论文部分内容阅读
为提高开关电流电路故障诊断的精度,提出了一种基于小波包优选和优化BP神经网路的开关电流电路特征抽取与识别方法.首先对开关电流电路原始响应信号进行多层次的小波包分解,接着计算N层分解后的归一化能量值,以特征偏离度作为评价选择最优小波包基,构建最优故障特征向量,最后将提取的最优故障特征通过遗传算法优化的BP神经网络进行分类.该方法以实例电路进行验证,结果表明所有的软故障均得到了有效的分类,说明了该方法在开关电流电路故障诊断中的优越性.
In order to improve the accuracy of circuit fault diagnosis of switched current circuits, a method of extracting and identifying the characteristics of switched current circuits based on wavelet packet optimization and BP neural network optimization is proposed.Firstly, wavelet packet decomposition of the original response signals of switched current circuits is carried out, Then, the normalized energy value after decomposition of N layer is calculated, and the optimal wavelet packet basis is selected based on the characteristic deviation degree, and the optimal fault feature vector is constructed. Finally, the optimal fault feature extracted is processed by BP neural network optimized by genetic algorithm The method is validated by an example circuit, and the results show that all soft faults are effectively classified, which shows the superiority of this method in fault diagnosis of switched current circuits.