Optimisation of Fusion and Decision Making Techniques for Affordable SPHM

Abstract

Over the past six years, Smiths and BAE Systems have launched collaborative work to evolve a certifiable practical SPHM system. The collaborative work has built on BAE Systems' vast advanced technology experience and on Smiths' unique experience that has produced intelligent Fleet and Usage Management Software (FUMStrade) including fusion, prognostic and decision support algorithms combining model-based and artificial intelligence (AI) techniques. This paper describes the recent advances and optimisation of the Smiths algorithms that include automatic data correction algorithms, mathematical networks and dynamic models. The algorithms have been developed to form the core of an affordable, certifiable SPHM system for legacy and modern aircraft. Therefore, following successful blind validation using legacy data covering 15 years of military operations, the algorithms have been optimised for airborne implementation. The algorithm optimisation efforts have been based on model-based knowledge, sensitivity analysis and genetic algorithms. The genetic optimisation has been targeted at data mining techniques and novel neural networks with unique activation functions that combine sigmoid, linear and inverse functions