Evaluation of quantitative ultrasonic C-scan testing for refill friction stir spot welding joints based on time-frequency analysis

This paper focuses on a comparative study of different ultrasonic feature-based C-scan testing imaging methods for characterizing refill friction stir spot welding (RFSSW) joints, and accurately identifying and measuring the nugget boundary. The aim is to explore a suitable method for nugget characterization and automatic size analysis. The research indicates that the frequency-domain C-scan imaging method outperforms the time-domain C-scan imaging method in accurately characterizing the nugget. Furthermore, the C-scan imaging method based on the feature value of the base material zone (BMZ) echo signal performs better than the method based on the nugget echo signal. The tested nugget sizes obtained by using the –6 dB drop-off method to identify the boundaries in the C-scan images are generally larger than the metallographic measurement values. A novel method is proposed in this paper, using the difference in main frequency amplitude values between the BMZ and the nugget as the feature value for the C-scan testing image, and employing the Hough circle transformation for an automatic extraction of joint size. This method achieves improved nugget characterization and higher accuracy in size analysis.


Introduction
Aluminum alloys have become a crucial material for making lightweight vehicle structures because of their comprehensive performance [1,2].However, the weldability of aluminum alloys is limited because of their material characteristics, such as large thermal deformation, high thermal conductivity, and susceptibility to oxidation [3,4].Refill friction stir spot welding (RFSSW), a low-temperature solid-state welding technique, has become an important welding method for joining aluminum alloys [5][6][7].It offers unique welding characteristics, including a small heat-affected zone, minimal welding deformation, and low welding residual stress, making it suitable for lap welding thin aluminum sheets [8,9].As the RFSSW process is complex and influenced by numerous parameters that affect joint quality, evaluating the quality of RFSSW joints using nondestructive testing has become an interesting topic in the academic community [8][9][10][11][12].In recent years, ultrasonic nondestructive testing methods have gained significant attention for evaluating joint quality due to their portability, reliability, safety, and other advantages [13][14][15][16][17].Although most joints are precisely inspected using ultrasonic techniques with high testing requirements in the laboratory after flattening, industrial applications still rely mainly on manual inspection and analysis, lacking convenient ultrasonic testing technology and devices that can provide automated imaging of testing results and quantitative analysis [17][18][19][20].This study aims to address the challenges associated with complex ultrasonic signals from RFSSW joints, the difficulty in identifying features of the base material zone (BMZ) and the nugget, the difficulty in quantitatively assessing the joint connection state, and the low degree of automation.To overcome these challenges, this study explores an ultrasonic C-scan testing imaging method based on an analysis of ultrasonic time-domain and frequency-domain characteristics.The proposed method uses the difference in the main frequency amplitude values between the BMZ and the nugget along with the application of Hough circle transformation.This method can realize the accurate identification of the BMZ and the nugget to achieve an automated quantitative analysis of the joint connection state.

Welded material and specimen preparation
The welded material used in this study is EN AW-5083 aluminum alloy, which is used in the manufacture of vehicles and rail applications.The chemical composition is shown in table 1.The thickness of the sheets is 2 mm, the width is 50 mm, and the length is 150 mm.Two sheets were lap welded with a width of 50 mm and the welded joint was located in the center of the overlap zone.The RFSSW specimens were welded by using an HT-FSW500MT RFSSW machine.The tool system consisted of an 18 mm diameter clamping ring, a 9 mm diameter sleeve, and a 5 mm diameter pin.The main welding parameters are shown in table 2, and the RFSSW process is illustrated in figure 1 [9][10][11][12].Figure 2 shows an actual photo of the RFSSW specimen.Figure 2(a) shows the welded surface of the RFSSW specimen, clearly showing the processing traces of the joint.Figure 2(b) shows the relatively flat bottom surface of the specimen.To ensure good coupling between the ultrasonic probe and the specimens during ultrasonic testing, the bottom surface of the specimen shown in figure 2(b) was selected as the surface for ultrasonic testing.

