Machine learning-based Scheme for Fault Detection for Turbine Engine Disk

Real-time fault detection of rotating engine components is a fundamental task for aero community, especially for commercial aircraft to cut maintenance costs for airline companies and enhance aviation security. Therefore, the aim of this paper is to develop a diagnostic framework for real-time fault detection of a turbine engine disk for commercial aircraft, particularly for state monitoring of rotating engine components. The proposed framework relies on a machine learning algorithm that can adjust to the shifts of the motion state of a commercial aircraft. In addition, feature selection techniques can reduce the repetitive, or unnecessary features, which might degrade the accuracy of classification. Accordingly, we consider the multi-layer perceptron algorithm to classify a sample between normal or fault and a binary particle swarm optimization-based feature selection scheme. Also, the paper offers comparative approaches such as the perceptron algorithm, the recursive feature elimination as another feature selection method, and so on. The simulation results show that the proposed framework is robust to changes in the operating conditions and achieves the best accuracy between all the methods considered.