Neurl Network-Based Real-Time Ontogenetic Engine Health

A problem of current interest to the Air Force is the automatic classification of potential critical component failures in machinery and structural parts. Current processing for health monitoring uses relatively simple metrics or features to measure and characterize changes in sensor data. An alternative solution is to use neural networks coupled with appropriate feature extractors to analyze and automatically extract rules for expert system classifier development. Radial Basis Function neural nets perform classification of input features by clustering and characterizing the feature data with a set of multidimenstional (for example Gaussian) basis functions. This set of basis functions forms rules that are used for classification. Proposed here is the development and application of a class-dependent elliptical basis function neural net developed at ORINCON for similar problems. The network is essentially a nearest-neighbor classifier and can perform automated rule extraction. It can be easily updated to develop new rules for new classes as data becomes available. A variety of feature extractors developed at ORINCON will be considered as input to the nets. The system performance will be assessed and demonstrated processing real data supplied by the Allison Engine Company, which is a partner on Phase I and Phase II.