基于灰度-梯度共生矩阵和模糊核聚类的振动图形识别方法

摘要

Taking the vibration parameter graphics of a reciprocating machinery as a study objective, a vibration image recognition method based on gray-gradient co-ocunence matrix and kernel-based fuzzy clustering was proposed. Using gray-gradient co-occurrence matrix directly extracted vibration features of a vibration image, the texture characteristics obtained were taken as samples to be inputted. With Mercer kernel method, the texture characteristics were mapped to a Gaussian feature space to have a better clustering in the high-dimensional feature space to realize an intelligent fault diagnosis of a reciprocating machinery. The test results showed that this method has higher diagnosis accuracy, and possesses a certain feasibility and effectiveness.%以往复机械振动参数图形为对象,提出了基于灰度-梯度共生矩阵和模糊核聚类的振动图形识别方法.利用灰度-梯度共生矩阵直接提取振动参数图形中的特征信息,将得到的纹理特征参量作为样本输入空间,通过Mercer核把输入样本映射到高斯特征空间后,在高维特征空间中进行聚类,从而实现往复机械故障智能诊断.实验结果表明,该方法可以获得较高的诊断精度,具有一定的可行性和有效性.