Abstract
We present a parametric approach for heat exchanger shape optimization utilizing Deep Reinforcement Learning (Deep RL) and Boundary Representation (BREP). In this study, we show that continuous geometric representation of the fluid and solid domain facilitates the implementation of boundary conditions and design space exploration in contrast to traditional Topology Optimization such as density-based methods. The proposed framework consists of a Deep Neural Network (DNN), a Computational Fluid Dynamics (CFD) solver with an automatic body-fitted mesh generation to solve a single fin shape optimization. The learning is performed using Proximal Policy Optimization (PPO) in combination with a CFD environment in FEniCS. The RL agent successfully explores the design space and maximizes heat transfer and minimizes pressure drop for geometric design with as low as 12 degrees of freedom represented by composite Bézier curves. Higher degree of freedom results in higher reward of the agent. This method alleviates the curse of dimensionality compared to voxel and pixel-based optimization of coupled thermal fluid-structure. Results show the manufacturability and efficiency of the output of our framework. Over 30 percent improvement in overall heat transfer while lowering the pressure drop by more than 60 percent compared to the rectangle reference geometry is achieved.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
