|
Abstract
|
Helicopter rotor failures pose significant threats to aviation safety, making accurate prediction and prevention crucial for minimizing accidents. This study delves into determining influential factors contributing to rotor failure classifications, utilizing machine learning-based feature selection methods. By analyzing 135 rotor failure incidents from a comprehensive dataset of 56,652 helicopter-related accidents, eight failure classes were identified. These include critical issues such as mast fatigue, control looseness, and rotor unbalance. The maximum takeoff weight, airframe hours since the last inspection, type of last inspection, power units, total airframe hours, altitude, wind speed, wind direction, and phase of flight are considered as the input features. By comparing five advanced methodologies, including deep learning, artificial neural networks, correlation matrix method, XGBoost (eXtreme Gradient Boosting) method and mutual information method the study identifies key features with robust justification in flight mechanics. Finally, it is concluded that “certification maximum gross weight”, “number of power units”, “phase of flight”, and “aircraft flight hours” have the highest degree of importance in prediction of the failure class of helicopter rotor. By prioritizing these variables, the findings aim to enhance predictive accuracy, reliability, and flight safety, paving the way for proactive measures in rotor failure prevention. This research marks a step forward in aviation safety and underscores the potential of machine learning in addressing complex challenges in rotorcraft operations.
|