首页|期刊导航|电子科技|基于CEEMD-APSO-LSTM的步进电机故障诊断

基于CEEMD-APSO-LSTM的步进电机故障诊断OA

Stepper Motor Fault Diagnosis Based on CEEMD-APSO-LSTM

中文摘要英文摘要

针对步进电机故障振动信号中噪声等无关数据干扰导致的故障诊断准确率降低问题,文中提出了一种基于互补集合经验模态分解(Complementary Ensemble Empirical Mode Decomposition,CEEMD)和自适应权重粒子群算法(Adaptive Particle Swarm Optimization,APSO)优化长短期记忆网络(Long Short-Term Memory,LSTM)的故障诊断模型.为了减小一维振动数据中的噪声干扰,采用CEEMD方法对数据进行分解,结合皮尔逊相关系数提取最相关的故障特征.基于注意力机制增强LSTM网络的学习能力,聚焦关键故障信息,通过自适应权重粒子群算法优化模型参数.将关键故障分量输入模型并重构后输出,实现故障诊断.实验结果表明,CEEMD-APSO-LSTM故障诊断模型的准确率为99.6%,损失函数值为0.021 6.

In view of the problem of reduced fault diagnosis accuracy caused by irrelevant data interference such as noise in the fault vibration signal of stepper motors,a fault diagnosis model for optimizing LSTM(Long Short-Term Memory Network)based on CEEMD(Complementary Ensemble Empirical Mode Decomposition)and APSO(Adaptive Particle Swarm Optimization)algorithm is proposed.To reduce the noise interference in one-di-mensional vibration data,the CEEMD(Complementary Ensemble Empirical Mode Decomposition)method is adopt-ed to decompose the data,and the most relevant fault features are extracted in combination with the Pearson correla-tion coefficient.Enhance the learning ability of the LSTM based on the attention mechanism,focus on key fault in-formation,and optimize the model parameters through the adaptive weight particle swarm optimization algorithm.The key fault components are input into the model,reconstructed and then output to achieve fault diagnosis.The ex-perimental results show that the accuracy rate of the CEEMD-APSO-LSTM fault diagnosis model is 99.6%,and the loss function value is 0.021 6.

王锐;张建新;吕刚

浙江理工大学 信息科学与工程学院,浙江 杭州 310018浙江理工大学 机械工程学院,浙江 杭州 310018湖州太平微特电机有限公司,浙江 湖州 313013

信息技术与安全科学

步进电机故障诊断互补集合经验模态分解自适应权重粒子群算法长短期记忆网络注意力机制

stepper motorfault diagnosiscomplementary ensemble empirical mode decompositionadap-tive weighted particle swarm algorithmlong short-term memory networkattention mechanism

《电子科技》 2026 (9)

78-88,11

浙江省科技厅重点研发项目(2022C01188)Key R&D Program of Zhejiang Science and Technology Department(2022C01188)

10.16180/j.cnki.issn1007-7820.2026.09.010

评论