基于小波变换和神经网络的机械故障检测方法

A Detection Method of Mechanical Failure Based on Wavelet Transforms and Neural Network

针对机械故障中遇到的问题,提出了基于小波变换和神经网络故障检测方法,小波变换理论能够根据被分析信号的特征,自适应地选择相应频带,使之与信号频谱相匹配,从而提高了时-频分辨力。利用小波变换对信号原始信息进行分解,得到信号的不同特征向量。将不同的特征向量送入不同子神经网络进行诊断,并通过神经网络做出最后的诊断。该系统具有知识自动获取、识别速度快、鲁棒性及容错能力强等特点,实例证明该系统是有效的。

Aimed at the problems of the mechanical failure, a detection method based on wavelet transform and the nerve network is proposed. Based on analyzing the characteristic of the signal, the theory of wavelet transform can adaptively choose correspondent frequency band, make it matched with the signal spectrum, thus raised the time-frequency resolution. Using wavelet transform carry on the decomposition of the original information of signal, get different characteristic vector of Signal. Send the different characteristic vector into the different sub-nerve network carries through the diagnosis. The nerve network make out a ultimate diagnosis. This system has characteristics of automatic knowledge obtain, quick identify, robustness and strong fault tolerance etc. The practical instance indicates that the system is valid.