Ultrasonic testing device and C-scan testing imaging method
In this study, the physical picture of the ultrasonic C-scan testing device is shown in figure 3(a), ultrasonic testing was performed using an S-type mechanical scanning pattern on an ultrasonic testing surface, as illustrated in figure 3(b).The ultrasonic testing device was an Olympus V-260SM focusing probe with a main frequency of 15 MHz and an ultrasonic signal acquisition card with a sampling rate of 100 MHz.The couplant is water.The testing zone was set to 20 mm × 20 mm, with a scanning step of 80 μm in both the X and Y directions.The ultrasonic probe performed ultrasonic testing at each mechanical scanning step node, and the ultrasonic C-scan testing image was generated as pixels by the characteristic values of the ultrasonic signals at each scanning step node.After ultrasonic testing, the specimens were cut along the cutting line, which corresponds to the axis of the  RFSSW joint, as shown in figure 3(b).Metallographic specimens were prepared to visualize the morphology of the RFSSW joint and the regional distribution of aluminum element content.

Results and discussion
3.1.Analysis of the RFSSW joint metallograph Figure 4(a) demonstrates a representative axial cross-sectional metallograph of the RFSSW specimen, in which zone A is the BMZ, zone B is the nugget, and zone C is the boundary zone including the heat-affected and thermomechanical affected zone [10][11][12].4(a) which is the metallographic measurement value of the joint's diameter, is determined to be 8.89 mm.This bowl-shaped characteristic, unique to RFSSW joints, is known as the bonding ligament and represents the remnant oxide coating from the faying surface [12,21,22].The analysis of the aluminum element content concludes that the bonding ligament also achieves an effective connection.

Ultrasonic time-domain and frequency-domain analysis of the RFSSW joint
The ultrasonic C-scan testing imaging method of the RFSSW joint is based on the principle that the ultrasonic testing signal will present different echo characteristics in different zones of the RFSSW joint.Figure 5 shows a schematic diagram of typical ultrasonic propagation and reflection states in different testing zones during the ultrasonic C-scan testing of the RFSSW joint.As shown in figure 5, when the ultrasound probe is positioned in the BMZ at position 1, the ultrasound entering the upper sheet is completely reflected at the faying surface.It  To assess the connection state of the RFSSW joint based on the characteristics of the ultrasonic A-echo signal, the sonic path distances of ultrasounds that propagate to the faying surface and the bottom surface of the specimen are respectively considered as target sonic path distances.The reflected ultrasonic amplitude values at these target sonic path distances are used as the characteristic values to construct characteristic curves along the diameter of the RFSSW joint.The characteristic curves are shown in figure 7, with the horizontal axis representing the scanning step nodes.Figure 7(a) shows the characteristic curve of the echo amplitude on the faying surface.The characteristic curve exhibits a noticeable transition in the boundary zone, which effectively distinguishes between the BMZ and the nugget.However, figure 7(b) shows that the characteristic curve of the echo amplitude on the bottom surface of the specimen is more complicated.The fluctuation of the characteristic curve is large due to the roughness of the bottom surface in the nugget.Additionally, the features of the characteristic curve in the boundary zone are difficult to discern due to interference from the second echo on the faying surface, so the amplitude at the sonic path distance of the specimen's bottom surface is not suitable for characterizing the connection state of the RFSSW joint.To accurately analyze the connection state of the RFSSW joint, the echo amplitude at the sonic path distance of the faying surface is utilized to distinguish between the BMZ and the nugget, and this study uses the -6 dB drop-off method to identify the boundary between the BMZ and the nugget.
To accurately determine the boundary between the BMZ and the nugget, the ultrasonic A-echo signals in figure 6        The C-scan testing images in figure 12 were processed using a Hough circle transformation [23,24] to automatically extract the boundary sizes of the nugget.This was done to compare the accuracy of boundary characterization using different characteristic values.The results of this processing are shown in figure 13.This figure demonstrates that the boundary of the nugget under the Hough circle transformation in figure 13(d) has a diameter of 8.96 mm, which is the closest to the metallographic measured value of 8.89 mm, followed by the analysis result in figure 13(b).The best C-scan testing imaging method is based on the main frequency amplitude difference between the BMZ and the nugget, and the boundary of the nugget is identified with higher accuracy.
This study conducted a verification experiment using 12 groups of RFSSW specimens to compare the ultrasonic testing results of the nugget diameters obtained through various ultrasonic C-scan testing imaging methods with the metallographic measured values.Figure 14 shows the comparison result between the metallographic measurement values and the tested values based on the C-scan testing images composed of characteristic values from ultrasonic time-domain echo signal amplitude values of the faying surface and the -6 dB drop-off method.Figure 15 shows the comparison result between the metallographic measurement values and the tested values based on the C-scan testing images composed of characteristic values from the main frequency amplitude values of the BMZ and the -6 dB drop-off method.Figure 16 shows the comparison result between the metallographic measurement values and the tested values based on the C-scan testing images composed of the main frequency amplitude differences between the BMZ and the nugget and the criterion A = 0 in equation (1).The average absolute errors are 1.464 mm, 1.312 mm, and 0.382 mm, with corresponding variances of 0.815 mm 2 , 0.748 mm 2 , and 0.241 mm 2 , respectively.The experimental results demonstrate that the testing error based on the analysis of the ultrasonic time-domain signal is significant and unstable compared to the error obtained through the analysis of the ultrasonic frequency-domain signal, which is more stable.Among the testing methods, the method that uses the main frequency amplitude difference between the BMZ and the     nugget as the characteristic value has the lowest error and the highest stability, making it the most reliable method for accurately determining the nugget diameter.

