基于PSO-SVM的点焊接头拉剪强度分类分析

Study on classification of tensile shear strength of spot welding joints based on PSO-SVM

点焊是汽车零部件的主要连接方式之一,点焊接头的拉剪强度是评价点焊质量的重要因素,论文在制备大量点焊试样的基础上,对各试样的焊点进行超声信号检测,并运用信号处理获得时域、频域和小波包特征值,随后对点焊试样在拉剪试验中的失效形式进行分析,建立点焊接头拉剪强度的分级标准.根据试验数据设计了BP(back-propagation)神经网络和基于粒子群优化支持向量机(particle swarm optimization support vector machine,PSO-SVM)的神经网络分类器,最后将试样的超声信号特征值作为输入参数,比较两种神经网络模型对点焊试样拉剪强度分类的准确率.试验结果表明,结合9个超声信号特征值的PSO-SVM神经网络具有最高的点焊强度分类准确率.

Spot welding is one of the main connection methods of automobile parts,and the tensile shear strength of spot welded joints is the most important factor to evaluate the quality of spot welding.In this paper,based on the preparation of a large number of spot welding samples,ultrasonic signal detection is carried out on the spot welding points of each sample,and the time-domain,frequency-domain and wavelet packet eigenvalues are obtained by using signal processing methods.Then,by analyzing the failure form of spot welding specimen in tension shear test,the grading standard of tensile shear strength of spot welding joint is established.BP neural network and neural network classifier based on PSO-SVM are designed according to the test data.Finally,the ultrasonic signal eigenvalue of the sample is used as the input parameter to compare the accuracy of the two neural network models for the classification of tensile shear strength of spot welding samples.The experimental results show that PSO-SVM neural network combined with 9 ultrasonic signal eigenvalues has the highest accuracy of spot welding strength classification.