Regular Article
Ultrasonic C-scan Detection for Stainless Steel Spot Welds Based on Signal Analysis in Frequency Domain
2014 Volume 54 Issue 8 Pages 1876-1882
Details
2014 Volume 54 Issue 8 Pages 1876-1882
In this paper, ultrasonic C-scan detection is conducted on stainless steel spot welds, and C-scan images are 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, can reflect the weld nugget morphology more truly. After C-scan image enhancement and edge detection processing, the dimension of spot weld nugget can be automatically obtained through equivalent diameter algorithm procedure. The dimension of nugget contains corona bond, and it is in good agreement with corona bond external diameter through metallographic measured value. So it can be taken as the reference to evaluate the quality of spot welds. Without the influence of corona bond, detection accuracy is very high. The error is less than 0.066 mm.
Along with the expansion of resistance spot welding technique and increasing requirements for welding quality, joint quality assessment has become an important issue in welding quality control. In the field of modern automobile and railway vehicles manufacturing, reliable joints are the key to guarantee safety and service life.1,2)
In recent years, many researchers at home and abroad are committed to the research on nondestructive testing technology of resistance spot welding. Ultrasonic nondestructive testing technology attracts special attention with its convenience, efficiency, reliability, security and economical efficiency.3) Traditional spot welding ultrasonic C-scan detecting mostly analyzes signal in time domain, which is greatly influenced by testing conditions, such as probe tilt, coupling of probe and the work piece, or uneven surface roughness of part of work piece and so on. These effects may cause echo signal severe attenuation, resulting in C-scan images distort to a certain degree or fuzzy character of nugget edge, etc. It not only affects the operator manual calibration accuracy, but also brings bigger difficult to image feature automatic recognition, not conducive to realize computer automation and intelligent detection.4,5,6) Based on characteristics analysis of A echo signal in frequency domain, this paper takes main frequency values as the frequency-domain feature signal for C-scan images, regardless of frequency amplitude, which effectively avoids the influence of testing conditions on ultrasonic echo intensity, gets closer to the real nugget morphology C-scan images, then eliminates errors brought by the operator manual calibration, and finally realizes computer automatic detection more accurately. Meanwhile, in this paper, a new kind of image processing method is applied. On the basis of getting ultrasonic C-scan image in the weld nugget zone, frequency domain C-scan images are processed on computer image enhancement, edge detection and nugget diameter measurement. It further highlights the nugget characteristics, greatly reduces the computation amount of data, and then quickly gets spot welding nugget size, finally completes real-time, automatic and intelligent assessment of the quality of spot welds. All of the above images processing methods are automatically done by computer, which can fully guarantee the stability and the accuracy of analysis results.
SUS304 stainless steel (0Cr18Ni9) is used in this study, with the chemical composition and mechanical properties shown in Tables 1 and 2, respectively. The specification of specimen is shown in Fig. 1. The thickness of the specimen is 2 mm, width is 30 mm, and length is 170 mm. To clear up the specimen surface by sand paper before welding to ensure that the specimen surface keeps in the same state. We use DC welder to weld the specimens. In the process of spot welding, the electrode couple use planar electrode + spherical electrode (R100 mm) combination, as shown in Fig. 2. Under the condition of normal spot welding process, welding specimen has obvious indentation in the side of the spherical electrode; the indentation in the side of the plane electrode is not obvious, close to the plane. In order to avoid welding indentation’s influence on test result, this study focuses on ultrasonic testing on the side without indentation, and uses water as the coupling agent. We use different welding specifications (see Table 3) to get four sets of specimens with different nugget diameter. There are three specimens in each group. Spot welding cycle diagram of the relationship between electrode force and welding current in the spot welding process is shown in Fig. 3.
| 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. |
| Yield Strength (N/mm2) | Tensile Strength (N/mm2) | Elongation % | Hardness HV |
|---|---|---|---|
| ≥205 | ≥520 | ≥40 | ≤200 |
Geometry and dimension of specimen (in mm).
