Feature Selection for Aero-Engine Fault Detection

Timely and accurate detection of aero-engine faults is crucial to preventing loss of lives and equipment. In recent times, there has been a focus on data-driven approaches to fault detection in aero-engines owing to the availability of numerous sensor information which addresses the complexities of model-based techniques. However, the increased use of sensors in aero-engines induces problems relating to multicollinearity and high dimensionality in developing fault detection models. Various feature selection approaches have been proposed for tackling dimensionality problems, with each offering advantages based on the peculiarity of the data. This study, therefore, investigates the use of feature-selection approaches to address the dimensionality problems associated with aeroengine data. Our study also reveals that careful evaluation of feature selection approaches is effective in achieving earlier fault detection in aero-engines with enhanced model performance.