模拟电路故障特征提取的小波基选取方法研究

Selection Method of Wavelet Bases on Feature Extraction for Analog Circuit Fault Data

小波技术在高维故障特征数据的压缩及敏感信号提取已被广泛应用,但小波基的选取没有一个统一的标准;通过实际采样信号数据的小波分解、特征向量计算、波动性函数比较等技术对小波基函数的选取进行了研究;最后通过综合小波分析、神经网络等技术的模拟电路故障诊断系统的诊断实例验证了所提选取方法的有效性;使用9种常用小波基函数,分别对采样信号进行分解并计算波动性函数,并在模拟电路故障诊断系统进行验证;小波基函数bior2.2的波动较小且与诊断结果一致。

Wavelet has been widely used in compressing dimension and extracting effective and sensitive feature in high--amount sample data of analog circuit fault diagnosis. However, there is no standard to select wavelet bases. A selection method combined wavelet decompo- sition, eigenvector computation and volatility function was proposed and the effectiveness was verified by a diagnosis example in analog circuit fault diagnosis system which integrated wavelet, BP neural networks. Taken 9 commonly used wavelet function to decompose of the sampled signal and calculate the volatility. The volatility of wavelet functions bior2. 2 is rain and it＇ s consistent with the diagnosis.