基于LMD-PE与神经网络的刀具故障诊断方法

Tool fault diagnosis method based on LMD-PE and neural network

摘要:

针对刀具故障诊断信号信噪比低、诊断结果不准确等问题,采用局域均值分解(LMD)结合排列熵(PE)来处理采集到的刀具加工时的振动信号,然后将提取到的特征向量输入到训练好的长短期记忆神经网络(LSTM)中得到诊断结果,为了提高LSTM的诊断效率,结合卷积神经网络(CNN)对LSTM进行了改造。试验表明,本文提出的方法诊断准确率比BP神经网络提高了将近12%,改进LSTM网络比传统LSTM的诊断时间缩短了50%。

Abstract:

Aiming at the problems of low signal-to-noise ratio of tool fault diagnosis signals and inaccurate diagnosis results, local mean value decomposition (LMD) combined with permutation entropy (PE) is used to process the collected vibration signals during tool processing, and then the extracted feature vectors Input to the trained long and short-term memory neural network (LSTM) to get the diagnosis result. In order to improve the diagnosis efficiency of LSTM, combined with convolutional neural network (CNN) to transform LSTM. Experiments show that the diagnostic accuracy of the method proposed in this paper is nearly 12% higher than that of the BP neural network, and the improved LSTM network reduces the diagnostic time of the traditional LSTM by 50%.