基于优化神经网络的空调系统未知类型故障诊断

Fault Diagnosis for Unknown Types of Faults of Air Conditioning System Based on Optimized Neural Network

目前基于数据驱动方法的制冷系统故障诊断模型,只能对参与建模训练的已知类型故障进行诊断,而对于未参与建模训练的未知类型故障,不能正确地诊断。针对这一问题,本文提出了一种优化神经网络的故障诊断策略。利用已知类型的故障数据建立BP神经网络模型,然后确定一个区分阈值,能够实现对未知类型故障的诊断识别。结果表明:对于包含所有故障类型的测试数据,模型的诊断正确率为88.62%,对于其中的未知类型故障,模型的诊断效果显著,正确率为99.48%。

At present, the fault diagnosis models based on the data driven for refrigeration system can only diagnose the known types of faults which are involved in the modeling process, but they fail to diagnose the unknown types of faults which are not involved in the modeling process. To solve this problem, the strategy based on an optimized neural network is presented. The known types of faults data are used to train the back propagation neural network model. Then, a threshold of distinction is established, which can diagnose the unknown types of faults successfully. The results show that the accuracy rate of the model is 88.62% for the testing data which includes all types of faults. For the unknown types of faults, the performance of the model is significant and the accuracy rate is 99.48%.