Gas Turbine Fault Diagnosis Based on Machine Learning Techniques

This paper proposes a method of diagnosis and prognosis of the imbalance fault of a gas turbine using machine learning techniques by putting its advantages over the old diagnostic procedures, mainly temporal analysis and frequency analysis of signals. Therefore, this work proposes a predictive maintenance approach to monitor and detect in realtime whether the power turbine is subject to an imbalance or is in normal operating condition using artificial intelligence techniques. For this purpose, a comparative study of three machine-learning algorithms, Support Vector Machine (SVM), Random Forest (RF), and K Nearest Neighbors (KNN), has been carried out. The results show that both RF and KNN have the best performance, with an F1 score of 99.95% and 99.97%, respectively. For the Linear SVM, we obtained an F1 score of 61.73%.