The research will develop and experimentally evaluate a sensor-based fault diagnostic method for helicopter drive trains. This method is based on an approach designed to cope with the variability and overlap of fault signatures, two major sources of difficulty for the existing diagnostic systems. This approach employs vectors of fault signatures in its diagnostic model; therefore, it can tolerate the overlap between individual signatures and, since it adjusts its diagnostic model in-process, it can cope with fault signature variations. This approach uses a multi-valued influence matrix (MVIM) as its diagnostic model with the columns representing the "average" values of the fault signatures. The MVIM approach can assess the diagnosability of the system with it s available measurements, select an optimal subset of measurements subject to the diagnosability constraint, perform on-line diagnostic reasoning, and update it diagnostic model (influence matrix) in response to process variations. The research will include (1) a detailed review of the current state of the art techniques for fault signature identification of power train systems, (2) a study of the available data on power train failure for possible tailoring of the MVIM method for a feasibility study, (3) development of a feasibility demonstration plan to support and justify a more extensive validation program, and (4) analysis of the results of the feasibility test to further modify the theory and develop a plan to fully validate the methodology.//
