Research On State Recognition Technology Of Key Components Of Gas Turbine

The gas turbine is applied to the aircraft as a power device and its working environment is bad.In order to ensure the effectiveness of its work,it is of great significance to study the state identification technology of its key components.Therefore,this paper uses a variety of methods to study the state identification technology of key components of gas turbine.There are many key components with different characteristics in the gas turbine.According to the demand of the actual research project,in order to study the state recognition technology,this paper takes the nozzle afterburner regulator as the specific object of study.On the special test platform of a certain type of gas turbine,the related operating parameters of the nozzle afterburner regulator of the key components are obtained through the test.Four kinds of states,such as normal,spout signal fault,abnormal swing of tail nozzle position and tension spring fracture,are determined and analyzed.Because of many parameters and unclear feature information,the kernel principal component analysis method is used to extract the 9 parameters,such as the exhaust temperature of the turbine and the low pressure rotor speed,to reduce the data dimension and extract the feature information that can reflect the state class better.Then,GRNN neural network,Elman neural network and deep learning are adopted respectively to create GRNN neural network,Elman neural network and DBN neural network state recognition model.The recognition effect of the established state recognition model is tested.The results show that the recognition accuracy of the three methods is 90%,85% and 91.25%,respectively,and the recognition effect of the deep learning method is the best.Finally,fuzzy integral and D-S evidence theory are used to study the state recognition results of three kinds of neural networks including GRNN,Elman and DBN in information fusion technology.The study shows that the recognition accuracy of the fusion based on fuzzy integral and D-S evidence theory is increased to 92.5% and 97.5%,respectively.These two information fusion methods have improved the accuracy of gas turbine key components status identification,especially the D-S evidence theory information fusion method is more effective.