Resistance spot welding sketch diagram.
| Group | I1 (KA) | I2 (KA) | t0 (cyc) | t1 (cyc) | t2 (cyc) | t3 (cyc) | t4 (cyc) | Electrode force (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 |
Resistance spot welding cycle diagram.
The ultrasonic testing device used in the experiment is comprised of portable industrial computer, 15 MHZ focused ultrasonic probe, ultrasound card, and X-Y mechanical scanning platform. The ultrasonic scanning test process operates automatically according to the set program by computer. 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 direction perpendicular to the nugget. The mode is S-type complete coverage scanning. And step length is 0.08 mm. The probe scanning path is shown in Fig. 4.
The probe scanning path schematic diagram.
Spot welding joints of specimens are scanned and detected using the ultrasonic testing device. A-scan waveform diagrams in different parts of the joint and the corresponding spectrum characteristic curves after fast Fourier transform have different characteristics, as shown in Fig. 5. In the position of non-welding zone a, since the impedance of the steel plate is far greater than that of the air, ultrasonic incident to the underside of the upper plate almost all reflect. At this time, the echo amplitude Am from upper plate bottom is larger and the distance between the sound waves is the sonic path distance of the upper plate thick. The maximum amplitude corresponding to the main frequency on the spectral characteristic curve is 2 MHz, as shown in Figs. 5(a1) and 5(a2). In the position of the combination line b, part of the ultrasonic wave enters into the lower plate through the combination line, and cause Am reduce. At this time, the maximum amplitude corresponding to the main frequency is 0.75 MHz, as shown in Figs. 5(b1) and 5(b2). In the position of the welding zone c, due to metal fusion in the joint without existence of phase interface, almost all ultrasound penetrate into the nugget. Reflection occurs at the bottom of the lower plate. At this time, the echo amplitude Am from lower plate bottom is larger and the distance between the sound waves is the sonic path distance of the two layers of the plate thick. The maximum amplitude corresponding to the main frequency is 1 MHz, as shown in Figs. 5(c1), 5(c2). The above analysis shows that in different positions of spot welding joint, echo amplitude Am and main frequency values both have obvious changes. Therefore, they can be respectively used as characteristic parameters representing joint internal fusion state. To analyze the signal from the angle of time domain and frequency domain and respectively to present C scan images in the form of grey value.
A scanning waveform diagrams in different parts of the joint and the corresponding spectrum characteristic curve after fast Fourier transform. (a1) non-welding zone, (b1) combination line, (c1) welding zone and (a2)–(c2) corresponding FFT curve. (Online version in color.)
Under the condition of other spot welding process parameters unchanged, nugget size increases with the increase of welding current within a certain range. Welding current and the formed nugget feature have a relatively fixed relationship. This article selects two welding specimens under the condition of different current. Take the echo amplitude Am as the characteristic signal to get C scan images. The result is shown in Fig. 6. As shown in the figure, two images both have a clear outer ring and an inner circle growing prior to the outer ring. And with the increase of current, the area of the inner circle grows. By testing the rest of the specimen, we can find that the ring characteristic is ubiquitous. The external ring occurs in the process of resistance spot welding7) and the inner circle corresponds to the nugget area. It is worth noting that some test conditions may cause impact on C scan images of time-domain signal , such as probe tilt, coupling of probe and the workpiece, or uneven surface roughness of part of workpiece and so on. These effects may cause echo signal severe attenuation, resulting in C scan images distort to a certain degree or fuzzy character of nugget edge, etc. It not only affects the visibility of the image, but also makes larger difficult to image feature automatic recognition, not conducive to realize computer automation, and intelligent detection. In this case, nugget edge needs testers to demarcate manually. Yet even if it is operated by engineers with some NDT experience, a large number of false inspection and leak inspection can also be found. So in order to improve the reliability of test results and to gain the real nugget size, another frequency domain signal analysis with low impact by testing conditions is taken into consideration.
Different welding current specimen time domain C scan images. (a) 9 KA, (b) 11 KA. (Online version in color.)
