Acps11th CD-hvbf Moisture Control

8/9/2019 Acps11th CD-hvbf Moisture Control

1/12

82

Paper C1 Horizontal Vacuum Belt Filter Control Using On-line Moisture Analysis at Gregory Coal Mine

paper C1

Horizontal Vacuum Belt Filter Control

Using On-line Moisture Analysis atGregory Coal MineG France, A von MuraltCallidan Instruments

Abstract

The paper summarises the recent ACARP project C14061 where a microwave moisture analyserwas tted across a horizontal vacuum belt lter at the Gregory coal preparation plant. The hori-zontal vacuum belt lter dries ne coal slurry which is typically 20% solids to a cake o 20-40%moisture. On-line analysis on the belt lter gives the opportunity or eed back control andminimal delay time.

The purpose o this project was to gain a better understanding o belt lter control by examin-ing the moisture on-line. The integration o a moisture result into the control loop may providea cake o more consistent moisture as the drying time can be adjusted real-time to maintain adesired outlet moisture range.

Ater determining the accuracy o the analyser was 1.0% to 2% at 2 standard deviations(depending on the range) it was possible to monitor moisture online. It was ound that themoisture analyser proved very useul to optimise process parameters such as focculant addi-tion and belt speed or a desired lter cake moisture.

Introduction

Horizontal vacuum belt lters are a common dewatering technique or ne coal applications inAustralian coal mines. Through the use o these lters, material as coarse as spiral product or asne as roth fotation concentrates can be dewatered eectively.

Horizontal vacuum belt lters work by applying vacuum below a lter media. As the nameimplies the lter is horizontal in the zone where the vacuum is applied. The lter media isthereore fat and, or continuous separation to occur, the media must also be moved continu-ously. Generally the horizontal vacuum belt lter is designed to allow about one-third o the

conveyor length or drainage o ree moisture and the remainder or drawing air through thedrying zone. Whilst there are a number o controls to adjust the perormance o horizontalvacuum belt lters such as belt speed, cake depth, focculant addition, and vacuum control, tooptimise the process requires extensive laboratory sampling to observe a parameters eect


2/12

8

11th Australian Coal Preparation Conference and ExhibitionProceedings 2007

on the output moisture. Figure 1 below displays a cross sectional view o a typical horizontalvacuum belt lter, the slurry is deposited on top o the cloth on the let side, and dropping othe right side o the belt as dewatered product.

Figure1Cross-sectionofaHorizontalVacuumBeltFilter

This project examines the installation o an on-line instrument to measure moisture in real-time.I proven successul, this will allow operators and process engineers to accurately adjust processparameters to meet the desired product moisture. It would also eliminate the costly need tomanually sample the lter belts, which in many cases are perormed only every two hours.

In the early stages o this project Callidan Instruments undertook to develop a microwavebased moisture analyser which would predict the moisture content o the ne coal bed acrossthe conveyor belt. The analyser was designed to independently determine the percentagemoisture at three dierent locations across the conveyor belt. For each location a microwavetransmitter was located below the conveyor and an opposing microwave receiver locatedabove the conveyor and lter cake product. One o the major design concerns or this applica-tion was to develop the correct microwave operating parameters to cater or approximately 20-40 mm o lter cake which could vary in moisture rom 20% to 50%.

Project Objectives

The objectives o this project were to:

Design, develop and supply an on-line microwave moisture analyser with 3 measurement headsacross a horizontal vacuum belt flter, and outputting three moisture results simultaneously.

Install commission and calibrate the analyser at the Gregory CHPP. Evaluate and determine the accuracy o the moisture result. Provide a report on the accuracy and eectiveness o the installed instrumentation and the

resultant improved eect upon belt lter control.

Identiy control strategies or belt lter control using an on-line moisture analyser.


3/12

8


Project Stages and Outcomes

Stage 1 Design and Development of an On-line Moisture Analyser for a Horizontal Vacuum

Belt Filter

The rst objective o this project was the design and development o a microwave moistureanalyser which would be capable o accurately predicting the moisture content at three inde-pendent points across the 5m span o the lter cloth. Whilst there are a number o techniquesavailable or the determination o moisture in material such as ne coal, it was decided to use amicrowave transmission technique. This technique passes a microwave signal rom below theconveyor abric, through the lter cake bed and is detected by the microwave receiver. Byusing statistical analysis tools (chiefy linear regression) on both the measured microwaveattenuation, phase shit, and the lter cake depth an algorithm was determined which couldpredict the total moisture content o the lter cake.

