Helicopter Rotor System Fault Detection Using Physics-Based Model and Neural Networks

Abstract

A comprehensive physics-based model of the helicopter rotor in forward flight is used to analyze the impact of selected faults on rotor system behavior. The rotor model is based on finite elements in space and time. The helicopter rotor model is used to develop a neural network-based damage detection methodology. Simulated data from the rotor system are contaminated with noise and used to train a feedforward neural network using backpropogation learning. Cases considered for training and testing the neural network include both single and multiple faults on the damaged blade. Results show that the neural network can detect and quantify both single and multiple faults on the blade from noise-contaminated simulated vibration and blade response test data. For accurate estimation of type and extent of damages, it is important to train the neural networks with noise-contaminated response data.