Fault Diagnosis Method of Gas Turbine Combustion Chamber Based on CNN-GRU Model Analysis

Abstract

The safety, reliability and economy of gas turbines all depend on the fault diagnosis of gas turbines. In order to solve the fault diagnosis accuracy problems of false alarm and false alarm in gas turbine combustion chamber, a fault diagnosis method of gas turbine combustion chamber based on gated recurrent unit (GRU) optimization convolutional neural network (CNN) model analysis is proposed. First, a sample set of combustion chamber failure data is generated by constructing a gas turbine thermodynamic model. The CNN model is then optimized using GRU to extract the spatial and temporal features of the data, using small convolution kernels and 2D convolution methods. Finally, the extracted features are fused and fed into a fully connected layer for fault type identification. Experimental results show that the proposed method is highly practical and feasible compared to the traditional combustor threshold-defined fault diagnosis methods and other artificial intelligence fault diagnosis methods, with an average diagnosis accuracy of 97.66%, which is higher than the identification accuracy.