The analyser needed to be designed such that three independent readings o moisture couldbe obtained rom the analyser, hence three pair o sensors where equally placed across theconveyor. The analyser then needed to accept the lter cake depth signal rom the alreadyinstalled ultrasonic sensor. Typically ultrasonic depth sensors would be provided or eachmicrowave sensor, however due to nancial and complexity constrictions, the single ultrasonicwas used and the thickness across the belt was assumed to be uniorm.

In addition to the sensor heads and measurement techniques o the analyser, the analyserneeded to incorporate an integral processor, user display, and industry standard analog, digitaland serial outputs.

Over a period o approx 2 months the analyser design specication was established, thenecessary components ordered and the analyser assembled and tested with Callidan Instru-ments R&D acility in Mackay. Figure 2 and Figure 3 below shows the analyser being manuac-tured in Mackay and the sensor head.

Figure2ManufactureofAnalyser


4/12

8


Figure3SensorHead

Stage 2 Installation and Commissioning

The second objective o this project was the installation, commissioning and calibration o theanalyser. The analyser was installed at Gregory in June 2005. The installation o the analyserneeded to occur during a routine shutdown period o the Gregory coal handling and prepara-tion plant. The necessary saety requirements or both the mechanical and electrical installa-tion were considered. Mechanical and electrical installations were then conducted saely andwithout any diculties.

The unit consists o the conveyor rame which supports the 3 pairs o transmitters/sensors.Located within 5m is the control cabinet which is designed to operate rom a standard 110v or240v power supply.

The analyser automatically begins reporting moisture when the lter cake reaches a pre-deter-mined minimum depth which was ound to be 10mm. The results o the analyser can besupplied to the control room via analog, serial, or Ethernet, enabling moisture trends to bedisplayed or the benet o the operators. The analyser can supply an instantaneous (1 second)or a congurable rolling average moisture result at each sensor location or alternatively an

average o the 3 results.The microwave transmitters located beneath the cake emit RF energy at less than 5mW or+3dBm. For comparison a mobile phone emits between 50mW and 800mW o RF energy.

Figure 4 and Figure 5 show the analyser location prior to the analyser being installed and aterthe installation.


5/12

8


Figure4DischargeEndofConveyorPriortoInstallation

Figure5InstallationofSensorHeadsDuringShutdownPeriod

Stage 3 Moisture Analyser Calibration

The analyser was calibrated by sampling lter cake directly rom the discharge conveyor. A

sample o approx 5kg was collected next to each sensor over a period o approx 10 secondsand samples were collected 10 minutes apart. The collected samples were analysed or totalmoisture content. In total 60 samples were taken or the purpose o calibration, 20 samples persensor. The analyser moisture result was also averaged over the same 10 second period.

It is important to understand the analyser carries out the on-line determination o moisture bymeasuring the microwave response through the lter cake, however or this to occur indepen-dently rom the depth o lter cake the depth signal must be used. For this particular lter beltit was assumed that measuring the cake depth at one location (i.e. the only existing ultrasonicsensor in line with sensor 1) would be sucient or all three measurements as ultrasonics

could not be provided or each sensor.

Figure 6 below highlights the calibration perormance o the rst sensor.


6/12

87


Figure6Sensor1SamplingData

The results rom Sensor 1 indicated that the analyser could operate to an accuracy o betterthan 2.7% (2 std dev) moisture over a range o 25% 45%. Sensors 2 & 3 did not achieve thislevel o accuracy and are included below. However or all data sets the F statistic or the data is

greater than the critical value indicating that there is a relationship between analyser and labresults, and that this relationship did not occur by chance. For the F statistic calculation, theprobability o erroneously concluding a relationship or alpha value was set at 0.05.



7/12

88



The correlations and signicance o the Moistscan versus laboratory samples or sensors 2 and3 are not as good as those or sensor 1.

