基于图像融合的燃气轮机故障诊断研究
Research on gas turbine fault diagnosis based on image fusion
利用孔探仪获得燃气轮机的运行图像，通过图像故障诊断判断重型燃气轮机的健康状态，是重型燃气轮机故障诊断的一种重要研究方法。为诊断出燃气轮机的运行故障，并针对燃气轮机压气机与透平的故障进行准确识别，提出了基于图像的燃气轮机的融合故障诊断的方法。利用孔探仪记录燃气轮机运行状态，获取燃气轮机运行的图片。分别输入至ResNet和GoogleNet网络中进行训练，分别保存诊断精度最好的模型得到各自的故障诊断结果，再利用D-S理论判据实现对两种模型结果的决策层融合。结果表明，图像融合故障诊断算法可以准确的诊断出燃气轮机的故障，能得到95%以上的诊断率，不到1%的虚警率。
It is an important research method for fault diagnosis of heavy gas turbine to obtain gas turbine running image by using hole probe and judge the health state of heavy gas turbine by image fault diagnosis.In order to diagnose the operating faults of gas turbine and identify the faults of gas turbine compressor and turbine accurately, a fusion fault diagnosis method of gas turbine based on image is proposed. The running state of the gas turbine is recorded by the hole probe, and the running picture of the gas turbine is obtained. They were input into ResNet and GoogleNet networks for training, and the models with the best diagnostic accuracy were saved respectively to obtain their own fault diagnosis results. The D-S theoretical criterion was used to realize the decision level fusion of the results of the two models. The results show that the image fusion fault diagnosis algorithm can accurately diagnose the fault of gas turbine, with a diagnosis rate of more than 95% and a false alarm rate of less than 1%.
程卓;曹云鹏;冯伟兴;
哈尔滨工程大学 智能科学与工程系,哈尔滨150001;哈尔滨工程大学,能源与动力学院,哈尔滨150001;哈尔滨工程大学 智能科学与工程系,哈尔滨150001;
TK4
gas turbine fault diagnosis decision fusion
1-7 / 7
评论