基于神经网络的故障诊断推理方法

针对传统诊断技术的局限性，研究了基于ＢＰ模型神经网络的故障诊断推理方法，它只需选择足够的具有代表性的故障样本训练神经网络，将代表故障的信息输入给训练好的神经网络，根据神经网络的输出结果，就可以判断发生故障的类型．神经网络一旦训练好，由于其具有容错性，不仅能诊断出已经出现过的故障，还能在一定范围内诊断出从未出现过的故障，使故障诊断智能化和简单化．仿真结果表明，基于神经网络的故障诊断方法是行之有效的

Fault Diagnosis Inference Based on Neural Network

In the light of limitation of traditional diagnosis technology, to start new effective methods for intelligence in fault diagnosis, the paper studied inference methods in fault diagnosis based on BP model neural network. The simulation results show that the fault diagnosis based on neural network is very effective. It need not set up a complicated mathematical model, and also need not do complicated mathematical calculation and data processing. All you need do is select enough typical fault samples to train the neural network. When information that represents fault is input to the trained neural network, the type of shown fault can be judged by the output of the neural network. Once the neural network training has finished, the neural network with fault tolerance can diagnose not only the shown faults but also some faults never shown in some extent. The fault diagnosis is intelligent and simple.