Research on detection method of micro cracks in mechanical welding based on Bayesian inference

Abstract

In view of the typical defects of the welding structure, the detection method of micro cracks in mechanical welding based on Bayesian inference is proposed. According to the micro crack detection signal of mechanical welding defects, combined with Bayesian inference, PCA and common classification algorithm, the micro crack features of mechanical welding are extracted. On the basis of this, denoise the micro crack detection signal and extract useful information. The processed signals are input into SVM classifier, and the penalty factor and RBF kernel function are optimized according to the process of simulating colony foraging. The global optimal value for crack detection is selected. The experimental results show that the detection accuracy of the research method is high, which fully meets the research requirements.