This paper presents an investigation to improve the design of a generic fin heat exchanger, using novel discrete adjoint solver tools in the computational fluid dynamics (CFD) software FLUENT. A baseline design is analyzed initially to evaluate flow resistance and heat transfer. Optimization is conducted by deploying the adjoint solver. The heat load and the drag force are combined into an objective function using a Reynolds analogy approach. Sensitivities of the objective function to geometric changes are predicted by the adjoint, and then the mesh is morphed, and the predictions are verified by the full CFD solutions. Predetermined, engineering driven, geometric changes are explored and compared, and the range of validity of the predictions is evaluated. An algorithm is then developed to implement steepest descent, constrained optimization based on the adjoint solution. The algorithm is applied iteratively on the fin heat exchanger, and a comparison is performed between the change in objective function predicted by the adjoint, and that calculated in full CFD solutions on morphed meshes. The insight gained on the directions of design changes and attending quantitative improvement of the design objective function is very useful to guide the optimization process. This is enabled by the adjoint solver’s capability to robustly evaluate the sensitivities of the objective function to all solution variables, and predict changes in observables.

This content is only available via PDF.
You do not currently have access to this content.