基于PSO-LSSVM的疲劳裂纹漏磁定量识别技术

To solve the problem of fatigue cracks quantitative identification, a modeling method combining principal component analysis(PCA)and particle swarm optimization least squares support vector machine(PSO-LSSVM)was proposed to establish a nonlinear mapping relationship between magnetic flux leakage signals and fatigue cracks for quantitative identification of the fatigue crack width and depth. Firstly, a magnetic flux leakage detection system was built, and a series of fatigue crack samples were prepared by fatigue tensile test. Then, the quantitative identification experiments of fatigue crack magnetic flux were carried out to establish a magnetic flux leakage defect sample library. Finally, the feasibility of the quantitative identification method of fatigue crack magnetic flux leakage based on PSO-LSSVM was verified. The results show that the method can effectively identify the width and depth of fatigue cracks with a size less than 1 mm, and the error is about 0.1 mm.