一种基于聚类的GRBF网络机械缺陷识别方法

A Method of Mining Machinery Flaw Identification Based on Clustering Algorithm Optimized GRBF

针对煤矿机械缺陷超声信号的非平稳特性,利用小波包和基于聚类的广义RBF神经网络进行缺陷的智能识别.重点研究了利用小波包方法提取反映不同缺陷性质的特征值和GRBF神经网络分类方法.并以机械焊接缺陷为研究对象,进行了实验研究.结果表明该方法与其他方法相比,具有较高的缺陷分类准确率.

According to the nonstationarity of mining machinery flaw signals in ultrasonic testing,the method used for defect identification based on WPT and Clustering Algorithm optimized GRBF.Focus on WPT extract the different defects characteristics and GRBF Algorithm classify the different flaws.Experimental study the mining machinery Welding flaws.Compared with the BP,experimental results shows that this algorithm has high accuracy of flaw classification.