基于灰色关联分析的芯片布局和散热器结构参数优化

Optimization of Multi-chip Layout and Heat Sink Structure Parameters Based on Grey Relational Analysis

目的降低芯片工作温升,提升芯片的热可靠性。方法利用CFD仿真工具,搭建多芯片共用散热器的热仿真分析模型,确定不同方案的芯片结点温升。以芯片横向和纵向间距、散热器基板厚度、翅片高度、翅片厚度、横向翅片间距、纵向翅片数等7个结构参数与芯片温升之间关系为研究对象,以降低芯片结点温升为优化目标,通过灰色关联分析,筛选出主要影响因素,并利用响应面回归分析优化。结果其中4个因素的灰色关联度大于0.6,是影响芯片温升的主要因素,排序为纵向翅片数>基板厚度>芯片横向间距>翅片厚度;横向翅片间隔、翅片高度、芯片纵向间距为次要因素。进一步通过响应面分析优化获取了最终组合优化参数,芯片纵向间隔为15 mm,翅片高度为18 mm,翅片间隔为6 mm;芯片横向间距为104 mm,基板厚度为11.2 mm,翅片厚度为1.13 mm,纵向翅片数为10,芯片组最大温升为48.959℃。结论灰色关联分析能较好地用于散热多因素影响分析,与响应面回归分析相结合,可以构建出较高精度的回归预测模型,该研究为多芯片共用散热器的布局和结构方案评估和优化提供了参考。

Aiming at reducing the temperature rise of the chip and improve the thermal reliability of the chip, CFD simulation tools were used to build a thermal simulation analysis model of a multi-chip shared heat sink to determine the junction temperature rise of the chips. The relationship between the seven structural parameters(the chips horizontal and vertical spacing,the thickness of the heat sink baseplate, the fin height, the fin thickness, and the horizontal fin spacing and the number of longitudinal fins) and the chips maximal temperature rise were analyzed. The main influencing factors were screened out through grey relational analysis and optimized by response surface analysis. The grey correlation degree of the four factors was greater than 0.6, which are the main factors that affected the temperature rise of the chips. The influence order was: the number of vertical fins>the thickness of the substrate>the horizontal spacing of chips>the thickness of the fin;the horizontal fin spacing. Fin height, and vertical chip spacing were secondary factors. The final combination optimization parameters were further obtained through response surface analysis optimization: the chip longitudinal spacing is 15 mm, the fin height is 18 mm, and the fin spacing is 6 mm;the chip lateral spacing is 104 mm, the substrate thickness is 11.2 mm, the fin thickness is 1.13 mm, the number of longitudinal fins is 10, and the maximum simulated temperature rise of the chipset is 48.959 ℃. Grey relational analysis can be better used to analyze the influence of multiple factors on heat dissipation, and build a high-precision regression prediction model when combining with response surface analysis. This research can provide reference for the evaluation and optimization of the chips layout and structure of the multi-chip shared heat sink.