Ultrasonic C-scan Detection for Stainless Steel Spot Welds

© 2014 ISIJ 1876

ISIJ International, Vol. 54 (2014), No. 8, pp. 1876–1882

Ultrasonic C-scan Detection for Stainless Steel Spot Welds Based on Signal Analysis in Frequency Domain

Jing LIU, Guocheng XU,* Xiaopeng GU, Guanghao ZHOU and Yongkui HAO

Key Laboratory of Automobile Materials of Ministry of Education and Department of Materials Science & Engineering, JilinUniversity, 5988 Renmin Street, Changchun City Jilin Province, China.

(Received on January 18, 2014; accepted on May 2, 2014)

In this paper, ultrasonic C-scan detection is conducted on stainless steel spot welds, and C-scan imagesare obtained respectively through characteristic signal analysis in time domain and frequency domain. C-scan image of frequency domain characteristics signal, which is less affected by detecting conditions, canreflect the weld nugget morphology more truly. After C-scan image enhancement and edge detection pro-cessing, the dimension of spot weld nugget can be automatically obtained through equivalent diameteralgorithm procedure. The dimension of nugget contains corona bond, and it is in good agreement withcorona bond external diameter through metallographic measured value. So it can be taken as the refer-ence to evaluate the quality of spot welds. Without the influence of corona bond, detection accuracy isvery high. The error is less than 0.066 mm.

KEY WORDS: stainless steel; spot welding; ultrasonic C-scan detection; signal analysis.

1. Introduction

Along with the expansion of resistance spot welding tech-nique and increasing requirements for welding quality, jointquality assessment has become an important issue in weld-ing quality control. In the field of modern automobile andrailway vehicles manufacturing, reliable joints are the key toguarantee safety and service life.1,2)

In recent years, many researchers at home and abroad arecommitted to the research on nondestructive testing technol-ogy of resistance spot welding. Ultrasonic nondestructivetesting technology attracts special attention with its conve-nience, efficiency, reliability, security and economical effi-ciency.3) Traditional spot welding ultrasonic C-scan detectingmostly analyzes signal in time domain, which is greatlyinfluenced by testing conditions, such as probe tilt, couplingof probe and the work piece, or uneven surface roughnessof part of work piece and so on. These effects may causeecho signal severe attenuation, resulting in C-scan imagesdistort to a certain degree or fuzzy character of nugget edge,etc. It not only affects the operator manual calibration accu-racy, but also brings bigger difficult to image feature automaticrecognition, not conducive to realize computer automationand intelligent detection.4–6) Based on characteristics analy-sis of A echo signal in frequency domain, this paper takesmain frequency values as the frequency-domain feature sig-nal for C-scan images, regardless of frequency amplitude,which effectively avoids the influence of testing conditionson ultrasonic echo intensity, gets closer to the real nuggetmorphology C-scan images, then eliminates errors brought

by the operator manual calibration, and finally realizes com-puter automatic detection more accurately. Meanwhile, inthis paper, a new kind of image processing method isapplied. On the basis of getting ultrasonic C-scan image inthe weld nugget zone, frequency domain C-scan images areprocessed on computer image enhancement, edge detectionand nugget diameter measurement. It further highlights thenugget characteristics, greatly reduces the computationamount of data, and then quickly gets spot welding nuggetsize, finally completes real-time, automatic and intelligentassessment of the quality of spot welds. All of the aboveimages processing methods are automatically done by com-puter, which can fully guarantee the stability and the accu-racy of analysis results.

2. Materials and Experiments

2.1. Specimen PreparationSUS304 stainless steel (0Cr18Ni9) is used in this study,

with the chemical composition and mechanical propertiesshown in Tables 1 and 2, respectively. The specification ofspecimen is shown in Fig. 1. The thickness of the specimenis 2 mm, width is 30 mm, and length is 170 mm. To clearup the specimen surface by sand paper before welding toensure that the specimen surface keeps in the same state. Weuse DC welder to weld the specimens. In the process of spotwelding, the electrode couple use planar electrode + spher-ical electrode (R100 mm) combination, as shown in Fig. 2.Under the condition of normal spot welding process, weld-ing specimen has obvious indentation in the side of thespherical electrode; the indentation in the side of the planeelectrode is not obvious, close to the plane. In order to avoid

