【作者】 曾涛；

【导师】 王琥；

【作者基本信息】 湖南大学 ， 机械工程， 2020， 硕士

【摘要】 随着电子元器件的尺寸不断减小,热功率密度不断增加,传统散热结构的发展遇到了瓶颈。拓扑优化作为高自由度的结构优化方法,相同模型的设计参数远多于尺寸优化,形状优化,形貌优化。尤其随着计算机技术的发展,拓扑优化也逐步应用于多物理场耦合,三维结构等大规模计算的结构优化。本文的仿真计算基于COMSOL计算平台。瞬态模型的验证引进了RC电路模型,该模型已被实验数据验证,并能较为精确地描述瞬态模型热源区域的平均温升。在热流耦合模型拓扑优化过程中,通过光滑Heaviside投影策略、偏微分方程过滤等方案,提高了热流耦合模型拓扑优化的稳定性,并降低了局部最优对优化结果的影响。在伪三维热流耦合模型拓扑优化过程中,通过引入人工换热系数以代替流体通道内散热结构在厚度方向的平均换热系数。伪三维热流耦合模型拓扑优化的精度是通过引入指标g来测定伪三维模型和三维模型芯片的温度分布差。热流耦合模型的仿真计算涉及流体流动、热传导等物理场的求解计算。因此,热流耦合模型拓扑优化可以有效针对散热结构进行优化。本文分别研究了稳态、瞬态热流耦合模型拓扑优化在芯片的散热结构优化中的应用,用以获得较优的散热性能。由于实际的散热元器件为三维结构,而三维热流耦合模型拓扑优化的计算量非常巨大,如具有125000个设计变量的三维强制对流模型的拓扑优化经过300个优化迭代共耗时304小时。本文引入人工换热系数,使得伪三维模型能较为精确地描述三维模型的流体流动和温度场分布,在保证一定精度的同时又能避免计算三维热流耦合模型高昂的计算成本,能在保证一定精度的同时又能使计算效率提升约50倍。其中,本文着重研究了瞬态的伪三维风冷和水冷模型的拓扑优化,并对其结果进行分析,发现瞬态拓扑优化模型在一定的时间段内能有效降低热源的平均温度:风冷模型降低了3.5K,水冷模型降低了5K,且在相同的工况下,水冷优化结构相对于参考结构泵功率下降了66.7%。更多还原

【Abstract】 As the size of electronic components continues to decrease and the thermal power density continues to increase,the development of traditional heat dissipation structures have encountered bottlenecks.Topology optimization,as a structural optimization method with higher degrees of freedom than the same problem of size optimization,shape optimization,and shape optimization.Especially with the development of computer technology,topology optimization is also gradually applied to the optimization of large-scale computing structures such as multi-physics coupling and three-dimensional structures.The simulation calculation in this thesis is based on the COMSOL computing platform.RC circuit model was introduced in the verification of the transient model,which has been verified by experimental data and can relatively accurately describe the average temperature rise of the heat source region of the transient model.In the process of topology optimization,through the smooth Heaviside projection strategy,partial differential equation filtering and so on,the stability of topology optimization of thermo-fluid model is improved,and the influence of local optimization on the optimization results is reduced.In the process of pseudo three-dimensional thermofluid topology optimization,the artificial heat transfer coefficient is introduced to replace the average heat transfer coefficient in the thickness direction of the heat dissipation fins in the flow channel.The accuracy of the pseudo-three-dimensional thermo-fluid topology optimization is to measure the temperature distribution difference between the pseudo three-dimensional model and the three-dimensional model chip by introducing the index g.The numerical calculation of the thermo-fluid model involves the calculation of physical fields such as fluid flow,heat conduction.Therefore,the thermo-fluid topology optimization can effectively optimize the heat dissipation structure.This paper studies the application of topology optimization for the steady-state and transient thermo-fluid model.Since the actual heat dissipation components are threedimensional structures,the calculation cost of the three-dimensional thermo-fluid model is very high.Therefore,the introduction of artificial heat transfer coefficients allows the pseudo three-dimensional model to relatively accurately describe the fluid flow and temperature field distribution in the corresponding three-dimensional model,while ensuring a certain accuracy and avoiding the high computational cost of the three-dimensional thermo-fluid model.The calculation efficiency of the pseudo threedimensional model increases by about 50 times of efficiency of three-dimensional model with a reasonable accuracy.Among them,this thesis focuses on the topology optimization of the transient pseudo-3D three-dimensional air-cooled and water-cooled models and analyzes the results.It is found that the transient topology optimization model can effectively reduce the average temperature of the heat source within a certain period of time: The air-cooled model is reduced by 3.5K,the water-cooled model is reduced by 5K,and under the same operating conditions,the pump power of the watercooled optimized structure is reduced by 66.7% relative to the reference fin structure.更多还原