Conclusion
(1) An analysis of the ultrasonic testing signal in both the time-domain and frequency-domain shows that extracting the ultrasonic echo signal characteristic value based on the sonic path distance of the faying surface in the BMZ can effectively avoid the interference of the echo noise, which is formed by the rough bottom surface and secondary echoes from the faying surface.The method of extracting the characteristic value based on the frequency-domain echo signal is more resistant to the interference of echo noise.
(2) The C-scan testing images based on the characteristic values of the time-domain and frequency-domain ultrasonic echo signals from the faying surface in the BMZ can effectively characterize the nugget of the RFSSW joint.However, the boundaries appear blurred in these images.In contrast, the ultrasonic C-scan testing imaging method proposed based on the main frequency amplitude difference between the BMZ and the nugget as characteristic value, has a clear pixel binarization in the BMZ and the nugget, resulting in a precise characterization of the nugget boundary.
(3) The testing results demonstrate that the boundary identification method using the C-scan testing image based on the characteristic value extracted from the ultrasonic frequency-domain echo signal, along with the -6 dB drop-off method, achieves higher accuracy compared to the time-domain ultrasonic echo signal.Furthermore, the method based on the main frequency amplitude difference between the BMZ and the nugget, with characteristic values of zero as the boundary, shows the highest level of accuracy in boundary identification.

Figure 1 .
Figure 1.The RFSSW process: (a) compression and tool rotation, (b) sleeve compression and pin retraction, (c) sleeve retraction and pin compression, and (d) unloading.

Figure 4 (
a) shows that the BMZ subjected to the clamping ring still has the original faying surface and that the faying surface in the nugget and the boundary zone is changed to a distinctive bowl-shaped characteristic resulting from stirring and refilling processes.To evaluate the actual connection state of the joint, the aluminum element content distribution was tested in the different zones of the RFSSW joint, as shown in figures 4(b)-(d).The aluminum element contents along line 1 in the BMZ (figure 4(b)) and line 2 in the boundary zone (figure 4(c)) shown in figures 4(e) and (f) respectively are significantly reduced, indicating an insufficient connection in these zones.The aluminum element content along line 3 in the boundary zone (figure 4(c)) shown in figure 4(g) exhibits only a slight decrease, indicating that a significant degree of connection has been formed in this zone.The aluminum element content along the bowl-shaped

Figure 2 .
Figure 2. Photos of the RFSSW specimen: (a) top surface of the specimen and (b) bottom surface of the specimen.