* Corresponding author: E-mail: [email protected]: http://dx.doi.org/10.2355/isijinternational.54.1876

ISIJ International, Vol. 54 (2014), No. 8

1877 © 2014 ISIJ

welding indentation’s influence on test result, this studyfocuses on ultrasonic testing on the side without indentation,and uses water as the coupling agent. We use different weld-ing specifications (see Table 3) to get four sets of specimenswith different nugget diameter. There are three specimens ineach group. Spot welding cycle diagram of the relationshipbetween electrode force and welding current in the spotwelding process is shown in Fig. 3.

2.2. Testing Device and MethodThe ultrasonic testing device used in the experiment is

comprised of portable industrial computer, 15 MHZ focusedultrasonic probe, ultrasound card, and X-Y mechanicalscanning platform. The ultrasonic scanning test processoperates automatically according to the set program by com-

puter. Stepping scanning precision is up to 0.02 mm.The probe driven by linear motor scanning platform will

scan in the X axis direction parallel to nugget and Y direc-tion perpendicular to the nugget. The mode is S-type com-plete coverage scanning. And step length is 0.08 mm. Theprobe scanning path is shown in Fig. 4.

3. Ultrasonic C Scan Image of Spot Welds

Spot welding joints of specimens are scanned and detect-ed using the ultrasonic testing device. A-scan waveform dia-grams in different parts of the joint and the correspondingspectrum characteristic curves after fast Fourier transformhave different characteristics, as shown in Fig. 5. In theposition of non-welding zone a, since the impedance of thesteel plate is far greater than that of the air, ultrasonic inci-

Table 1. Chemical composition of SUS304 austenitic stainless steel (wt.%).

C Si Cr Ni N Mn P S Fe

≤0.07 ≤1.00 18.00–20.00 8.00–11.00 0.01–0.25 ≤2.50 ≤0.035 ≤0.030 Bal.

Table 2. Mechanical properties of SUS304 austenitic stainless steel.

Yield Strength(N/mm2)

Tensile Strength(N/mm2)

Elongation%

HardnessHV

≥205 ≥520 ≥40 ≤200

Fig. 1. Geometry and dimension of specimen (in mm).

Fig. 2. Resistance spot welding sketch diagram.

Table 3. Welding parameters used for the stainless steel.

Group I1(KA)

I2(KA)

t0(cyc)

t1(cyc)

t2(cyc)

t3(cyc)

t4(cyc)

Electrodeforce (KN)

1 5 6 15 6 5 20 15 10

2 5 7.5 15 6 5 18 15 10

3 5 9.5 15 6 5 16 15 10

4 5 11 15 6 5 13 15 10

Fig. 3. Resistance spot welding cycle diagram.

Fig. 4. The probe scanning path schematic diagram.

© 2014 ISIJ 1878


dent to the underside of the upper plate almost all reflect. Atthis time, the echo amplitude Am from upper plate bottomis larger and the distance between the sound waves is thesonic path distance of the upper plate thick. The maximumamplitude corresponding to the main frequency on the spec-tral characteristic curve is 2 MHz, as shown in Figs. 5(a1)and 5(a2). In the position of the combination line b, part ofthe ultrasonic wave enters into the lower plate through thecombination line, and cause Am reduce. At this time, themaximum amplitude corresponding to the main frequency is0.75 MHz, as shown in Figs. 5(b1) and 5(b2). In the positionof the welding zone c, due to metal fusion in the joint with-out existence of phase interface, almost all ultrasound pen-etrate into the nugget. Reflection occurs at the bottom of thelower plate. At this time, the echo amplitude Am from lowerplate bottom is larger and the distance between the soundwaves is the sonic path distance of the two layers of theplate thick. The maximum amplitude corresponding to themain frequency is 1 MHz, as shown in Figs. 5(c1), 5(c2). The

above analysis shows that in different positions of spotwelding joint, echo amplitude Am and main frequency val-ues both have obvious changes. Therefore, they can berespectively used as characteristic parameters representingjoint internal fusion state. To analyze the signal from theangle of time domain and frequency domain and respectivelyto present C scan images in the form of grey value.

