LCD模糊熵和SOM神经网络在液压泵故障诊断中的应用

针对液压泵故障诊断问题,提出了一种基于局部特征尺度分解(Local Characteristic-scale Decomposition,LCD)、模糊熵和SOM神经网络三者相结合的故障诊断方法。对液压泵振动信号进行LCD分解,得到若干个内禀尺度分量(Intrinsic Scale Component,ISC);将ISC分量分别与原信号进行相关分析,筛选出包含主要故障信息的前几个ISC分量,计算其模糊熵并组成特征矩阵;将特征矩阵输入SOM神经网络进行分类识别。液压泵故障诊断实例表明,该方法能够准确识别液压泵典型故障,具有一定优势。通过与BP神经网络分类结果相对比,显示了SOM神经网络在特征分类方面的优越性。

Application of LCD Fuzzy Entropy and SOM Neural Network in Fault Diagnosis of Hydraulic Pump

Aiming at the fault diagnosis problem of hydraulic pump, a method that based on the combination of Local characteris-tic-scale decomposition ( LCD) , Fuzzy Entropy and SOM Neural Network was proposed. Firstly, the vibration signal of hydraulic pump was decomposed with LCD into several Intrinsic scale components (ISC). Secondly, the first few ISC components, which contained main fault information, were chosen by correlation coefficient analysis with the original signal and then the fuzzy entropy of these com-ponents was calculated to compose characteristic matrix. Finally, the characteristic entropy matrix used to train the SOM neural network to recognize different fault types was ensured. The experiment results of fault diagnosis of hydraulic pump show that the method can classify typical fault types of hydraulic pump exactly, and has certain superiority. By comparing with the classfied results of BP neural network, the superiority of the SOM neural network in fault feature classification is confirmed.