关于电机故障诊断方法的优化研究  

Optimization studies of motor fault diagnosis method

作　　者：王惠中[1] 夏雨婷[1,2] 乔林翰 刘联涛[1] WANG Hui - zhong XIA Yu - ting QIAO Lin - han LIU Lian - tao(Lanzhou University of technology, Lanzhou Gansu ,730050, China Graduate Students in Wenzhou Branch, Lanzhou University of Technology, Wenzhou Zhejiang,325000, China)

机构地区：[1]兰州理工大学,甘肃兰州730050 [2]兰州理工大学温州研究生分院,浙江温州325000

出　　处：《计算机仿真》2017年第6期361-366,共6页Computer Simulation

摘　　要：针对高维数据的维数困扰、参数影响性能以及噪声干扰,采用等谱流形学习算法进行数据的降维,并采用狼群算法对等谱流形学习算法和最小二乘支持向量机的参数进行优化,提出了一种电机故障诊断的方法。结果表明:首先采用狼群算法优化等谱流形学习算法的特征维数d以及最小二乘支持向量机的惩罚参数C和核函数γ,并用等谱流形学习算法对数据进行降维处理,以提取其低维的特征空间,最后在建立最小二乘支持向量机模型,从而进行分类。仿真结果验证了所建模型的正确性和有效性,而且等谱流形学习算法的降维结果优化于主成分分析,其抗噪能力也很强,并且狼群算法优化的分类结果也优于粒子群算法。This paper focuses on diagnosis method of motor fault. Firstly,wolf pack algorithm is used to optimize feature dimension d of equal spectrum manifold learning algorithm and penalty parameter C and kernel function γ of least squares support vector machine （ LS - SVM ）, then the equal spectrum manifold learning algorithm is used to carried out dimension reduction process for data to extract its feature space with low dimension. Finally, LS - SVM model is built, thus classification is carried out. Simulation results prove validity and effectiveness of the model. The dimension reduction result of equal spectrum manifold learning algorithm is superior to the one of principal component analysis. Anti - noise capacity is excellent and classification results optimized by the wolf pack algorithm are superior to particle swarm algorithm.

关 键 词：等谱流形学习算法 狼群算法 最小二乘支持向量机 主成分分析 粒子群算法 

分 类 号：TN929.52[电子电信—通信与信息系统]