Probleemstelling:

Design optimization is a trending engineering topic. It involves improving designs towards specific goals, with the aim of increasing performance. In the design of cooling systems, a much-asked question is which pin fin heat sink is the best for a certain application, e.g. the cooling of a processor unit. Pin fin heat sink are simple, yet tuning the number of fins and the shape of the fins can give large performance difference. The standard question is which shape and how many fins to use for the problem at hand.

This is a case of parametric optimization, where a limited set of parameters describes the problem to be solved. More specifically, this is a mixed integer linear programming problem as the number of fins is an integer, while dimensions can be continuous. Specific algorithms and optimization packages exist to solve this kind of problems.

As outlined, parametric optimization can give simple answers to simple problems. However, there are several challenges in making its results relevant:

How to best parametrize the design of heat sinks, also if they are more complex than the one above (e.g. skewed fins and how skewed they are)
How to still be computationally efficient without using a brute-force approach
How to fully automate the whole process, from design generation to CFD analysis and interpretation of the results.

Doelstelling:

For three selected test cases, quantify the performance gain of a parametrically optimized design w.r.t. a reference design. In four steps, this thesis will address the challenges outlined above. First, you will look into standard heat sinks designs as well as some designs of Diabatix (which are typically more complex) and how to properly parametrize them. Literature can be a guide here, but it will probably be necessary to test some of the approaches.

Second, the parametrization will give a clue on which algorithms best to apply. This will require a thorough study of available algorithms and some testing as well. Third, the whole process is automated to allow for easy testing. This means a.o. a design will have to be generated based on the input parameters. Lastly, the whole chain will be tested, which is expected to uncover several limitations and areas of improvement.

This project aids in achieving sustainable development goal (SDG) 12: responsible consumption and production, by improving performance and energy efficiency in industry.