Fin Design Topology Optimization for Direct Liquid Cooling of Multi-Chip Power Modules

This article demonstrates the use of topology optimization (TO) to generate optimal fin designs capable of high heat flux removal without increasing the pumping power. We use pseudo-3D modeling validated against 3D simulations for thermal-hydraulic performance, which reduces the computational intensity and allows for a wider range of designs to be considered. The accuracy of the pseudo-3D modeling is aided by allowing the heat transfer coefficient to vary locally with the flow conditions, in contrast to other studies that often assume constant heat transfer coefficient independent of surface shape. We generated TO designs with density fields representing six different conventional fin designs including straight, zigzag, and pin fins of different dimensions along with the traditional constant intermediate density field. TO designs outperformed their conventional counterparts by delivering lower chip temperatures for similar or lower pressure drops at a fixed flow rate. In addition to the lower chip temperatures, the difference of average temperatures for the chip rows along the flow direction was 76.3 % lower on average for the TO designs compared to the conventional designs. We produced and tested prototype finned base plates using conventional straight fin design and a TO design which showed similar pressure drop to the straight fin design in 3D simulations. Heating was provided using ceramic resistor chips as surrogates for power semiconductor devices. Experimental testing validated the predictions for device temperatures and pressure drops. The experiments demonstrate the effectiveness of the TO design which gave a 12.3 % (from 0.47 to 0.41 °C.cm
²
/W) reduction in thermal resistance and 81.8 % (from 2.75°C to 0.49°C) lower difference in chip temperatures along the flow direction compared to the conventional design. The performance improvements were consistent for the range of flow rates in laminar flow regime for which the optimization was performed as well as turbulent flow regime.