Hybrid Approach for Fault Detection of Control Actuation System in a Launch Vehicle

The hybridization of model based and data driven methods for a small satellite launch vehicle (SSLV) control actuation system is proposed in this paper. The flex nozzle control actuation system (FNC) uses electro-mechanical actuators (EMAs) to steer the launch vehicle by deflecting the nozzle. The fault detection of EMA is crucial to ensure the successful completion of any mission. The common fault detection approaches utilise only one method and thus fails to detect complete faults. In this work, Jaccard similarity coefficient is used to generate residuals for fault detection and a deep neural network has been used to identify the location at which the fault has occurred. This hybrid method helps to improve the reliability of fault detection and isolation (FDI). Once a fault has been detected and isolated, reconfiguration has been carried out with the help of redundant hardware components. The parameter variations caused by reconfiguration has been taken care by a model reference adaptive controller (MRAC).