摘要
电感式磨粒检测方法采用电磁感应原理,能够实时地检测机械设备所产生的金属磨粒,是目前具有潜力的在线磨损监测方法;然而,多个磨粒同时通过传感器引起的信号混叠是造成检测误差的一个重要原因。提出一种基于神经网络的混叠误差修正方法,通过采用单个磨粒的实际信号构造混叠样本,对神经网络进行训练,从而获得磨粒信号的混叠模型,并且能够解决传感器差异所造成的方法适应性问题。最后,采用正弦波模拟磨粒信号验证了方法的有效性。
Inductive oil debris detection can timely monitor the generation of wear debris particles of mechanical system through electromagnetic induction,which is a potential online method.However,the superimposition of induced voltages of multiple particles seriously limit the detection's accuracy.A correction method based on artificial neural network for the aliasing error is proposed.The superimposition samples are constructed by using an actual waveform to train the network,which can overcome the modelling difficulty and adaptability problem.Finally,the effectiveness is validated by a simulation testing with sine waveforms.
作者
李彤阳
洪葳
LI Tong-yang;HONG Wei(China Civil Aviation Engineering Consulting Co.,Ltd.,Beijing 100621;School of Physics,Huazhong University of Science and Technology,Wuhan,Hubei 430074)
出处
《液压与气动》
北大核心
2021年第5期57-61,共5页
Chinese Hydraulics & Pneumatics
基金
国家自然科学基金(51905187)
关键词
油液磨粒监测
混叠误差
人工神经网络
oil debris monitoring
aliasing error
artificial neural network
