This paper presents a comparison of alternative reliability curve fitting techniques for components of two model types of helicopters. Both mathematical function-based and neural network models were investigated. Preliminary results suggest that the neural network models compare very favorably with standard curve fitting techniques, and may provide better curve fitting for component reliability data from sparse data sets where the hazard rates are either constant or monotonically increasing.
Relation:
Reliability Engineering & System Safety 48(3), pp.229-234
