摘要
地铁转向架轴承系统出现故障将会严重影响列车的安全运行。针对于极限学习机(Extreme Learning Machine,ELM)的不足,提出基于粒子群优化(Particle Swarm Optimization,PSO)算法的自编码ELM(Auto-encoder-PSO-ELM,AP-ELM)算法。首先构建多个结构相同的自编码器,将自编码器训练好后,对每个自编码器初始化权重,然后通过PSO算法优化权重,优化完成后将其与ELM算法融合。该方法被用于西储大学公共轴承故障试验台和地铁转向架轴承系统试验台中验证,并与ELM算法和单层自编码ELM算法对比。结果表明:AP-ELM算法在2个数据集中的分类效果均优于另外2种算法,验证了AP-ELM的有效性与优越性。
The failure of metro bogie bearing system will seriously affect the safe operation of the metro.Aiming at the deficiency of Extreme Learning Machine(ELM),an auto-encoding ELM algorithm based on Particle swarm optimization(AP-ELM)was proposed.Firstly,several auto-encoders with the same structure were constructed.After the auto-encoder was trained,the weight of each auto-encoder was initialized.Then,the weight was optimized by Particle Swarm Optimization(PSO)algorithm.After the optimization,it was fused with ELM algorithm.The proposed method was used in the public bearing fault data set of Western Reserve University and the test-bed of metro bogie bearing system.Compared with ELM algorithm and single auto-encoder ELM algorithm,the results show that the classification effect of AP-ELM method in two datasets is better than the other two algorithms,which verifies the effectiveness and superiority of AP-ELM.
作者
吴宝存
WU Baocun(Tianjin Binhai New Area Rail Transit Investment Development Company,Tianjin 300450,China)
出处
《现代制造工程》
CSCD
北大核心
2023年第1期122-126,共5页
Modern Manufacturing Engineering
关键词
故障诊断
极限学习机
自编码
粒子群优化
fault diagnosis
Extreme Learning Machine(ELM)
auto-encoder
Particle Swarm Optimization(PSO)
