小波网络在振动信号时间序列预报中的应用

介绍了离散仿射小波神经网络的原理及其构建方法 ,用正交最小二乘算法进行小波基函数的选择 ,用最小二乘算法确定小波网络权值。说明小波神经网络具有高度的非线性函数逼近能力 ,可以用于建立时间序列预报模型 ,最后给出了小波网络用于建立旋转机械振动信号时间序列预报模型实例。

Application of Wavelet Networks to the Prediction of Time Series of Vibration Data

The principle and construction of discrete affine wavelet networks are discussed. In the paper, the proper wavelet and the weights in the networks are determined by the orthogonal least square (OLS) method and the least square method respectively. It is illustrated by a practical example in the vibration of rotating machinary that the wavelet networks can be used to model a time series for prediction due to its good feature in the approximation of a nonlinear function.