3.1. C Scan Images of Time Domain Characteristic Sig-nal (Echo Amplitude Am)

Under the condition of other spot welding process param-eters unchanged, nugget size increases with the increase ofwelding current within a certain range. Welding current andthe formed nugget feature have a relatively fixed relation-ship. This article selects two welding specimens under thecondition of different current. Take the echo amplitude Amas the characteristic signal to get C scan images. The resultis shown in Fig. 6. As shown in the figure, two images bothhave a clear outer ring and an inner circle growing prior to

Fig. 5. A scanning waveform diagrams in different parts of the joint and the corresponding spectrum characteristic curveafter fast Fourier transform. (a1) non-welding zone, (b1) combination line, (c1) welding zone and (a2)–(c2) corre-sponding FFT curve. (Online version in color.)


1879 © 2014 ISIJ

the outer ring. And with the increase of current, the area ofthe inner circle grows. By testing the rest of the specimen,we can find that the ring characteristic is ubiquitous. Theexternal ring occurs in the process of resistance spotwelding7) and the inner circle corresponds to the nuggetarea. It is worth noting that some test conditions may causeimpact on C scan images of time-domain signal , such asprobe tilt, coupling of probe and the workpiece, or unevensurface roughness of part of workpiece and so on. Theseeffects may cause echo signal severe attenuation, resultingin C scan images distort to a certain degree or fuzzy char-acter of nugget edge, etc. It not only affects the visibility ofthe image, but also makes larger difficult to image featureautomatic recognition, not conducive to realize computerautomation, and intelligent detection. In this case, nuggetedge needs testers to demarcate manually. Yet even if it isoperated by engineers with some NDT experience, a largenumber of false inspection and leak inspection can also befound. So in order to improve the reliability of test resultsand to gain the real nugget size, another frequency domainsignal analysis with low impact by testing conditions is tak-en into consideration.

3.2. C Scan Images of Frequency Domain Characteris-tic Signal (Main Frequency Value)

To make frequency domain analyses of A-echo signal oftwo specimens in Fig. 6 through the Fourier transform.Without regard to the main frequency amplitude, take mainfrequency values as the frequency domain feature signal forC scan images, which avoid the effect of testing conditionson the intensity of ultrasonic. The result is shown in Fig. 7.

From the figure we can see, the circle in the characteristicnugget area is similar to the inner circle in Fig. 6. The colorof the area outside the nugget is single, without the interfer-ence of parti-colored. Besides, in the figure, boundary char-acteristics of nugget area are clear, which can truly reflectthe weld nugget appearance, measure nugget diameter accu-rately, and evaluate the quality of spot welding reliably.

4. C Scan Image Data Processing of Frequency DomainCharacteristic Signal

In the process of ultrasonic testing, the influence of manyfactors, such as detection system resolution, signal-to-noiseratio and materials with uneven texture, etc., can bring noiseto the echo signal. In order to make the image show nuggetmorphology more clearly and realize the quantitative mea-surement of the weld nugget diameter, in this paper, C scanimages in frequency domain are processed through comput-er image enhancement, edge detection and nugget diametermeasurement. Finally to complete real-time, automaticassessment of the quality of spot welding.

4.1. Image EnhancementImage enhancement is a processing method which selec-

tively highlights interested image features while ignoressome needless features through adding some information onthe original image or transforming data by certain means soas to improve image quality and strengthen the effect ofimage recognition. In order to filter out interference noiseeffectively, and protect the nugget edge information verywell, a variety of filter methods are used on dealing with C

Fig. 6. Different welding current specimen time domain C scan images. (a) 9 KA, (b) 11 KA. (Online version in color.)

Fig. 7. Different welding current specimen frequency domain C scan images. (a) 9 KA, (b) 11 KA. (Online version incolor.)