To make frequency domain analyses of A-echo signal of two specimens in Fig. 6 through the Fourier transform. Without regard to the main frequency amplitude, take main frequency values as the frequency domain feature signal for C scan images, which avoid the effect of testing conditions on the intensity of ultrasonic. The result is shown in Fig. 7. From the figure we can see, the circle in the characteristic nugget area is similar to the inner circle in Fig. 6. The color of the area outside the nugget is single, without the interference of parti-colored. Besides, in the figure, boundary characteristics of nugget area are clear, which can truly reflect the weld nugget appearance, measure nugget diameter accurately, and evaluate the quality of spot welding reliably.
Different welding current specimen frequency domain C scan images. (a) 9 KA, (b) 11 KA. (Online version in color.)
In the process of ultrasonic testing, the influence of many factors, such as detection system resolution, signal-to-noise ratio and materials with uneven texture, etc., can bring noise to the echo signal. In order to make the image show nugget morphology more clearly and realize the quantitative measurement of the weld nugget diameter, in this paper, C scan images in frequency domain are processed through computer image enhancement, edge detection and nugget diameter measurement. Finally to complete real-time, automatic assessment of the quality of spot welding.
4.1. Image EnhancementImage enhancement is a processing method which selectively highlights interested image features while ignores some needless features through adding some information on the original image or transforming data by certain means so as to improve image quality and strengthen the effect of image recognition. In order to filter out interference noise effectively, and protect the nugget edge information very well, a variety of filter methods are used on dealing with C scan images in this paper. After a large number of experimental comparison and analysis, we find that the effect of median filtering with nonlinear smooth characteristics on spot welding C scan image processing is best. After median filtering processing in 7 * 7 modules, the output image of Fig. 7 is shown in Fig. 8. We can see, after median filtering, it not only filters out sharp wave interference noise and makes nugget boundary more clearly in C scan images, but also further improves the quality of C scan images, which lays foundation for subsequent image edge extraction.
Frequency domain C scan images after median filtering. (a) 9 KA, (b) 11 KA. (Online version in color.)
This paper further chooses Roberts-operator, which is sensitive to noise signal, to make edge detection on the image and obtain accurate nugget boundaries. The results are as shown in Fig. 9. In order to conveniently identify and analyze nugget area in the image, morphology is used to segment the needed feature information, as shown in Fig. 10. After segmentation, not only does the nugget feature become prominent in the image, but also greatly reduces the computation amount of data and provides favorable conditions to get the real-time nugget dimension and assess spot welding quality rapidly.
Frequency domain C scan images after edge detection. (a) 9 KA, (b) 11 KA. (Online version in color.)
Frequency domain C scan images after edge segmentation. (a) 9 KA, (b) 11 KA.
Spot welding quality is characterized by the strength of joint. And the strength of the joint mainly depends on the dimension of the nugget, especially the size of nugget diameters.8) How to get connection dimension of the joint quantitatively is the key of the spot weld diameter evaluation. This paper adopts equivalent diameter method, namely calculates average nugget diameter on the spot weld joint surface to represent the welding quality of joint.
Pick up N point on the nugget area boundary. Take two points of the N points as midperpendicular. The intersection point is the center of the circle. Then centre point cluster is got. Take the average of this point cluster as the center. The average distance from the coordinate to the N boundary point is the weld nugget diameter. Figure 11 is the schematic diagram of any two midperpendicular for center of the circle.
The schematic diagram of equivalent diameter method. (Online version in color.)
C scan images of the twelve spot welding specimens with different weld quality are obtained after signal data processing. In order to evaluate the accuracy of weld nugget dimension by ultrasonic C scan image detecting, the specimens, after ultrasonic testing, are split along the combining surface of the upper and lower plate to get the nugget cross section’s metallographic detective area (as shown in Fig. 12). Measured value of specimen nugget dimension is gained. Make comparison between specimen ultrasonic detection value and metallographic measured value. The results are shown in Fig. 13(a). De refers to ultrasonic testing nugget diameter detection value; Dm refers to spot weld specimen nugget diameter measured value.
