为提升轴承故障诊断的准确度及信号的处理能力，针对小波分析在轴承局部损伤诊断中的紧支性特点，研究了正交紧支撑函数的构建方法，以小波多分辨分析法和多元时间序列分析为基础，构建出小波神经网络，并对模型中的权值与缩放、平移数进行了修正。基于MATLAB对故障轴承的非平稳信号加速度波形进行了数值模拟计算，通过对神经网络参数修正前后的收敛性对比可知，修正后的迭代误差在幅值和收敛效率上均优于修正前，明显提升了轴承故障诊断的准确度及信号的处理能力。%In order to improve the accuracy of bearing fault diagnosis and the signal processing,aiming at the com⁃pactly-supported characteristics of wavelet analysis in the local damage diagnosis of bearing, the establishment method of the orthogonal compactly⁃supported function was studied. With the wavelet multi⁃resolution analysis and multivariate time series analysis as the basis, a wavelet nerve network was established;in addition, the parameters of weights, scaling and translation in the model were modified. Based on MATLAB, a numerically simulated calcu⁃lation was carried out for the non⁃stationary signal acceleration waveform of the failed bearing. By comparing the convergence attained before and after the parameter revision, it shows that the iteration error after revision is superi⁃or to the value before revision whether on the aspect of amplitude or convergence efficiency, and that the accuracy of fault diagnosis of bearing and signal processing capacity are significantly improved.
展开▼
