As turbofan technology advances, periodic engine inspection and maintenance still remain a significant part of aircraft operational costs. Operators are thus looking to engine condition-based maintenance (ECBM), leveraging sensor data and numerical engine models for continuous diagnostics. KLM Engine Services developed a surrogate model based on the High Dimensional Model Representation approach, capable of processing a large volume of engine data at a lower computational cost to estimate engine component condition. This study proposes enhancements to expand the model's capabilities, incorporating additional engine parameters and extending the operational envelope. The augmented surrogate model was then combined with a Long Short Term Memory network capable of predicting component condition based on trends generated from the surrogate model. The framework proposed has demonstrated the potential advantages of combining the surrogate and prediction model for engine diagnostics and prognostics, serving as a valuable starting point for future ECBM projects at KLM ES.

Over the years, aero engines have evolved into the efficient turbofans present on commercial airliners today. Although these engines are very reliable, they still experience degradations in performance and reductions in component health. The degradation affects the operation of the engine, which can be measured using the pressure, temperature, and rotational speed sensors. Using accurate engine models together with engine sensor measurements, the condition of individual components can be determined. Surrogate models of a non-linear Gas Path Analysis model have been developed for the GEnx-1B turbofan in the form of High Dimensional Model Representations. It is determined that the surrogate models developed are able to determine component conditions with a high accuracy and low computational complexity. The models are able to properly identify the effects of water-washes and turbine failures when applied to real-world on-wing data. Due to the low computational complexity of the models, they provide the possibility for fleet-wide continuous engine diagnostics.