© 2014 ISIJ 1880


scan images in this paper. After a large number of experi-mental comparison and analysis, we find that the effect ofmedian filtering with nonlinear smooth characteristics onspot welding C scan image processing is best. After medianfiltering processing in 7 * 7 modules, the output image ofFig. 7 is shown in Fig. 8. We can see, after median filtering,it not only filters out sharp wave interference noise andmakes nugget boundary more clearly in C scan images, butalso further improves the quality of C scan images, whichlays foundation for subsequent image edge extraction.

4.2. Edge DetectionThis paper further chooses Roberts-operator, which is

sensitive to noise signal, to make edge detection on theimage and obtain accurate nugget boundaries. The resultsare as shown in Fig. 9. In order to conveniently identify andanalyze nugget area in the image, morphology is used tosegment the needed feature information, as shown in Fig.10. After segmentation, not only does the nugget featurebecome prominent in the image, but also greatly reduces thecomputation amount of data and provides favorable condi-tions to get the real-time nugget dimension and assess spotwelding quality rapidly.

4.3. Nugget Diameter Evaluation_Equivalent DiameterMethod

Spot welding quality is characterized by the strength ofjoint. And the strength of the joint mainly depends on thedimension of the nugget, especially the size of nugget diam-eters.8) How to get connection dimension of the joint quan-titatively is the key of the spot weld diameter evaluation.This paper adopts equivalent diameter method, namelycalculates average nugget diameter on the spot weld jointsurface to represent the welding quality of joint.

Pick up N point on the nugget area boundary. Take twopoints of the N points as midperpendicular. The intersection

point is the center of the circle. Then centre point cluster isgot. Take the average of this point cluster as the center. Theaverage distance from the coordinate to the N boundarypoint is the weld nugget diameter. Figure 11 is the schematicdiagram of any two midperpendicular for center of the circle.

5. Test Results Analysis

C scan images of the twelve spot welding specimens withdifferent weld quality are obtained after signal data process-ing. In order to evaluate the accuracy of weld nugget dimen-sion by ultrasonic C scan image detecting, the specimens,after ultrasonic testing, are split along the combining surfaceof the upper and lower plate to get the nugget cross section’smetallographic detective area (as shown in Fig. 12).Measured value of specimen nugget dimension is gained.Make comparison between specimen ultrasonic detectionvalue and metallographic measured value. The results areshown in Fig. 13(a). De refers to ultrasonic testing nuggetdiameter detection value; Dm refers to spot weld specimennugget diameter measured value.

From Fig. 13(a) we can see detection value De is gener-ally larger than measured value Dm. The average relativeerror is 12.05%. The reasons for this phenomenon can bespeculated on the existence of the corona bond. Take out thesample of the maximum error and observe the metallograph-ic photos, as shown in Fig. 14. A round of wide corona bondis found outside the nugget. In addition, external diametervalue Dc of the corona bond (corona bond plus nugget over-all dimension) is close to the detection value De. In thispaper, the rest of the specimens are carried on the compar-ative analysis. The external diameter value of spot weldingcorona bond is in good agreement with the ultrasonic detec-tion value, as shown in Fig. 13(b). De refers to ultrasonic

Fig. 8. Frequency domain C scan images after median filtering. (a)9 KA, (b) 11 KA. (Online version in color.)

Fig. 9. Frequency domain C scan images after edge detection. (a) 9KA, (b) 11 KA. (Online version in color.)

Fig. 10. Frequency domain C scan images after edge segmenta-tion. (a) 9 KA, (b) 11 KA.

Fig. 11. The schematic diagram of equivalent diameter method.(Online version in color.)


1881 © 2014 ISIJ

testing nugget diameter detection value; Dc refers to exter-nal diameter measured value of spot welding specimencorona bond. The test results show that the value obtainedfrom C scan images detection based on frequency domain

Fig. 12. Solder joint macro metallographic photos.Fig. 14. Maximum error specimen welding spot macro metallo-

graphic photos. (Online version in color.)

Fig. 13. Experimental results comparison: (a) comparison figure of detection value De and diameter measured value Dm,(b) comparison figure of detection value De and the plastic ring measured value Dc. (Online version in color.)

Fig. 15. (a) Spot welding specimens longitudinal section metallographic photos, (b)–(c) separately represents the magni-fication of the corresponding position of (a).

