摘要
针对传统的故障识别方法存在信号质量低和诊断精度差等问题,提出一种参数自适应变分模式提取(PA-VME)和稀疏保持投影(SPP)相结合的数据驱动机械故障诊断新方法。结合相关系数、L-峭度和信息熵构造一个新的指标L_(FCI)并将其作为适应度函数,采用粒子群算法对变分模式提取的内部参数进行优化,从而形成PA-VME模型并将其用于振动信号的模式分解;根据构造的指标能够反映信息有序度的原则,选取有效的模式分量并计算得到高维特征数据集;利用SPP将数据集通过权重矩阵投影到低维空间,实现对高维特征数据的降维和聚类分析。通过对仿真信号和实验台的故障信号进行分析,证明其对不同类型机械故障的识别精度可以达到96.87%。
Aiming at the low signal quality and poor diagnosis accuracy in traditional fault identification methods,this paper proposes a new data-driven mechanical fault diagnosis method combining Parameter Adaptive Variational Mode Extraction(PA-VME) and Sparse Preserving Projection(SPP).A new index L_(FCI) is constructed by combining the correlation coefficient,L-kurtosis and information entropy as a fitness function.Particle Swarm Optimization algorithm is used to optimize the internal parameters of VME,so as to form a novel PA-VME model and use it for the mode decomposition of vibration signals.According to the principle that the constructed index can reflect the order of information,the interested model components are selected and the high-dimensional feature data set is calculated.SPP is applied to project the data set into the low dimensional space through the weight matrix to achieve dimension reduction and clustering analysis of high-dimensional feature data.The analysis of simulation signals and test-bed fault signals proves that the recognition accuracy of the proposed model for different faults can reach 96.87%.
作者
柯伟
金仲平
董灵军
吕信策
KE Wei;JIN Zhongping;DONG Lingjun;LYU Xince(Taizhou Special Equipment Inspection and Testing Institute,Taizhou 318001,China)
出处
《机械制造与自动化》
2024年第2期60-66,74,共8页
Machine Building & Automation
基金
台州市科技局工业类科技计划项目(21gyb13)。
关键词
参数自适应变分模式提取
稀疏保持投影
特征提取
机械设备
故障诊断
parameter adaptive variational mode extraction
sparsity preserving projections
feature extraction
mechanical equipment
fault diagnosis
