粗集神经网络及其在齿轮箱故障识别中的应用

Rough neural network and its application to gearbox-fault recognition

为有效解决故障诊断中的故障识别问题,提出了一种以强耦合方式进行集成的粗集-神经网络的故障识别方法。该方法通过粗糙集对获得的故障属性进行降维预处理并获取粗集规则,再以所得的粗集规则与BP神经网络进行强耦合对故障进行识别。对该方法以齿轮箱故障识别进行了仿真实验,与BP神经网络的识别效果进行对比,结果表明了粗集神经网络在训练速度、测试精度方面的有效性。

To effectively solve the problem of the fault recognition in fault diagnosis, a method of fault recognition by rough neural network through the way of strong coupling is presented. This method reduces attribute-dimension of fault by rough set for preprocessing and obtains rough set rules, with the rough set rules obtained and the BP neural network assembled by a way of strong-coupling for fault recognition. Application in gearbox fault recognition are used to verify this algorithm and the result of simulation, which is compared to that of BP neutral network algorithms, validated the effectiveness of training speed and testing precision by such rough neural network.