Prognostics of gas turbine engine: An integrated approach

Condition-based maintenance is an emerging paradigm of modern system health monitoring, where maintenance operations are based on diagnostics and prognostics. Condition-based maintenance strategies in the industry benefit from accurate predictions of the remaining useful life (RUL) of an asset in order to optimise maintenance scheduling, resources and supply chain management. Due to the substantial costs involved, small improvements in efficiency, result in the significant cost reductions for overall maintenance services as well as its impact on energy consumption and the environment. In this paper, we present a data-driven methodology combining the hierarchical Bayesian data modelling techniques with an information-theoretic direct density ratio based change point detection algorithm to address two very generic issues namely dealing with irregular events and dealing with recoverable degradation, which are often encountered in the prognosis of complex systems such as the modern gas turbine engines. Its performance is compared with that of an existing Bayesian Hierarchical Model technique and is found to be superior in typical (heterogeneous) and non-typical scenarios. First, the technique is illustrated by an example on the simulation data and later on, it is also validated on the real-world civil aerospace gas turbine fleet data.