Here is the abstract you requested from the Thermal_2010 technical program page. This is the original abstract submitted by the author. Any changes to the technical content of the final manuscript published by IMAPS or the presentation that is given during the event is done by the author, not IMAPS.
|Design Optimization Framework for Integrated Cooling Systems|
|Keywords: Heat Sink, Optimization, Design|
|Conventional air-cooled heat sink performance is becoming inadequate to meet the growing thermal loads and heat fluxes produced by modern electronics systems while still maintaining required component temperature limits. Furthermore, air-cooling systems are generally assembled from individual components (extended surface heat sink, air mover, blower motor, etc) that are designed and optimized independently. This approach neglects coupling between components, frequently resulting in non-optimal system performance. Enhancing system performance to meet growing in cooling demand requires an integrated approach to maximize performance at the cooling system level and exploit interactions between components within the system. A modeling framework has been developed that incorporates all key elements of an air-cooling system. This approach has been used to identify a cooling system capable of decreasing heat sink thermal resistance by 2X for a system with equivalent input power. In general, thermal systems can include a large number of continuous and discrete parameters and performance that exhibits both non-linear and non-convex behavior. This results in a solution space that contains local minimas that prevent the identification of global optimum solutions and infeasible zones that inhibit the use of computationally-efficient gradient-based methods. In the present work, discrete and non-convex performance effectors are identified and used to drive an optimization framework that combines gradient-based and genetic algorithms in a computationally efficient and robust manner. The optimization methodology includes parameterization strategies for heat sink surface geometry and component performance. To preserve continuity of the objective function, slack variables are used to allow for the closure of infeasible designs but also provide the necessary gradient information to drive these intermediate solutions out of infeasible regions. Experimental results will be shown that validate the modeling methodology.|
|Brian St. Rock,
United Technologies Research Center
East Hartford, CT