Design of heat exchangers via a bioinspired topology optimization framework with physics-informed underlying fields

A topology optimization framework with physics-informed fields is presented to bridge the gap between implicit and explicit bioinspired techniques. The methodology takes advantage of the reduced optimization space unique to the explicit analysis while intelligently adjusting the topologies through underlying fields. A heat exchanger design problem was used to illustrate the effectiveness of the proposed methodology. The objective was to minimize the average temperature while remaining below the pressure drop of a traditional parallel channel design. Pressure and velocity fields from an open cavity design domain were used to control the length scale and orientation, respectively, of the topology generation process. This resulted in designs that had a minimal pressure drop while following the natural flow path of the heat exchanger. The Darcy flow model was used to permit a rapid analysis during the optimization routine, while the final designs were validated using a high-fidelity RANS model. Four different design cases were run to observe the effect that the underlying fields had on the optimization process. The cases ranged from no underlying fields and a three-parameter optimization space to two underlying fields and a five-parameter optimization space. A Bayesian optimization model was used to explore and exploit each parameter space. Ultimately, the two-field design case highlighted the value of the physics-informed fields as it produced the top performing heat exchanger design. Compared to the traditional parallel channel design, it reduced the average temperature by 68% while maintaining a similar pressure drop.