Grain Size Measurement in Steel by Laser Ultrasonics Basedon Time Domain Energy

Fei He+, Yin Anmin and Yang Quan

National Engineering Research Center of Flat Rolling Equipment, University of Science and Technology Beijing,Beijing 100083, China

Laser ultrasonics, a technique based on the generation of ultrasonic waves by a pulsed laser and on their detection by a laser interferometer,can be used to determine grain sizes in steels. The absolute values of the average grain size can be calculated directly from the attenuationmeasurements of ultrasonic longitudinal bulk waves. The general methods of attenuation measurements are computed based on the amplitude oflongitudinal waves. In fact, the amplitude as only one value is infected by the noise. Besides, the attenuation should be considered as the energyreduction. The method of time domain energy is proposed to compute the attenuation. The results indicate that time domain energy can improvethe accuracy of the grain size prediction. The laser ultrasonic technique may be incorporated online for direct measurements of grain size duringsteel production. [doi:10.2320/matertrans.M2014445]

(Received December 11, 2014; Accepted March 5, 2015; Published April 10, 2015)

Keywords: laser ultrasonics, ultrasonic attenuation, grain size, steel sheet, time domain energy

1. Introduction

The grain size and crystalline texture of metal sheetsstrongly affects its formability. The grain size and textureplays an important role in determining the subsequentformability into such finished parts as beverage cans andautomotive components. Since grain size is one of the factorswhich influence the material properties of steel products, it isimportant to maintain grain size in permissible ranges.

Characterization of material, during production as well asdevelopment, is usually carried out using time-consumingand costly methods such as microscopy and mechanicaltest.1) It is well known that laser-ultrasonic (LU) informationabout microstructural and metallurgical properties such astexture and grain size can be used to extract from themeasured ultrasonic wave velocity and attenuation in thematerials.2,3) Results have also been published on austeniterecrystallization and grain growth in steel alloys.4­6)

In rolled strip material, the grain size is influenced byvarious kinds of thermo-mechanical processes applied in therolling mill (hot and cold rolling and annealing). It is of greatimportance that the quality should be homogeneous over theentire length of a strip; this can only be monitored using acontinuous, on-line measuring method. The combined resultsof the studies presented in the literature have firmly shownthat the laser-ultrasonic techniques (LU) can be used asthis.6)

Ultrasonic waves can penetrate a few percent of meters inmetals, although the wave amplitude becomes progressivelyattenuated by the microstructure as it travels over longerdistances. The attenuation experienced is dependent on grainsize such that large grains have a dominating influence.Depending on the wavelength/grain size ratio (­/D), theattenuation ¡ can be described as a Rayleigh region(¡(­,D) = CrD3f 4, ­ º D), a stochastic region (¡(­,D) =CsDf 2, ­ μ D,) and a diffusive region (¡(­,D) = Cd/D,­ ¹ D), where Cr, Cs, and Cd are constants of the materialand D is the diameter for spherical grains.

One way to obtain the attenuation from ultrasonicmeasurement is from the relative difference between theamplitude of two successive echoes. In this way, theattenuation parameter can be calculated the followingrelationship.1,7)

¡ ¼ 1

2LlogðU1=U2Þ ð1Þ

Where, U1 and U2 are the amplitude of the 1P and 3P. Thepropagation path of the first echo and the second echo arenamed as 1P and 3P. L is the thickness of the sample.

In this method, the amplitude is selected as the maximumvalue of the 1P and 3P echoes from the time domain signals.And then, the attenuation parameters of each sample werecalculated. In real measurement, the results are affected bythe noise from the equipment stability, the sample surfacestate, the environmental difference, the second phases,precipitation, dislocations and so on. If only the maximumvalues of the 1P and 3P echoes are selected as the intensity,maybe the accuracy is influenced by the noise seriously.Besides, the attenuation should be considered as the energyreduction. In this paper, time domain energy method isproposed to compute the attenuation, which was shown in2.2.

