基于RBF网络的旋转机械故障诊断方法

Fault diagnosis method of rotary machinery based on RBF networks

针对 BP网络在故障诊断中存在的训练收敛速度慢且容易陷入局部极小、网络初值对学习性能影响比较大等缺陷 ,提出了一种基于 RBF网络的故障诊断方法 ,介绍了 RBF网络的结构和自适应正交最小方差算法 ( SROSL法 ) ,并应用于旋转机械的故障诊断中 .应用结果表明 ,RBF网络训练速度快、分类性能良好 ,在设备故障诊断领域具有很好的实用性 .

Aiming at the insufficiency of BP learning algorithm in machinery fault diagnosis, such as the low learning convergence speed, the easily appearing local minimum , instability learning performance caused by initial value, the authors propose a new diagnosis method based on RBF networks, and introduce the structure and self adaptive orthogonal least squares learning algorithm (SROSL) of RBF networks. And RBF networks is applied to rotary machinery fault diagnosis. The result shows that RBF networks has very high learning convergence speed and better classifying performance. RBF networks has good practicality in the field of equipment fault diagnosis.