MATHEMATICAL NETWORKS FOR THERMAL TRANSIENT AND NON-TRANSIENT PROGRESSIVE FATIGUE OF ENGINE COMPONENTS

Engine components can experience varying centrifugal loads, gas loads, oxidation, microstructure transformation at high temperatures and str esses induced by temperature gradients. The life consumption of hot en gine components depends not only on these factors but also on the time spent at constant-amplitude loads. The damage mechanism of engine com ponents is therefore complex and requires formidable models. These mod els are not suitable for fatigue management or on-board systems becaus e of their high computational costs. There is a need for efficient sim ulations that can accurately portray this complex damage mechanism and , at the same time, can be embedded in fatigue management and on-board systems. Mathematical networks were developed to fulfil this need and successfully synthesized the fatigue damage of aircraft structural co mponents from flight parameters. In this paper, the feasibility of tra ining the mathematical networks to synthesize fatigue of engine compon ents is demonstrated. The mathematical attributes of the networks were based on information supplied by Rolls-Royce. The networks' training mechanism was targeted at the minimization of errors in synthesized ac cumulative damage values. The mathematical networks synthesized the ac cumulative fatigue damage of three engine components successfully. One component was subject to non-thermal transient stresses and two compo nents were subject to thermal transient stresses.