Shape Optimization of Pin-Fins by Free-Form Deformation and Active Learning Surrogate Model

Abstract

The shapes of pin-fins play an essential role in enhancing overall and local forced convective heat removal. This study presents a new design optimization scheme based on the free shape deformation method, which gives rise to nonuniform pin-fin shapes depending on local flow structures interacting with each local pin-fin. The grid of pin-fins is embedded in a parameter space formed by basic function and control points. By moving the position of the control points, the shapes of the pin-fins and the neighborhood grid are deformed during optimization. The optimization objective function is the heat transfer coefficient of the pin-fin arrays at a limited pressure drop to ensure cycle efficiency. In addition, to reduce the computational resources, the active learning surrogate model for constraint global optimization is adopted, which is combined with computationally expensive computational fluid dynamics simulations. To exemplify the effectiveness of the proposed method on optimizing the pin-fin shape, two-dimensional pin-fins in inline arrangements are considered. Finally, an improvement in the heat transfer of 33.4% is achieved as compared to the baseline. Therefore, the proposed method of pin-fin optimization is effective in improving thermal performance.