2. Experimental and Time Domain Energy Method

2.1 ExperimentalThe samples inspected were steel plate heat commercial

(SPHC) steel. The samples were taken from coils directlyafter hot rolling. Grain size was an important propertyof these hot-rolled strips. Four samples with dimensions75mm © 65mm © 3.35mm were cut from the coil forexperimental study. Three of them were annealed in orderto adjust grain size, the holding temperature is 800°C and theholding time is 30­90min, then they were cooled to roomtemperature and a small piece 8mm © 9mm was cut forEBSD measurement. The samples were prepared usingstandard mechanical polishing and electro-polishing (5%perchloric acid solution under an applied potential of 15V+Corresponding author, E-mail: hefei@ustb.edu.cn

Materials Transactions, Vol. 56, No. 6 (2015) pp. 808 to 812©2015 The Japan Institute of Metals and Materials

during 60 s). After acid pickling and electrolytic polishing thesamples were cleaned for 2min in an ultrasonic bath, TheElectron Backscattered Scattering Detection (EBSD) datawere obtained using a ZEISS ULTRA 55 Field-EmissionScanning Electron Microscope (SEM). The measured area inthe middle of the samples was 1000 µm © 400 µm on a stepsize 1.5 µm. The samples had high indexing rates (over 99%)after noise reduction, indicating a low level of errors/noisein the map. The grain sizes of the four samples weredetermined from EBSD analyses by measuring the meanlinear intercepts.

Figure 1 shows the laser-ultrasonic measurement model.The laser generator uses a Q-switched Neodium: Yttrium-Aluminum-Garnet (Nd:YAG) pulsed lasers as a transmitter ofthe ultrasonic wave, the wavelength was 532 nm, the laserpulse energy was 200mJ and pulse width was 8 ns, thediameter of the pulse laser beam is 6mm, after focused by thelens, the diameter of the pulse laser spot on the middle of thesample surface was about 2mm, this will not damage thesample surface and the laser-ultrasonic measured area wasabout the same size with the EBSD measured area. Theultrasonic detection equipment was based on two wavemixing interferometer (bandwidth 120MHz, wavelength1550 nm). The sample rate was 50MHz and 1000 sample

points were collected each time. In order to reduce the noise,in the sample position the signals collection were repeated 8times and then the average values in the time domain were asthe output ultrasonic signals. The received ultrasonic signalswere input to a differential circuit (equivalent to a high-passfilter) and a digital oscilloscope. Finally, the signals werestoried in a computer.

And then, the attenuations were got from the laser-ultrasonic measurement. One sample was used to compute Cr,constant of the material. The grain size of the other threesamples can be calculated.

2.2 Time domain energyAfter time-averaging and frequency filtering, the noise

from the equipment stability, the sample surface state, theenvironmental difference can be reduced. The noise fromthe equipment stability that is low frequency signal, can bereduced by frequency filtering. The noise from the samplesurface state and environmental difference can be reduced bytime averaging.

But the noise from the microstructure difference indifferent samples (in other words, the microstructure of eachsamples are not strictly same, the Second phases, Precip-itation, Dislocations which also effect the ultrasonic waveattenuation are slightly different) cannot be reduced.

In order to reduce the influence, the time domain energy isproposed. The analysis benefits from some post-processing,which was depicted in Fig. 2. Signal ‘echo’, in the upperleft corner, was an experimentally obtained time signal inFig. 2(a). The first echo and the second echo spectrum,labeled “zoom of 1P” and “zoom of 3P” in Fig. 2(b),contained the acoustic information. After invert the signalslabeled “zoom of 1P” and “zoom of 3P” by “time = 0” as asymmetry axis, signals “inverted 1P” and “inverted 3P” wereobtained shown in Fig. 2(c). A final step reduced the noise byfitting the signals “inverted 1P” and “inverted 3P” using theMATLAB toolbox with a Gaussian function, labeled “fitting

Pulse Laser 532nm

Detection Laser 1550nm

Two wave mixing interferometer

Control UnitDigital

oscilloscopePC

200mJ/8ns pulse, 20Hz

100mW continuous

Sample

lens

Fig. 1 Laser-ultrasonic measurement model.

(a) (b)

(c) (d)

Fig. 2 Sketch of Time domain energy; (a) echo signals, (b) zoom of signals, (c) inverted signals, (d) integral signals.

Grain Size Measurement in Steel by Laser Ultrasonics Based on Time Domain Energy 809

of inverted 1P” and “fitting of inverted 3P” in Fig. 2(d). Inthe final spectrum, the full width at half maximum (FWHM)was chosen to compute the “integral of 1P” and “integral of3P”. And the integral of 1P and 3P named as A1, A2, that werecalculated as the energy of 1P and 3P. Then, the attenuationparameter can be calculated the following relationship.

¡ ¼ 1

2LlogðA1=A2Þ ð2Þ

Where, A1 was the energy of 1P; A2 was the energy of 3P;L is the thickness of the sample.

