja je smanjena zbog slučajnih promena u karakteristikama atmosferskog vazduha koji se razdvaja. U ovom radu je formulisan i rešen problem optimizacije režima rada PSA postrojenja sa zeolitnim adsorbentom 13X prema kriterijumu brzine izdvajanja kiseonika u uslovima periodične nesigurnosti sastava, temperature i pritiska atmosfer-skog vazduha. Problem optimizacije uzima u obzir, takođe, zahtev za čistoćom kiseo-nika, produktivnošću postrojenja i uštedom granulisanog adsorbenta zbog abrazije gra-nula. Predlaže se obezbeđivanje uštede adsorbenta ograničavanjem ulaznog protoka u frontalnom sloju adsorbenta pomoću „meke“ stepenaste promene otvaranja kontrolnih ulaznih i izlaznih ventila postrojenja. Problem (uključujući traženje programa za pro-menu vremena za stepen otvaranja upravljačkih ventila) rešen je primenom razvijenog matematičkog modela cikličnih procesa toplotne i masene razmene adsorpcije-desorp-cije u PSA postrojenja i heurističkog iterativnog algoritma. Izvršena je uporedna analiza rezultata rešenja problema optimizacije sa i bez uzimanja u obzir ograničenja brzine protoka gasa u frontalnom sloju adsorbenta. Izučavan je uticaj određenih zahteva na performanse PSA postrojenja i čistoće kiseonika na stepen njegovog izdvajanja.

Ključne reči: adsorpcija sa promenljivim pritiskom, zeolit, matematičko modelo-vanje, optimizacija, numerička simulacija, nesigurnost.

Documents

ISSN 1451 - 9372(Print) ISSN 2217 - 7434(Online ... No1_p1-104...čevi, Novi Sad, Serbia (45 20’ N, 19 51’ E). The content of water in grain was determined according to the method