Defect Length Prediction of Aluminum Alloy Sheet by Using Differential Evolution-Support Vector Regression (DE-SVR)

Aluminum alloy sheet has been widely applied in transport manufacturing industry due to its good mechanical properties. However, aluminum alloy sheet can inevitably generate defects in the process of processing and forming. In this paper, through the Support Vector Regression (SVR) model optimized by Differential Evolution (DE) algorithm, the collected Lamb wave signal in aluminum alloy sheet is analyzed and processed to detect the defect length in aluminum alloy sheet. The error penalty parameter $C$ and kernel function $g$ of SVR can be optimized constantly by using the selection, crossover, mutation operators and greedy selection strategies of DE algorithm. The feature matrix of Lamb wave signal is extracted and introduced into Particle Swarm Optimization-Support Vector Regression (PSO-SVR), Genetic Algorithm-Support Vector Regression (GA-SVR) and DE-SVR to compare and analyze the defect length error evaluation indexes. The results show that DE-SVR can greatly improve the speed and accuracy of defect length prediction of aluminum alloy sheet.