Reduced Order Modelling of Optimized Heat Exchangers for Maximum Mass-Specific Performance of Airborne ORC Waste Heat Recovery Units

Abstract

Waste heat recovery (WHR) from aero engines via compact organic Rankine cycle (ORC) units may increase the fuel efficiency of air transportation. Heat exchangers are arguably the key components of ORC systems for aeronautical applications and their design must be optimized to guarantee the best trade-off between fluid pressure drop, weight, and induced aircraft drag. At present, no heat exchanger design guidelines are available for waste heat recovery systems aboard aircraft. This study, thus, contributes to defining a proper design methodology for ORC systems of such applications. The chosen
test case is a supercritical ORC system with cyclopentane as the working fluid, which recovers waste heat from the auxiliary power unit of an aircraft. The exhaust gas temperature and mass flow rate of the power unit are known and kept constant in the analysis, and so are the ambient conditions, which define the cold sink of the ORC turbogenerator. Three design strategies targeting the minimum mass and maximum net power output of the ORC unit have been assessed. In the first one, the multi-objective optimization is performed by prescribing a priori the geometry and frontal area of the heat exchangers.
Thus, only the cycle parameters are optimized. The second method tackles, instead, the simultaneous optimization of the geometric parameters of the condenser and the cycle parameters. It was found that the integrated design allows for system mass reduction by 10 - 12% for a given ORC power output, highlighting the importance of performing the simultaneous optimization of the thermodynamic process and the heat exchanger geometry. Finally, the third method addresses the same optimal design problem by leveraging a reduced-order model of the condenser to predict the optimal design space of this component. The generated Pareto front obtained with this method is very similar to that found by optimizing simultaneously the complete condenser geometry and the cycle parameters. The mean deviation is about 2%. With just one heat exchanger surrogate model, the Pareto front was generated in one-fourth of the computational time. This is due to the lower number of optimization variables and the faster objective function evaluation.