基于CV-ELM方法的机电作动器故障预测研究  

Research on Fault Prognosis of Electromechanical Actuator Based on CV-ELM Method

作　　者：曹宇燕 吕永玺 王新民[1] CAO Yu-yan;LV Yong-xi;WANG Xin-min(School of Automation,Northwestern Polytechnical University,Xi'an Shanxi 710129,China)

机构地区：[1]西北工业大学自动化学院,陕西西安710129

出　　处：《计算机仿真》2022年第8期432-436,共5页Computer Simulation

基　　金：国家自然科学基金(61703341);陕西省自然科学基金(2020JQ-218);航空科学基金(20180753006)。

摘　　要：为提高机电作动器的可维护性,提出了一种基于交叉验证-极限学习机(CV-ELM)的机电作动器故障预测方法。首先通过分析机电作动器模型,确定了轴承寿命是影响机动作动器寿命的重要因素。其次,针对传统ELM方法中参数随机选择导致的精度不高的问题,将交叉验证方法与ELM结合,并证明了改进ELM方法的有效性,给出了基于CV-ELM方法的故障预测流程;最后,基于轴承全寿命试验数据,通过与反向传播神经网络(BPNN)方法和支持向量回归(SVR)方法的对比仿真,验证了本文所提方法的快速性和有效性,为机电作动器的维护提供了数据参考和技术支撑。In order to improve the maintainability of electromechanical actuators,a fault prognosis method based on cross-validation extreme learning machine(CV-ELM)is proposed.Firstly,through the analysis of the electromechanical actuator model,the remaining useful life of the bearing was regarded as an important factor affecting the life of the actuator.Secondly,aiming at the problem of low accuracy caused by the random selection of parameters in the traditional ELM method,the cross-validation method was combined with ELM.The effectiveness of the improved ELM method was proved and the fault prognosis process based on CV-ELM was given.Finally,the validation and effectiveness of the proposed method were verified by simulations in comparison with the backpropagation neural network(BPNN)method and support vector regression(SVR)method on the basis of the bearing life test data.The CVELM method will provide data reference and technical support for the maintenance of the electromechanical actuator.

关 键 词：机电作动器 故障预测 极限学习机 交叉验证 

分 类 号：U226.8[交通运输工程—道路与铁道工程]