Residual Selection for Observer-Based Fault Detection and Isolation in a Multi-Engine Propulsion Cluster

For complex systems, the number of residual candidates generated by Structural Analysis could be in the order of tens of thousands, and implementing all candidates is infeasible. This paper addresses the residual generator candidate selection problem from a state-observer perspective. First, the most suitable candidates to derive state-observers are selected based on two criteria related to the state-space form and a low number of equations. Then, a novel algorithm finds the minimal subset of residual generator candidates capable of detecting and isolating all faults. A procedure is introduced to compare the fault sensitivity of the selected candidates. This residual selection method is applied to the multi-engine propulsion cluster of a reusable launcher to illustrate its benefits.