粗糙集及PSO优化BP网络的故障诊断研究

Fault Diagnosis Research by Rough Set Theory and the PSO-BP Neural Network

针对BP神经网络故障诊断存在网络结构复杂、训练时间长、精度不高的问题,将粗糙集、微粒群算法、遗传算法引入到柴油机故障诊断中,提出了基于粗糙集理论与改进BP神经网络相结合的柴油机故障诊断算法。算法采用自组织映射方法对连续属性离散化,利用粗糙集理论对特征参数进行属性约简,使用微粒群算法优化BP网络结构,从而缩短训练时间,有效提高故障诊断的准确度。最后用柴油机的实际诊断结果验证了该算法的可行性、快速性和准确性。

For the imperfections of BP network fault diagnosis model,including the complexity of the network structure,the long time of training,and the low precision,this article introduced rough set（RS）,particle swarm optimization（PSO） and genetic algorithm（GA） into the diesel engine fault diagnosis,then proposed a new algorithm that is based on rough set theory and the improved BP neural network.The algorithm uses self-organization mapping net（SOM） to discretize the continuous attributes,rough set theory to make a reduction on the properties for characteristic parameters,and the particle swarm optimization（PSO） to optimize the BP network structure,so that it can shorten training time and improve the accuracy of fault diagnosis effectively.Finally,the result of the diesel engine＇s diagnosis proves the feasibi-lity,rapidity,veracity of the algorithm.