故障诊断中基于神经网络的特征提取方法研究

在电路状态检测与故障诊断过程中,恰当地选择特征参数是诊断成败的关键。本文研究了基于神经网络的特征评价和特征提取方法,利用神经网络的训练结果对特征参数进行合理的评价。由于神经网络满足高分辨率信息压缩所需的非线性映射条件,通过特征提取将电路故障模式识别中复杂的分类问题转移到特征处理阶段,利用神经网络有效地实现了特征参数的提取。诊断实例验证了该方法的有效性。

Feature extraction method in fault diagnosis based on neural network

In circuit state detection and fault diagnosis, choosing proper feature parameters is vital to diagnosis. The feature estimation and extraction methods are presented; the results trained by neural network are used to evaluate feature parameters in reason. Because neural network satisfies the nonlinear mapping requirement for high-resolution information compression, the complex classification problem in circuit fault pattern recognition is transferred to feature processing stage, and feature extraction is realized by neural network effectively. An illustration validates this method.