Solder joint macro metallographic photos.
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.)
From Fig. 13(a) we can see detection value De is generally larger than measured value Dm. The average relative error is 12.05%. The reasons for this phenomenon can be speculated on the existence of the corona bond. Take out the sample of the maximum error and observe the metallographic photos, as shown in Fig. 14. A round of wide corona bond is found outside the nugget. In addition, external diameter value Dc of the corona bond (corona bond plus nugget overall dimension) is close to the detection value De. In this paper, the rest of the specimens are carried on the comparative analysis. The external diameter value of spot welding corona bond is in good agreement with the ultrasonic detection value, as shown in Fig. 13(b). De refers to ultrasonic testing nugget diameter detection value; Dc refers to external diameter measured value of spot welding specimen corona bond. The test results show that the value obtained from C scan images detection based on frequency domain characteristic signal processing contain corona bond dimension. It can be used as the reference value of evaluating spot weld quality.
Maximum error specimen welding spot macro metallographic photos. (Online version in color.)
Spot welding specimens longitudinal section metallographic photos, as shown in Fig. 15. From the micrograph in the spot welding plastic ring zone in Fig. 15(b), we can see although specimen two board interface fit very close, it is still incomplete fusion area, not real nugget area. As shown in Fig. 15(c), nugget organizations are composed of columnar austenitic plus ferrite along the distribution of the columnar crystal. Developed columnar crystal grown up from nugget edge to joint surface touching each other. Phase interface does not exist. If not considering the influence of the organization on ultrasonic, we can put the nugget and the base as a unified whole theoretically. In order to evaluate precision and stability of the weld nugget size by ultrasonic C scan image detecting, under the circumstance of small plastic ring or no plastic ring, we remove the not connected lower plate. A set of standard specimen is made in this paper which simulates ideal weld nugget appearance without plastic ring at the most extent. Its nugget dimension ranges from 4.0 mm to 8.5 mm, and the interval is 0.5 mm, as shown in Fig. 16. The standard specimen avoids the possibility of damaging nuclear metal due to stretching or tearing, and reduces the measuring error. It has better accuracy and repeatability compared with joints after tensile fracture.
(a) Spot welding specimens longitudinal section metallographic photos, (b)–(c) separately represents the magnification of the corresponding position of (a).
Standard specimen. (Online version in color.)
Figure 17 is the comparison figures of standard specimen and ultrasonic test results. We can see that without the influence of the corona bond, C scan images based on frequency domain feature signal has very high accuracy on detection of weld nugget diameter. The error is less than 0.066 mm. It also indirectly proves that the error of the test method mainly comes from the corona bond. Further research on how to distinguish the nugget from corona bond to obtain more accurate nugget diameter is what we are working on at present. Of course, if some strong norms are used to make the corona bond very narrow in the process of resistance welding or the ultrasonic detection accuracy requirement is not high, the detection error by using the method is negligible.
Comparison figures of standard specimen and ultrasonic test results. (Online version in color.)
(1) In this paper, stepping mechanical scanning method is used on stainless steel spot welding joint to conduct ultrasonic C scan detection. Then signal characteristics are extracted respectively from time domain and frequency domain and weld nugget zone C scan images are obtained. After comparative analysis, we find that C scan images obtained by frequency domain characteristics signal are less affected by detecting conditions, which can reflect the weld nugget morphology more truly.
(2) Image enhancement and edge detection processing on C scan images in frequency domain can provide clear nugget boundary morphology. Not only does nugget feature become prominent in the image, but also the computation amount of data is greatly reduced. The dimension of spot weld nugget can be automatically obtained through equivalent diameter algorithm procedure.
(3) Stainless steel spot welding ultrasonic C scan value De based on frequency domain signal analysis contains corona bond dimension, and it is in good agreement with corona bond external diameter through metallographic measured value Dc. So it can be taken as the reference to evaluate the quality of spot welds. Without the influence of corona bond, detection accuracy is very high. The error is less than 0.066 mm.
