摘要
为了提高纯电动汽车电驱总成的故障诊断准确率,提出了一种基于粒子群优化(Particle Swarm Optimizing,PSO)算法的改进BP(Improved Back Propagation,IBP)神经网络(PSO-IBP)故障诊断方法。应用线性整流单元(Rectified Linear Unit,ReLU)作为BP神经网络的激活函数,通过粒子群优化算法对BP神经网络权值和阈值进行动态寻优,构建PSO-IBP模型。通过采集纯电动汽车电驱总成故障数据,分别对PSO-IBP神经网络模型、BP神经网络模型和概率神经网络(Probabilistic Neural Network,PNN)模型进行训练与仿真,结果表明,相比于BP神经网络方法及概率神经网络方法,基于PSO-IBP神经网络模型的纯电动汽车电驱总成故障诊断方法具有更高的准确率。
In order to improve the accuracy of fault diagnosis for the electric drive assembly of pure electric vehicles,a fault diagnosis method based on Particle Swarm Optimizing(PSO)algorithm was proposed to optimize the Improved Back Propagation(IBP)neural network.The Rectified Linear Unit(ReLU)was used as the activation function for the BP neural network.Through the Particle Swarm Optimizing algorithm,the weights and thresholds of the BP neural network were dynamically optimized to build the PSO-IBP model.By collecting fault data from the electric drive assembly of pure electric vehicles,PSO-IBP model,along with the BP neural network model and the Probabilistic Neural Network(PNN)model,were trained and simulated.The results showed that compared to the BP neural network methods and PNN methods,fault diagnosis method for pure electric vehicle electric drive assembly based on PSO-IBP neural network model has higher accuracy.
作者
肖伟
李泽军
管天福
贺路
陈绪兵
XIAO Wei;LI Zejun;GUAN Tianfu;HE Lu;CHEN Xubing(School of Mechanical&Electrical Engineering,Wuhan Institute of Technology,Wuhan 430205,China;College of Automobile and Electromechanical,Hubei Land Resources Vocational College,Wuhan 430090,China)
出处
《现代制造工程》
CSCD
北大核心
2024年第1期137-141,共5页
Modern Manufacturing Engineering
基金
国家自然科学基金项目(51875415)
中国电子劳动学会“产教融合、校企合作”教育改革发展课题项目(Ciel2022139)。
关键词
纯电动汽车
粒子群算法
BP神经网络
故障诊断
pure electric vehicle
Particle Swarm Optimizing algorithm
BP neural network
fault diagnosis