Figure 3 .
Figure 3.The schematic diagram of ultrasonic testing (a) ultrasonic C-scan testing device and (b) schematic diagram of the probe scanning path.

Figure 4 .
Figure 4. Metallograph and aluminum distribution diagrams of a typical RFSSW joint: (a) macroscopy of the joint, (b)-(d) separately represent the magnification of the different zones in (a): (b) the BMZ, (c) the boundary zone, and (d) the nugget zone, (e)-(h) corresponding to the aluminum element content diagrams of Line 1, Line 2, Line 3, Line 4 respectively.

Figure 5 .
Figure 5. Schematic diagram of the ultrasonic propagation paths in different zones of the RFSSW joint.
were converted into envelope detection signals, as shown in figure 8. Then spectrograms of the envelope detection signals were generated through the Fourier transform, as shown in figure 9.In these spectrograms, clear main frequency features and different main frequency values are observed in both the BMZ and the nugget.However, the boundary zone lacks a clear main frequency feature.Characteristic curves are generated along the diameter direction of the RFSSW joint using the main frequency amplitude values of the BMZ and the nugget, as shown in figures 10(a) and (b) respectively.
Figure 10(a) demonstrates that the main frequency amplitude value of the BMZ decreases from high to low through the boundary zone until it reaches its lowest value in the nugget.Conversely, figure 10(b) demonstrates that the main frequency amplitude value of the nugget decreases from high to low through the boundary zone until it reaches its lowest value in the BMZ.The characteristic curves of

Figure 6 .Figure 7 .
Figure 6.Ultrasonic A-echo signal diagrams acquired in different zones of the RFSSW joint: (a) the BMZ, (b) the boundary zone, and (c) the nugget.

Figure 8 .
Figure 8. Ultrasonic envelope detection signal diagrams acquired in different zones of the RFSSW joint: (a) the BMZ, (b) the boundary zone, and (c) the nugget.

Figure 9 .
Figure 9. Spectrograms acquired in different zones of the RFSSW joint: (a) the BMZ, (b) the boundary zone, and (c) the nugget.

Figure 10 .
Figure 10.Spectrum characteristic curves along the diameter of the joint: (a) the main frequency of the BMZ, (b) the main frequency of the nugget, and (c) overlay of (a) and (b).

Figure 11 .
Figure 11.C-scan testing images obtained by different ultrasonic characteristic value methods: (a) characteristic value in the time domain, (b) main frequency amplitude of the BMZ, (c) main frequency amplitude of the nugget, and (d) the difference between the main frequency amplitude values of the BMZ and the nugget.

Figure 12 .
Figure 12.Boundaries of the RFSSW joints: (a)-(c) the -6 dB drop-off method of figures 11(a)-(c), and (d) the boundary where the difference between the ultrasonic characteristic values of the BMZ and the nugget in the spectrum is zero.

Figure 13 .
Figure 13.The Hough circle transformation extracts the diameters from the boundaries of the RFSSW joints: (a)-(d) are the corresponding results for figures 12(a)-(d).

Figure 14 .
Figure 14.Diameters tested by the C-scan testing imaging method of the characteristic value of ultrasonic time-domain echo signal amplitude from the faying surface: (a) comparison between the actual and tested values, and (b) error statistics.

Figure 15 .
Figure 15.Diameters tested by the C-scan testing imaging method of the main frequency amplitude value of the BMZ: (a) comparison between the actual and tested values, and (b) error statistics.

Figure 16 .
Figure 16.Diameters tested by the C-scan testing imaging method of the main frequency amplitude difference between the BMZ and the nugget: (a) comparison between the actual and tested values, and (b) error statistics.