摘要
Abstract
Fault diagnosis of axial piston pumps based on convolutional neural networks can eliminate the problems of manual design and extraction of signal features in traditional fault diagnosis and imperfect signal feature extraction.Vibration signals were collected in five working states of the axial piston pump,including normal state,loose boot,worn slider,worn distributor plate,and effective center spring,as data samples for fault detection.The original signals were then input into the CNN model.The D-1DCNN was used for fault diagnosis,which could directly input the original signals,set network model parameters,and optimize network parameters.Through experiments,it was found that D-1DCNN has strong performance in terms of training time and accuracy,and that the fault diagnosis accuracy can reach 100%,and the training time is 126 seconds.关键词
卷积神经网络/轴向柱塞泵/故障诊断/模型/信号采集Key words
convolutional neural network/axial piston pump/fault diagnosis/model/signal acquisition分类
机械制造