3. Experimental Results and Analysis

3.1 EBSD measurement resultsThe microstructure of the four steel samples measured by

EBSD was shown in Fig. 3. The mean linear interceptmethod for measuring “grain size” was well establishedaccording to ASTM E1128) and has been implemented. Amajor difference between this implementation and moreconventional ones is that the crystallographic orientation datais being used rather than a processed image from an etchedspecimen. The advantage of EBSD is that there is noambiguity about the grain boundaries since they are measuredin terms of crystallographic data. Parallel test lines are“drawn” over the map and the points where the lines intercepta grain boundary are noted. The mean linear intercept iscalculated by adding all the line segments together anddividing by the number of complete grains the test linespassed through. Incomplete intercepts, from grains that touch

the edges of the map, are not included in the statistics. Toavoid sampling a grain more than once, the lines should beat least the maximum grain diameter apart. In this measure-ment, 10 parallel test lines were drawn over the map andthe average grain size and error bar were automaticallycalculated, which were shown in Table 1.

3.2 Laser ultrasonic measurement resultsThe surface vibration caused by ultrasonic wave was

detected by the laser ultrasonic receiver. The receivedultrasonic signals were shown in Fig. 4. The vertical axisof the detected ultrasonic waveforms represented theamplitude of ultrasonic wave detected by the laser ultrasonicreceiver, and the horizontal axis showed the time followingirradiation by the pulsed laser. It was found that the echoesbecome weaker as the grain size increases.

3.3 Analysis and discussionIt was assumed that the main contribution to the

attenuation was the scattering of the ultrasonic pulse at thegrain boundaries, other contributions such as absorption atdislocations and precipitates were neglected.

In general, the attenuation of the two successive ultrasonicechoes can be obtained from eq. (1), the amplitude wasselected as the maximum value of the 1P and 3P echoes inFig. 4. The attenuation of the four steel samples calculated bythe general method was listed in Table 1. From Table 1 it wasfound that attenuation parameters increase with the increaseof the average grain size. From above, the average grain sizesand the attenuation parameters of each sample were got, and

Fig. 3 Microstructure of SPHC steel samples.

Table 1 Attenuation parameters and the grain sizes.

SamplesThickness

/mm

MeasuredGrain size

/µm

General method Energy method

AttenuationPredictedGrain size

AttenuationPredictedGrain size

SPHC-1 3.940 13.2 0.0251 / 0.0279 /

SPHC-2 3.872 13.1 0.0243 13.06 0.0273 13.10

SPHC-3 3.777 16.9 0.0461 16.17 0.0540 16.45

SPHC-4 3.719 20 0.0573 17.38 0.0710 18.02

F. He, Y. Anmin and Y. Quan810

then the material constants Cr can be calculated using theequation in Rayleigh region (¡(­,D) = CrD3f 4, ­ º D). Inthis paper, the material constant Cr was calculated using theSPHC-1 that called as calibration sample. Then the grainsizes of the other three samples were computed and listed inTable 1.

Here, the attenuation of the four steel samples was alsocalculated by the time domain energy method and listed inTable 1. Similarly as above, the material constants Cr inRayleigh region were calculated. The SPHC-1 was used ascalibration sample to predict the grain size of other samples.The results were also listed in Table 1.

The measurement grain size by EBSD, calculated valuesby the general method and energy method were shown inFig. 5 and Table 1. Then it was found that:

(1) The time domain energy method can improve theaccuracy of the calculated grain size. The calculated grainsizes of the other three samples from LU were more close tothe EBSD compared with the general method. For example,the average grain size of SPHC-3 from EBSD was 16.9 µm.The grain size from the time domain energy method was16.45 µm, compared with 16.17 µm from the general method.

(2) The average grain size of SPHC-2 from EBSD was13.1 µm, and the calculated average grain size from LU wasvery close to the EBSD. With the increasing of the differenceof the grain size between the calibration sample and testingsamples, the error also increased. For example, the calculatedaverage grain size of sample SPHC-4 was 18.02 µmcompared with measured by EBSD 20 µm, it was also moreerror than sample SPHC-2 and SPHC-3.

(3) The average grain size of the steel samples predicted bytime domain energy method had a higher accuracy than thegeneral method, especially when the average grain size of thetesting sample had a big difference with the calibrationsample.

