Non-linear Model-based Stochastic Fault Diagnosis of 2 DoF Helicopter

Fault diagnosis of non-linear helicopter systems are affected by inherent characteristics such as non-linear behaviour, high cross coupling effects, external disturbances such as atmospheric turbulence and wind effects. Fault diagnosis in non-linear systems gains importance due to its high complexity and this work focuses on fault detection of helicopter system with the consideration of the inherent non-linearity effects. This paper deals with the detection, identification and classification of sensor, actuator and component faults in nonlinear helicopter systems using model-based state estimation approaches. Approaches include Interacting Multiple Model based Extended Kalman Filter and Interacting Multiple Model based Unscented Kalman Filter. To address problem of fault detection, statistical measures of residual analysis, stochastic likelihood ratio and model probability is proposed. A Comparison of these approaches is presented based on the ability to detect, identify and classify faults in spite of system non-linearity. Algorithm is applied to 2 degrees of freedom helicopter and the results for various fault cases are presented. The results yield better fault detection performance using Interacting Multiple Model based Unscented Kalman Filter.