基于改进K均值聚类的机械故障智能检测  

Machinery Fault Intelligent Detection Based on Improved K-means Clustering

作　　者：费贤举[1] 

机构地区：[1]常州工学院,江苏常州213002

出　　处：《计算机测量与控制》2015年第4期1121-1123,共3页Computer Measurement &Control

摘　　要：针对传统的K均值聚类算法在机械故障检测的过程中,由于对K值的选择具有较强的主观性,最后极易得到局部最优解,而非全局最优解,降低了机械故障检测的准确性。提出一种改进K均值聚类的机械故障智能检测方法;将K均值聚类算法与粒子群算法相结合,在迭代处理的过程中,结合K均值进行优化,即将粒子群算法中的子代个体利用K均值聚类进行运算获取局部最优解,并使用这些个体继续参与迭代处理,这样能够提高算法的收敛速度,避免陷入局部最优解,获得准确的机械故障信号特征;实验结果表明,利用K均值倾斜特征提取的机械故障智能检测算法进行机械故障检测,能够有效提高故障检测的准确性,取得了令人满意的效果。In view of the traditional k--means clustering algorithm in the process of mechanical fault detection, due to the choice of K value has strong subjectivity, finally to get local optimal solution, rather than the global optimal solution, reduces the accuracy of the mechanical fault detection. An improved k-means clustering of machinery fault intelligent detection method is prposed, k--means clustering algorithm is combined with particle swarm algorithm, in the process of iteration process, in combination with k--means is optimized, the particle swarm algorithm of individuals using k--means clustering to obtain the local optimal solution, and use these individuals continue to participate in the iteration process, it can improve the convergence speed of the algorithm and avoid falling into local optimal solution, obtain accurate mechanical fault signal characteristics. Experimental results show that using k--means tilt feature extraction of machinery fault intelligent detection algorithm for fault detection, can effectively improve the accuracy of fault detection, and satisfactory results have been achieved.

关 键 词：K均值聚类算法 特征提取 机械故障检测 

分 类 号：TH128[机械工程—机械设计及理论]