一种聚类优化融合故障诊断方法及其应用

针对单一聚类诊断方法难以准确、全面识别不同故障状态的问题，提出了一种聚类优化融合故障诊断方法。分别利用社团聚类、K均值聚类及粒子群聚类三种方法对故障进行识别，得出三种聚类方法对应的故障识别准确率，在此基础上构建初始权值矩阵，并通过遗传算法对初始判断矩阵与三种聚类方法进行优化，得到最优权值矩阵与优化的聚类模型，用于融合诊断。轴承故障诊断实例结果表明，该聚类融合诊断方法能够有效提高故障识别准确率。

Clustering Optimization Fusion Method for Fault Diagnosis and Its Applications

Single community diagnosis clustering methods were difficult to identify different fault states, in order to improve diagnostic accuracy, a fusion clustering method was proposed herein based on genetic optimization algorithm. Three clustering methods, the community clustering, the Kmeans clustering and the particle swarm clustering, were used to identify the fault states respectively. The diagnostic accuracies were used to construct an initial weight matrix. The genetic optimization algorithm was used to optimize the weight matrix. The examples of bearing fault diagnosis show that the clustering optimization fusion method may improve diagnostic accuracy.