| Résumé | Cooling channels are a crucial component of many processes which rely on controlled and efficient heat transfer, such as die casting, which constitutes a major large-scale fabrication method of metallic parts. Yet, the design of the cooling channels is still limited by the constraints of classical subtractive fabrication methods. The advent of additive manufacturing paradigm for metallic parts opens the door to cost-effective cooling channel designs via topology optimization techniques. By selecting and assessing proper interpolation and projection schemes, we propose a density-based, adjoint approach for the 3D Navier–Stokes-energy equations. This work complements the literature by considering geometries with heated cavities, representative of die casting molds, by comparing two different objective functions: the domain-averaged and surface-averaged temperatures (resulting in drastically different designs) and by discussing validation aspects such as the calibration of the Darcy friction coefficient and the comparison of body-fitted vs density-based solutions of the optimal design. Furthermore, the sensitivity of the optimal design to the hyper-parameters of the problem, i.e. fluid volume constraint, density filter radius and pressure penalization coefficient, is showcased in low- (100) and high-Reynolds (1000) laminar regimes. The pitfalls of improper tuning of these hyper-parameters are exhibited and it is shown that the methodology can achieve 52% improvement in cost function with regards to a straight-drilled baseline design. |
|---|