The results presented in Table 1 shown that there was aclose relation between grain size obtained from EBSD imageanalysis and grain size calculated from ultrasonic wavesanalysis. The stress state, orientation distribution state andmetallurgical state were also changed for the detectedsamples. Any changes in the detected samples would hamperthe analyses and become a source of error to the analysisresults.

4. Online Measurement Based on Laser-Ultrasonic

The modern iron and steel production and application oftechnical requires the steel products with high performanceand uniformity quality. The conventional quality and me-chanical properties testing mainly relies on a static, offlineand destructive methods. Although these methods can obtainthe performance parameters of the final products, they areless efficient and great randomness, and cannot detect themicrostructure, mechanical properties and control the prod-ucts quality online throughout the production process. So, itis of strategic significance to investigate and develop theonline determination and quality control technology, for

(a) (b)

(c) (d)

Fig. 4 Detected ultrasonic waveforms of the 1P and 3P echoes; (a) SPHC-1, (b) SPHC-2, (c) SPHC-3, (d) SPHC-4.

Fig. 5 Comparison of grain size calculation.

Grain Size Measurement in Steel by Laser Ultrasonics Based on Time Domain Energy 811

online, whole process, real time and non-destructivedetection the plate quality to improve the overall quality ofthe steel industry and steel products, and ensure the steelproducts with a high stability, high uniformity, highconsistency of quality and performance.7)

The ultimate purpose of on-line measurement in theproduction line is feedback control of the production processitself. This is possible if the measured quantity on thecontrollable production parameter is quantitatively known.Figure 6 shows the sketch of on-line laser ultrasonicmeasurement system. A Neodium:Yttrium-Aluminum-Garnet(Nd:YAG) laser is employed to simultaneously generate theseultrasonic modes. For the ultrasonic detection, a continuouslaser is coupled to a photorefractive interferometer by opticalfibers.

5. Conclusion

The average grain sizes and material constants Cr of steelswere determined and analyzed by laser ultrasonic. Using thecalculated material constants Cr the average grain sizes ofother steel samples were calculated, and there was a closerelation between grain size obtained from EBSD imageanalysis and grain size calculated from ultrasonic waveanalysis, and these results may be useful for online measure-ment based on laser-ultrasonic. The time domain energy

method can improve the accuracy of the calculated grain size.With the increasing of the difference of the grain size betweenthe calibration sample and testing sample, the error alsoincrease. In the future the on-line measurement in theproduction line can be used for feedback control during therolling, and the proposed time domain energy method in thispaper can improve the accuracy of the grain size prediction.

Acknowledgments

This research is supported by National Natural ScienceFoundation of China (Grant no. 51204018), the National KeyTechnology R&D Program of the 12th Five-year Plan ofChina (Grant no. 2012BAF04B02, 2015BAF30B01), theResearch Fund for the Doctoral Program of Higher Educationof China (Grant no. 20110006110027), the Beijing YouthTalents Program of University (Grant nos. YETP0422), theFundamental Research Funds for the Central Universities(Grant no. FRF-TP-14-124A2).

REFERENCES

1) S. Sundin and D. Artymowicz: Metall. Mater. Trans. A 33 (2002) 687­691.

2) M. Falkenström, M. Engman, E. Lindh-Ulmgren and B. Hutchinson:Nondest. Test. Eval. 26 (2011) 237­252.

3) C. S. Lim, S. T. Hong, J. K. Yi, S. G. Choi, K. J. Oh, Y. Nagata, H.Yamada and N. Hamada: Rev. Quant. Nondest. Eval. 26 (2007) 1175­1182.

4) A. Smith, S. E. Kruger, J. Sietsma and S. van der Zwaag: ISIJ Int. 46(2006) 1223­1232.

5) M. Maalekian, R. Radis, M. Militzer, A. Moreau and W. J. Poole: ActaMater. 60 (2012) 1015­1026.

6) D. Lévesque, S. E. Kruger, G. Lamouche, R. Kolarik, II, G. Jeskey, M.Choquet and J.-P. Monchalin: NDT&E Int. 39 (2006) 622­626.

7) A. Yin, Q. Yang, F. He and H. Xiao: Mater. Trans. 55 (2014) 994­997.8) ASTM E112-12:2012, Standard Test Methods for Determining Average

Grain Size.

Detector

Coile

Roller

Generation LaserLaminar Cooling

Rolling Mill

Heating Furnace

Fig. 6 Sketch of on-line laser ultrasonic measurement.

F. He, Y. Anmin and Y. Quan812