Ultrasonic Signal Characteristic Analysis and Intelligent Defect Identification of Thin-wall Small-diameter Column/Plate Diffusion Welding Interface

To address the problem of aliasing in amplitude signals from the defects and interface during ultrasonic inspection of thin-wall and small-diameter column/plate diffusion welding quality, which complicates the determination of small defects in the welded joint, a support vector machine based on particle swarm optimization (PSO-SVM) was used to identify the defects at the diffusion welding interface with multi-feature parameters of different interface types as input. Firstly, the C-scan data from the sample was collected using a water immersion ultrasonic detection system. Taking the welding cross-section obtained from metallographic testing as a reference, the time and frequency domain characteristic values for three types of interfaces, i.e. defect-free, weld bead, and lack of fusion were extracted by using fast Fourier transform and empirical mode decomposition methods. Afterwards, principal component analysis (PCA) was used to integrate the multi-feature parameters to obtain the fusion eigenvalues. Finally, the defects were inputed into the PSO-SVM model for intelligent identification, and the prediction results were compared with those without extensive feature fusion. The results show that after principal component analysis processing, the recognition accuracy of the three types of interfaces in the test results is 100%, demonstrating 4.5% improvement over the accuracy of the test results without PCA processing.