Abstract
We provide an algorithm to optimize the geometry of the fins in an array of longitudinal-fin heat sinks (HSs) in, e.g., a blade server, which is a prohibitively long task using computational fluid dynamics (CFD). First, banks of CFD simulations are run to precompute dimensionless thermal resistances (conjugate Nusselt numbers) as a function of dimensionless HS geometry, thermophysical properties, and external parameters. These precomputed CFD results are embedded in flow network models (FNMs) in the form of look-up tables. This preserves much of the accuracy of CFD and the speed of FNM. The FNMs are, in turn, embedded in a multivariable optimization algorithm (MVO). Our hybrid numerical algorithm is provided, and we exercise it for an example problem.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have