© 2014 ISIJ 1882


characteristic signal processing contain corona bond dimen-sion. It can be used as the reference value of evaluating spotweld quality.

Spot welding specimens longitudinal section metallo-graphic photos, as shown in Fig. 15. From the micrographin the spot welding plastic ring zone in Fig. 15(b), we cansee although specimen two board interface fit very close, itis still incomplete fusion area, not real nugget area. Asshown in Fig. 15(c), nugget organizations are composed ofcolumnar austenitic plus ferrite along the distribution of thecolumnar crystal. Developed columnar crystal grown upfrom nugget edge to joint surface touching each other. Phaseinterface does not exist. If not considering the influence ofthe organization on ultrasonic, we can put the nugget and thebase as a unified whole theoretically. In order to evaluateprecision and stability of the weld nugget size by ultrasonicC scan image detecting, under the circumstance of smallplastic ring or no plastic ring, we remove the not connectedlower plate. A set of standard specimen is made in this paper

which simulates ideal weld nugget appearance without plas-tic ring at the most extent. Its nugget dimension ranges from4.0 mm to 8.5 mm, and the interval is 0.5 mm, as shown inFig. 16. The standard specimen avoids the possibility ofdamaging nuclear metal due to stretching or tearing, andreduces the measuring error. It has better accuracy andrepeatability compared with joints after tensile fracture.

Figure 17 is the comparison figures of standard specimenand ultrasonic test results. We can see that without the influ-ence of the corona bond, C scan images based on frequencydomain feature signal has very high accuracy on detectionof weld nugget diameter. The error is less than 0.066 mm.It also indirectly proves that the error of the test methodmainly comes from the corona bond. Further research onhow to distinguish the nugget from corona bond to obtainmore accurate nugget diameter is what we are working onat present. Of course, if some strong norms are used to makethe corona bond very narrow in the process of resistancewelding or the ultrasonic detection accuracy requirement isnot high, the detection error by using the method is negligible.

6. Conclusions

(1) In this paper, stepping mechanical scanning methodis used on stainless steel spot welding joint to conduct ultra-sonic C scan detection. Then signal characteristics areextracted respectively from time domain and frequencydomain and weld nugget zone C scan images are obtained.After comparative analysis, we find that C scan imagesobtained by frequency domain characteristics signal are lessaffected by detecting conditions, which can reflect the weldnugget morphology more truly.

(2) Image enhancement and edge detection processingon C scan images in frequency domain can provide clearnugget boundary morphology. Not only does nugget featurebecome prominent in the image, but also the computationamount of data is greatly reduced. The dimension of spotweld nugget can be automatically obtained through equiva-lent diameter algorithm procedure.

(3) Stainless steel spot welding ultrasonic C scan valueDe based on frequency domain signal analysis containscorona bond dimension, and it is in good agreement withcorona bond external diameter through metallographic mea-sured value Dc. So it can be taken as the reference to eval-uate the quality of spot welds. Without the influence ofcorona bond, detection accuracy is very high. The error isless than 0.066 mm.

REFERENCES

1) M. Jou: J. Mater. Process. Technol., 132 (2003), 102.2) S. Agashe and H. Zhang: Weld. J., 82 (2003), 179.3) R. J. Ditchburn, S. K. Burke and C. M. Scala: NDT&E Int., 29 (1996),

111.4) L. F. Shi, X. M. Zheng and Y. W. Shi: Nondestr. Test., 26 (2004), 34.5) M. Thornton, L. Han and M. Shergold: NDT&E. Int., 48 (2012), 30.6) H. Takada and T. Hirose: JFE Tech. Rep., 10 (2007), 26.7) M. Thornton, L. Han and M. Shergold: NDT&E. Int., 48 (2012), 30.8) I. C. Syal and R. Syal: Indian Concrete J., 7 (1997), 367.

Fig. 16. Standard specimen. (Online version in color.)

Fig. 17. Comparison figures of standard specimen and ultrasonictest results. (Online version in color.)

Documents

Ultrasonic C-scan Detection for Stainless Steel Spot Welds