基于ICEEMDAN和PSO-LSSVM的滚动轴承故障诊断方法研究

Research on the Application of ICEEMDAN and PSO-LSSVM Algorithms in Rolling Bearing Fault DiagnosisZheng Lizhao1, Song Hongzhi1, Gu Qilin1,Zhang Baoling1,An Hongxin1,

摘要:

针对滚动轴承疲劳故障振动信号具有能量弱、特征稀疏等特点，提出了一种通过改进自适应噪声完备经验模态分解方法与粒子群优化的最小二乘支持向量机结合的故障识别方法。对轴承不同故障信号利用改进的自适应噪声完备经验模态算法分解为一系列固有模态函数分量；根据相关系数-方差贡献率准则筛选出最能表征原始信号状态的分量，并计算重构分量的奇异谱熵值构成特征向量；将提取的特征向量集合输入到基于粒子群优化的最小二乘支持向量机分类器中，进行模型的训练和故障模式的识别，与SVM和LSSVM分类器模型进行准确率和效率比较。试验结果表明，该方法在滚动轴承故障信号中能有效提取故障特征，准确率达98.75%，具有一定可靠性和实用性。

Abstract:

In view of the weak energy and sparse features of fatigue fault vibration signals of rolling bearings, a fault identification method combining improved adaptive noise complete empirical mode decomposition (ICEEMDAN)and particle swarm optimization least-squares support vector machine was proposed (PSO-LSSVM). Different bearing fault signals are decomposed into a series of inherent modal function (IMF) components by an improved adaptive noise complete empirical mode algorithm; The component that can best represent the original signal state is selected according to the correlation core-variance contribution ratio criterion, and the singular spectrum entropy of the reconstructed component is calculated to form the feature vector; The extracted feature vector set is input into the least square support vector machine classifier based on particle swarm optimization, and the model is trained and the fault mode is identified. The accuracy and efficiency of the model are compared with that of SVM and LSSVM classifier。The test results show that the method can effectively extract fault characteristics from rolling bearing fault signals with an accuracy of 98.75%, which has certain reliability and practicability.