基于邻域粗糙集和并行神经网络的故障诊断

针对目前使用神经网络诊断故障时出现的输入向量选择困难、网络结构复杂、对并发故障诊断效果不好等问题,提出了基于邻域粗糙集和并行神经网络的故障诊断方法。先利用邻域粗糙集对初始征兆进行约简,留下有价值的征兆作为神经网络的输入向量,然后针对每种故障类型设计一个神经网络。用多个训练好的神经网络来并行诊断故障,综合每个神经网络的结果给出最终的诊断结论。用转子实验台的实验数据对这种故障诊断方法进行验证,结果显示该方法能优化神经网络结构,且神经网络具有训练速度快、诊断正确率高的特点。

A Fault Diagnosis Method Based on Neighborhood Rough Sets and Parallel Neural Networks

Using neural network to diagnose the faults may occur the problems such as difficult selection of input vector, complex structure of network and ineffective for simultaneous fault diagnosis. For that reason, this paper proposes a fault diagnosis method based on neighborhood rough sets and parallel neural networks. We first use neighborhood rough sets to reduce the initial signs. The remaining valuable signs will be used as the input vector of neural network. Then we design neural networks for each type of fault. We use the trained neural networks to diagnose the faults in parallel and give the final diagnosis conclusion according to the results of each network. We have tested the method by using the experimental data of rotor test stand and found that this method can optimize the structure of neural network and the networks need less training time and can ensure the accuracy of fault diagnosis.