Rocket Engine Failure Detection Using System Identification Techniques

HPFT HPFTP This paper presents the; theoretical foundation and HPOP application of two univariate failure detection HPOT algorithms to Space Shuttl,; Main Engine (SSME) test LPFP firing data. Both algorithms were applied to data LPOP collected during steady-sutte operation of the engine, m One algorithm, the time s_'.ries algorithm, is based on MCC time series techniques and involves the computation of p(x) autoregressive models. Tim_ series techniques have been previously applied to SSMF data. The second algorithm Pxx(f) is based on standard sigmd processing techniques. It PBP consists of tracking the variations in the average signal PID power with time. The average signal power algorithm is PSD a newly proposed SSME failure detection algorithm, q-I Seven nominal test firings were used to develop failure rxx [m] indication thresholds fcr each algorithm. These thresholds were tested using four anomalous firings and RPL one additional nominal fi_g. Both algorithms provided SSME significantly earlier failure indication times than did the T current redline limit sysUm. Neither algorithm gave u(t) false failure indications Ior the nominal firing. The x[n] strengths and weaknesses of the two algorithms are y(t) discussed and compared. The average signal power algorithm was found to have several advantages over the time series algorithm.