| With the rapid development of science and technology,such novel technologies as artificial intelligence,industrial Internet,and big data centers have emerged in succession.In this regard,chips have played a vital role in the advancement of technology.According to Moore’s Law,the number of components on an integrated circuit doubles every 18 to 24 months,as does their performance.At the same time,with the rapid increase in the integration of chip transistors,the overall amount of power consumption will rise sharply,which places higher requirements on the structure and heat dissipation performance of the thermal management device.Against this background,the optimal design of chip heat dissipation device has now become a research hotspot across the industry.Compared with shape optimization and parameter optimization,topology optimization allows the highest level of discretion in design.Given the bounds,it is even possible to design the heat sinks that are close to optimal heat dissipation performance.As the traditional topology optimization of chip heat sinks relies on a specific design domain,the shape of the design domain determines the upper limit on the heat dissipation performance of the optimized heat sink.When the optimized design domain is unspecified or partially specified,it is worth considering how to determine the design domain with a higher upper limit of optimization.In addition,the sticking point of topology optimization lies in the huge workload of computation and the difficulty to achieve convergence.Especially for the topology optimization in three-dimensional space or when convective heat transfer is taken into account,the calculation is likely to be a lengthy process.Therefore,it is necessary to impose limit on the process of topology optimization.For example,only two-dimensional space or only heat conduction is considered to reduce the difficulty in optimization and to improve the convergence of calculation.In this paper,the topology optimization design of the heat sink is studied under three different circumstances.The first one is that convection is considered in two-dimensional space,the second one is that convection is not considered in two-dimensional space,and the lat one is that convection is not considered in three-dimensional space.The main points of this research are as follows:Firstly,topology optimization is performed on different types of heat sinks in two-dimensional space.Through the topology optimization performed on heat sinks when convection is considered or not in two-dimensional space,the effect of mesh and penalty factors on optimization results is explored.It is discovered that the optimization results are irrelevant to grid when the maximum grid element in the design reaches 3 mm,and that the optimization results are the best in bifurcation with the fewest intermediate value areas when the penalty factor is 3.Compared with the optimized heat sink considering convection,the optimized heat sink without convection shows more and denser bifurcations,the interface with the air is more visible,and the intermediate value is smaller.The computation time without convection is only one third that when convection is considered,and the outcome of convergence is better.In addition,topology optimization is performed for liquid cooling runners.There are two objective functions designed.The minimum pressure drop is effective in ensuring the fluidity of the runner,while the maximum outlet temperature is sufficient to ensure the most satisfactory heat dissipation effect of the runner.According to the temperature cloud map and streamline diagram,the heat dissipation mechanism is discussed in depth.Secondly,a study is conducted on the design of topology optimization for the three-dimensional heat sink.In order to determine the design domain with a relatively high upper limit of optimization,topology optimization is designed for three design domains including cylindrical,forward conical and inverted conical as well as for the design domains of different heights,respectively.Furthermore,the heat transfer mechanism is investigated in depth.It is found out that when the design domain is wide at the bottom but narrow at the top,the heat dissipation effect of the optimized heat sink improves,and that the height of design domain has effects mainly on the shape of heat dissipation bifurcation.The higher the design domain,the finer and fewer the thermal bifurcations.SLM 3D printing technology is applied to produce three types of optimized heat sinks,namely cylindrical,forward-conical and inverted-conical,with 3 W and 5 W LED chips taken as heat source for the design experiments.It is discovered that compared with traditional fin heat sinks,the performance of the topology optimized heat sink is significantly better.More specifically,the positive conical heat sink is the best,the cylindrical heat sink ranks second,and the inverted conical heat sink is the worst.It is illustrated that the optimal optimization design domain is close to the shape of a hemisphere,which maximizes the extraction of heat from the chip,while reaching balance between the major channels and branches of heat dissipation.This study provides reference for the industrial-scale application of topology optimization. |
