59289 - A Novel Gas Path Fault Diagnostic Model for Gas Turbine Based on Explainable Convolutional Neural Network With Lime Method 

Gas turbine is widely used in aviation and energy industry. Gas path fault diagnosis is an important tool used for gas turbine operation and maintenance. With the development of information technology, especially deep learning methods, data-driven approaches for gas path diagnosis are developing rapidly in recent years. However, most data-driven models are hard to explain in mechanism, which decreases the credibility of the methods. In this paper, a novel explainable data-driven model for gas path fault diagnosis based on Convolutional Neural Network (CNN) using Local Interpretable Model-agnostic Explanations (LIME) method is proposed. Prior mechanism information of gas turbine fault effects and fault modes is used to build the measurement sensor matrix as the input of CNN model. The relationship between the measurement parameters and fault modes is adoptedto arrange the relative position in the matrix. Various types of fault data generated by mechanism model are used to train the model. The regions in the matrix which contributes to fault recognization can be visualized with LIME method, and the prior mechanism information is used to verify the fault diagnostic process and to improve the arrangement of measurement sensor matrix. This method can express the relevance of the fault mode and its high-correlation measurement sensors in the model, which can greatly improve the interpretability and application of the model.