The calibration or sensor 2 yields a good correlation, assisted heavily by the 50% moisture sample.The error is quite high at 7.3% (2 std dev) this is due to a number o outliers at ~35% moisture.

The calibration or sensor 3 above yields a poor correlation, though a surprisingly small error o2.9%. The reason or this small error is the narrower range o sample moistures compared tothe calibration results o the other two sensors. There are 2 points that could be considered asoutliers whose removal would improve the statistics, these are at 24% and 31%.

The calibration data was certainly good or sensor 1 and poor or sensors 2 and 3. The reasonor this is likely the assumption o consistent bed depth across the lter. Given that the clothwas quite old and prone to blinding, this assumption is unlikely to be true or cases where thisis occurring. Unortunately due to time constraints, cost, system complication and limited

downtime, additional ultrasonics were not tted or the other 2 sensors.

Stage 4 Accuracy Determination

1 week ater calibration, this project attempted to determine what the actual accuracy o theanalyser is via the operation o a Grubbs estimator approach. The Grubbs estimator approachconsists o taking two independent samples rom the discharge end o the conveyor at sensor1 and then comparing these with the average Moistscan result or the same sensor over thesame time period. This Grubbs estimator sampling program was only conducted or Sensor 1due to the depth measurement being in line with this sensor and not the others.

The results o the sampling program indicated an Analyser error o approx 2.0% at 2 standarddeviations. Figure 9 below highlights a tracking plot o the analyser, sample 1 (Lab1), andsample 2 (Lab2).


8/12

89


Figure9TrendsofMoistureAnalyserandLaboratories

When applying the 3 data sets to the Grubbs Estimator program it can be determined that theanalyser error is approx 1.0% moisture at a 95% condence level whilst the two sampling tech-niques yield a slightly greater error (approx 1.1% and 1.2%). Figure 10 below shows the resul-tant Grubbs calculated instrument and laboratory techniques.

Figure10GrubbsEstimatorReportingofInstrumentandLaboratoryErrors

Stage 5 Investigation into Control Mechanisms and their Optimisation

The horizontal vacuum belt lter used or this trial unortunately had some serious maintenanceconcerns. The lter belt was scheduled or replacement in late September as the current beltwas ouling and partially blinded. In addition the spray bar nozzles at both the ront end o thebelt lter (prior to slurry addition) and mid way along the belt lter were partially inoperative.


9/12

90


The result o this poorly perorming belt lter was that a large variation in moisture was experi-enced across the width o the belt lter. Typically sensor 1 & 2 experienced extremely highmoisture variations and variable depth o product. Sensor 3 experienced much more stablecake depths which resulted in a much more consistent moisture reading. Figure 11 below

highlights the vastly diering moisture readings experienced over a 2 hour period.

Figure11TwoHourTrendofMoisturesandCakeDepth

What can be observed rom this data is the clear relationship the cake depth has with moisturecontent. Considering that the depth measurement only occurs near Sensor 1, it is interestingto observe the relationship between the two below in Figure 12.

Figure12AnalyserMoisturevsFilterCakeDepth


10/12

91


Whilst depth o lter cake is only a part o the complexity o controlling moisture it can beobserved rom Figure 12 that typical perormance plots or moisture control can be now easilyobtained using the calibrated and veried moisture analyser where previously this exerciserequired extensive sampling and analysis.

Ater the bed depth trial above, the lter cloth was replaced and no more data was collectedor 2-3 months. With the installation o the new lter cloth the lter cake proles appearedmuch more consistent and o much lower moisture content.

Figure 13 displays an example o the eect o belt speed on the on-line moisture result. Typi-cally the plant operators run the belt speed at 0%, however during this test the belt speed wasincrementally adjusted up to 50%. A speed o 0% simply represents the minimum belt speedo 30mm/s whilst 100% represents the belts maximum belt speed o 200mm/s or 60t/h. It canbe observed that as the belt speed approaches 35% o its maximum speed the moisturecontent begins to rise rapidly.

Whilst this graph clearly displays a relationship between belt speed and moisture it should beunderstood that the belt speed is not the only contributor to the change in lter cake mois-ture. It was observed during the test that as the belt speed increases the bed depth decreases,hence two eects are contributing together during the this test.

