Development and Application of a Parallel Two-Stage RLS-based Fault Detection for Nonlinear Thrust Vector Control Actuator on a Sounding Rocket

A Recursive Least Squares-based (RLS-based) fault detection algorithm was developed, namely Parallel Two-Stage RLS-based Fault Detection (PTSRLS-FD), to safely operate against false-positive cases by improving the detection for a specific fault type, namely loss of effectiveness (LOE). The algorithm was compared to a simpler, time domain residual model error-based detection algorithm (TDME) on several simulation results through actuator subsystem-level simulations, system-level Monte-Carlo (MC) analyses, Hardware-in-the-Loop (HIL) tests and flight tests all of which include both healthy and faulty actuator cases. The PTSRLS-FD have higher true-positive rates and lower false-positive rates in comparison with the TDME in the numerical experiments, and provided a safer residual margin in healthy actuator cases, and finally, the algorithm turned out to be fast enough to run online with at most 1.4 ms loop time.