论文部分内容阅读
提出了一种粒子群算法(PSO)优化支持向量机(SVM)的方法,对焊接转子环焊缝的超声回波信号进行缺陷识别.对消噪后的超声回波缺陷信号进行4层小波包分解及结点重构,提取结点重构信号中近似部分的波峰系数和波形系数,并与细节部分的积分超声值、有效值和绝对值方差组成样本的特征向量;采用PSO算法对SVM的惩罚因子和核函数参数进行优化选择,最后完成缺陷识别.结果表明:PSO-SVM模型对预测样本具有很好的识别效果,与其他常用的SVM模型相比,PSO-SVM模型无论是识别率还是识别时间上都具有良好的效果.
A particle swarm optimization (PSO) optimization support vector machine (SVM) method is proposed to identify the defects of ultrasonic echo signals of welded rotor girth welds. The four-layer wavelet packet Decomposition and node reconstruction, extract the peak coefficients and waveform coefficients of the approximate part of the reconstructed signal, and compose the feature vectors of the samples with the integrated ultrasonic, RMS and absolute variance of the detail part; Penalty factor and kernel function parameter.Finally, the defect identification is completed.The results show that the PSO-SVM model has a good recognition effect on the predicted samples.Compared with other commonly used SVM models, the PSO-SVM model has no effect on the recognition rate Recognition time has a good effect.