Year: 2012
Type: dissertation
Abstract: This research investigates an approach to finding the optimal geometry of convective cooling system structures for the enhancement of cooling performance. To predict the cooling effect of convective heat transfer, flow analysis is performed by solving the Brinkman-penalized Navier-Stokes equation, and the temperature profile is obtained from the homogenized thermal-transport equation. For accurate and cost-effective analysis, stabilized finite element methods (FEM) and the adjoint sensitivity method for the multiphysics system are implemented. Several stabilization methods with different definitions of their stabilization tensors and the Newton-Raphson iteration method are introduced to solve the governing equations. This study investigates numerical instabilities, such as velocity and pressure oscillation at the fluid-solid interfaces, which result from the fact that the non body-conforming mesh for the topology optimization method fails to capture the sharp change in velocity gradient with a high Reynolds number flow. These oscillations are not problematic at the system analysis level, but prevent the design from converging to an optimized shape at the design optimization level, creating element-scale cavities near the solid boundaries. Several stabilization methods are examined for their ability to alleviate the instabilities. The Galerkin/least-square method produces less oscillation in most cases but it is insufficient in resolving the convergence issue. The density and sensitivity filters do not effectively suppress the cavities at the design optimization level, while a move-limit scheme easily prevents this instability without significant increase in computational cost. The topology optimization method is applied to the convective cooling system design, by using the same configuration that was successfully used in designing the Navier-Stokes flow system. The main design purpose is to design a flow channel to maximize cooling efficiency. A numerical issue concerning the behavior of the Brinkman penalization is presented with example designs. The optimizer frequently ignores the Brinkman penalization and creates infeasible designs. To resolve this issue, a multi-objective function that also minimizes pressure drop is suggested. As design examples, 2D and 3D cooling channels are designed by the multi-objective function, and the effect of Reynolds and Prandtl number change is discussed.
Author: Kyungjun Lee
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Keywords: topology optimization, convective cooling, heat sinks, flow channel design, Brinkman penalization, stabilized finite element method, Galerkin/least-squares stabilization, adjoint sensitivity analysis, multiphysics optimization, density filter, pressure drop minimization, multi-objective optimization, Reynolds number effects, Prandtl number effects, cooling efficiency
Open Access status: green
OpenAlex ID: W183609586
