An improved support vector machine based on particle swarm optimization in laser ultrasonic defect detection

Laser ultrasonic defect detection and classification has been widely used in engineering and material defect detection, so detecting and classifying the defect targets accurately is significant. In order to obtain the higher classification accuracy, an improved support vector machine (SVM) based on particle swarm optimization algorithm is used as classifier in this paper. To search the optimal parameters of SVM, a new Tangent Decreasing Inertia Weight strategy particle swarm optimization (TPSO) algorithm is proposed to determine the optimal parameters for SVM. In addition, to further improve the classification accuracy, sparse representation is used to extract the target features from the real target echo waveform in experiment. Experimental results show that the proposed TPSO-SVM can achieve higher classification accuracy compared to the commonly PSO-SVM, classical SVM and BP neural network (BPNN) in the laser ultrasonic defect signals classification. Key-words: Laser ultrasonic defect detection; Classification method; Support vector machine; Particle swarm optimization; Sparse representation.