A radial basis function networks appraoch in turbine engine prognosis

摘要

Army Research Laboratory (ARL) and Texas A&M University (TAMU) have been developing a Turbine Engine Prognostic sysem for M1A1 tank turbine engine. This paper describes an approach to predict failures of the AGT-1500 gas turbine engine of the tank by detecting the presence of air leaks in the engine. We analyze the problem, identify appropriate discriminator signals from sensor data, and then solve the problem using Radial Basis Function networks approach. The results of the experiemtnal evaluation indicate the effectiveness of the approach. After learning from a set of training signals representing the conditions of normal engines and engines that introduce small and large air leakds, the system is able to detect correctly the presence and the category of the problem.