基于概率神经网络模型的异步电机故障诊断

Based on Probabilistic Neural Network Model for Asynchronous Motor Fault Diagnosis

针对传统故障诊断方法的局限性,提出一种基于概率神经网络(PNN)的诊断方法。以异步电机转子断条、偏心、失电残压等故障为例进行了诊断研究,通过选取故障样本来训练PNN,将故障信息输入训练好的PNN模型后,由输出结果即可判断发生的故障种类。MATLAB仿真表明,基于PNN的电机故障诊断方法能有效识别出电机故障,故障诊断准确率高,易于工程实现。但神经网络还处于发展阶段,仍有不少问题需进一步研究。

According to the limitation of traditional fault diagnosis method, a diagnosis method based on probabilistic neural network was proposed. An example of asynchronous motor rotor with broken, eccentric, electric residual pressure fault was done. By choosing fault samples to train PNN, and then inputting the diagnosis information to the trained model of PNN, the occurred fault types could be judged from the output results. MATLAB simulation showed that diagnosis method of the motor based on probabilistic neural network could effectively identify motor fault and the fault diagnostic accuracy rate was so high that it could be easily implemented in engineering projection. But as neural network itself was undergoing developing, many problems need to be further studied.