基于粗糙遗传BP神经网络的滚动轴承故障诊断

Fault Diagnosis on Rolling Bearing Based on Rough Genetic BP Neural Network

为了提高滚动轴承的故障诊断效率和精度,将粗糙集理论和遗传BP神经网络相结合,提出了一种新的滚动轴承故障诊断方法。首先利用粗糙集理论对故障诊断决策表进行属性约简,以简化BP神经网络的结构及降低计算量。然后利用遗传算法来优化BP神经网络的参数以达到最优泛化能力,从而建立粗糙遗传BP神经网络故障诊断模型。以美国凯斯西储大学的轴承数据为例,通过MATLAB仿真,结果表明该方法不仅可以克服BP神经网络的缺陷、减少遗传寻优迭代次数,还能提高故障诊断精确度。

In order to improve the fault diagnosis efficiency and precision of rolling bearing, a new fault diagnosis method for rolling bearing was proposed by combining rough set theory with genetic BP neural network. Firstly, the attribute reduction of fault diagnosis decision table was made by rough set theory to simplify the structure of BP neural network and reduce the computation. Then, the genetic algorithm was used to optimize the parameters of BP neural network to achieve optimal generalization ability, so as to establish the fault diagnosis model of the rough genetic BP neural network. Taking the case of Case Western Reserve University, through MATLAB simulation , the results show that this method can not only overcome the defects of BP neural network and reduce the number of genetic optimization iterations, but also improve the accuracy of fault diagnosis.