Multi-objective Optimization of a Heat Exchanger Using the Discrete Adjoint Method

A 2D model of a pin-fin heat exchanger was optimized using the discrete adjoint method in ANSYS Fluent 16. The initial heat exchanger shape was modeled as staggered cylinders in a cross-flow. Two observables were monitored during the optimization cycles: the heat transfer and the pressure drop, and the objectives are the maximization of heat transfer and the minimization of pressure drop. However, improving the performance of the heat exchanger poses its own challenges since the heat transfer and pressure drop are usually two contradicting observables. In order to successfully improve both observables, single objective and multi-objective shape optimizations were studied. Both single and multi-objective optimizations were conducted under steady laminar flow conditions at Re = 10 and Re = 100. The single objective optimizations were done for different step sizes of the geometry change, e.g. different changes of pressure drop or heat transfer. While the optimized observable was set to improve linearly, the other unconstrained observable shows a nonlinear deterioration. The multi-objective optimizations were performed for different weight factors, leading to different end shapes. For the final optimized geometry, we could achieve up to 11% reduction in pressure drop and 11% increase in heat transfer.