Data-driven modeling of high-speed centrifugal compressors for aircraft Environmental Control Systems.

The Environmental Control System (ECS) is the main consumer of non-propulsive power onboard aircraft. The use of an electrically-driven Vapor Compression Cycle (VCC) system, in place of the conventional air cycle machine, can lead to a substantial increase of the coefficient of performance. This work documents the development of an integrated design optimization method for VCC-based aircraft ECS, where the sizing of the system is performed along with the conceptual design of the compact heat exchangers and the high-speed centrifugal compressor. A data-driven model of the compressor has been developed to reduce the complexity of the VCC system model and the computational cost of the associated optimization problem. The model is based on artificial neural networks and has been trained on a synthetic dataset of 165k centrifugal compressor designs, generated with an in-house tool. The case study selected to demonstrate the capabilities of the proposed methodology is the multi-objective design optimization of an electrically-driven VCC system for the ECS of a single-aisle, short-haul aircraft, flying at cruise conditions. The results show that the number of function evaluations needed to identify the Pareto front reduces by a factor of three when using the data-driven model, in place of a meanline method. At the same time, the robustness of the numerical solver is improved, leading to the identification of optimal solutions covering a wider design space. Finally, the proposed methodology enables the analysis of the trends established between the system performance metrics and the design of the individual components. • The development of a data-driven model for centrifugal compressors is presented. • A novel integrated design optimization framework for aircraft ECS is developed. • The multi-objective optimization of an electrically-driven VCC system is documented. • The computational cost reduces by a factor of three when using the data-driven model. • The robustness of the numerical method is improved when using the data-driven model. [ABSTRACT FROM AUTHOR]