Figure13AnalyserMoisturevsBeltSpeed

The eect o focculant on on-line moisture was tested (Figure 14). Plant operators typicallyoperate the belt lter with an addition o 100% focculant, meaning the focculant pump isoperating at 100% o its capacity. Figure 14 indicates that focculant addition could be reduced

to 80% beore any appreciable change in moisture was observed on that occasion.


11/12

92


Figure14AnalyserMoisturevsFlocculantAddition

Typically operators run belt lters or good discharge and adjust the right bed thickness toachieve this. There is currently no easy way o knowing what the best approach or moisturecontrol should be. Traditional practices usually involve a manual sample being taken directlyrom the belt discharge every two hours. Not only is this sample indicative o perhaps a snap-shot o 2-3 seconds o belt lter operation the analysis results can take up to 4 hours to bereceived by the operator. By this time the operation o the belt lter can and usually doeschange considerably.

This project demonstrates the value o an instrument measuring belt lter moisture online sothat eective control strategies can be implemented. It is the authors belie that the accuracyachieved rom the analyser would clearly add benet to strategies or belt lter control. Havingan instrument reliably provide direct on-line inormation to an accuracy o better than 1.0%moisture would be a clear advantage over previous labour intensive sampling practices.

The control strategies or optimum belt lter control will vary considerable depending uponthe objectives o each operation. Not all plants may require the lowest moisture possible assome operations will consider throughput and focculant usage as equally important param-eters. There may even be the case o wanting the product to be o higher moisture content tomeet client specications. Regardless o plant objectives, there is little debate that the use oan accurate and reliable on-line moisture result will enable plant operators to tune the beltlter or their own objectives.

The ndings o this project suggest there is opportunity or this type o analyser in almostevery belt lter application world wide. The analyser could be used in industry or more eec-

tive means to monitor product quality and ecient use o focculant.


12/12

9


Conclusions and Recommendations

The authors eel they met the majority o the project objectives and conclude that the design,development and implementation o the analyser were very successul. Due to the age o thecloth and the serious blinding eect on some o the sensors, the plant condition variations

caused fooding on the belt, the result o which was that some o the collected data orsensors 1 and 2 was invalid. The eect o belt speed, cake depth and focculant levels weremeasured. It can be clearly seen that using the moisture analyser as a process tool would allowoperators to condently adjust belt lter parameters to achieve the plant objectives.

Monitoring the validity o the calibration is the only regular maintenance or the analyser. It issuggested that the analyser moisture be compared to shit moisture results or ailing thatmonthly spot checks to determine whether recalibration is required to ensure accurate on-linemoistures are produced.

Further research into the ability to take the same technology into other coal processingapplications such as ilter presses, drum ilters, centriuges, and other dewatering techniquesis recommended.

Technology Transfer Issues

Whilst this project creates an opportunity to introduce existing technology into a new applica-tion, it is the project teams belie that there is insucient novelty in this concept to warrantpatentability. Callidan Instruments currently hold a patent or their microwave technology,which constitutes measuring attenuation and phase shit o a received microwave signal or aswept requency range through a material to determine the moisture o the material. This proj-

ect merely explores another avenue to apply the currently patented technology.

Acknowledgments

The project team would like to acknowledge the efforts of the Gregory CHPP staff who assisted withthe installation and commissioning of the analyser and also allowed process changes to occur upontheir belt filter for the purposes of the project.

Annexes

Microwave Emission Information

The MOISTSCAN Moisture Analyser is designed to use low level non-ionizing microwave radia-tion. The microwave power emitted rom the MOISTSCAN is less than 10mW (10dBm). Thiscomplies with AS/NZS 4268 which species the maximum Equivalent Isotropically RadiatedPower (EIRP) or short range radio equipment.

This radiation level exists directly between the two antennas, which in almost all cases is inac-cessible due to the conveyor belt. Microwave radiation urther than 1m rom the analyser isvirtually undetectable.

ReferencesiEdward, D and Clarkson, C., 1999, Sampling Statistics Toolkit,ACARP Project C3090, Australian Coal Research Pty Ltd.

Documents

Acps11th CD-hvbf Moisture Control