首页|基于CMIE与参数优化KELM的旋转机械故障诊断策略

基于CMIE与参数优化KELM的旋转机械故障诊断策略

扫码查看
原文链接
  • 万方数据
  • 维普
针对多尺度排列熵忽略信号幅值信息以及粗粒化处理存在不足,造成旋转机械故障识别准确率不稳定和不可靠等缺陷,提出了 一种基于复合多尺度增长熵(CMIE)和算术优化算法(AOA)优化核极限学习机(KELM)的旋转机械故障诊断策略(方法).首先,引入增长熵代替排列熵,进行了故障特征提取,同时采用复合粗粒化处理进行了信号的多尺度分析,提出了复合多尺度增长熵指标,将其用于提取旋转机械振动信号的非线性故障特征;随后,利用AOA对KELM的核心参数进行了 自适应优化,建立了网络结构最优的分类模型;最后,将故障特征输入至AOA-KELM分类器,进行了训练和测试,根据分类器的输出标签完成了样本的故障识别任务;利用旋转机械故障数据集对所提策略的性能进行了实验和分析.研究结果表明:CMIE方法可以有效地识别旋转机械的故障类型和故障程度,两种数据集的识别精度均达到了 99.2％,在特征提取效率和识别精度方面均优于比较方法;AOA-KELM模型的识别准确率和识别效率优于遗传算法优化核极限学习机、粒子群算法优化极限学习机、网格算法优化核极限学习机和灰狼算法优化核极限学习机.
Fault diagnosis strategy of rotating machinery based on CMIE and parameter optimization KELM
Aiming at the problem that multiscale permutation entropy ignored signal amplitude information and coarse-grained processing was insufficient,which led to unstable and unreliable fault identification accuracy of rotating machinery,a new fault diagnosis strategy based on composite multiscale increment entropy(CMIE)and arithmetic optimization algorithm(AOA)optimized kernel extreme learning machine(KELM)for rotating machinery was proposed.Firstly,introducing increment entropy instead of permutation entropy for fault feature extraction and using composite coursed processing for multi-scale analysis of signals,a composite multiscale increment entropy index was proposed for extracting nonlinear fault features of rotating machinery vibration signals.Then,the core parameters of KELM were adaptive optimized by AOA,and the optimal classification model of network structure was established.Finally,the fault features were input to the AOA-KELM classifier for training and testing,and the fault identification of samples was realized according to the output label of the classifier.The performance of the proposed strategy was tested and analyzed using rotating machinery fault data sets.The research results indicate that the CMIE method can effectively identify the types and degrees of faults in rotating machinery,and the recognition accuracy of both datasets reaches 99.2％,which is superior to the comparative method in terms of feature extraction efficiency and recognition accuracy.The recognition accuracy and efficiency of AOA-KELM are better than that of genetic algorithm optimized kernel extreme learning machine,particle swarm optimization optimized kernel extreme learning machine,grid search optimized kernel extreme learning machine and grey wolf algorithm optimized kernel extreme learning machine.

composite multiscale increment entropy(CMIE)arithmetic optimization algorithm(AOA)kernel extreme learning machine(KELM)rotating bearinggearboxcomposite coursed processingsignals multi-scale analysis

连璞、吴磊、伍永豪

展开 >

长治职业技术学院机械电子工程系,山西长治 046000

西安交通大学机械工程学院,陕西西安 710049

联合传动及轴承技术研究中心,宁夏石嘴山 753000

复合多尺度增长熵 算术优化算法 核极限学习机 滚动轴承 齿轮箱 复合粗粒化处理 信号多尺度分析

国家自然科学基金资助项目

51575421

2024

10.3969/j.issn.1001-4551.2024.01.007

机电工程
浙江大学　浙江省机电集团有限公司

机电工程

CSTPCD北大核心
影响因子：0.785
ISSN：1001-4551
年,卷(期)：2024.41(1)
  • 23