بررسی عددی و تجزیه و تحلیل حساسیت از کولرهای میکرو کانال مانیفولد

This paper presents a numerical investigation of a single-phase manifold microchannel cooler (MMC) heat exchanger demonstrating a reduction in fluid pressure drop while improving chip-temperature uniformity. This modeling work includes the entire manifold length with multiple microchannels, whereas previous models have only focused on individual microchannels, ignoring complex manifold effects. Computational Fluid Dynamic (CFD) models were used to identify the impact of varying both the manifold and microchannel fin and channel dimensions, and a sensitivity analysis was performed with respect to system pressure drop, rise in device temperature, and thermal uniformity. This modeling work demonstrated both large velocity gradients between microchannels, as well as fluidic swirling in the microchannels that significantly improved the heat transfer coefficient. These results are absent from unit-cell type models. The results of the full MMC model showed significantly improved chip-temperature uniformity when large (approximately 10X) differences in velocity occurred between microchannels. The simulations also showed that, for equivalent thermal performance, the MMC design resulted in a 97% reduction in system pressure drop when compared to an equivalent straight microchannel cooler. Finally, the numerical pressure drop results were compared to a simpler, one-dimensional approximation based on the Hagen–Poiseuille equation. While under-predicting total pressure drop, the analytical equation does capture prevailing trends of the effects of channel dimensions on the pressure drop and can be used for rapid evaluation of numerous tradeoffs from a system perspective.

The electronics industry is moving to smaller products with increased functional density. This equates to electronics with higher heat generation requiring more effective cooling. Power electronics have heat fluxes approaching 1000 W/cm2[1]. Without proper cooling, the increased heat fluxes will result in higher device operating temperatures, which can adversely affect device performance [2]. These high heat fluxes also require the use of forced convection liquid cooling [3]. Microchannel systems are a desirable cooling solution because they have been shown to reduce overall system size and improve the heat transfer coefficient [4]. But the two disadvantages to a microchannel system are typically the high pressure drop caused by small hydraulic diameter channels and non-uniform cooling caused by the fluid absorbing heat along the length of the channel. Many researchers have looked into ways of reducing these effects: multi-layered microchannel heat sinks [5] and [6], single plane manifold configurations [7] and [8], and fractal channels [9] and [10]. While each of these designs looks to improve the problems of microchannel cooling, they each have disadvantages. The multi-layered microchannel heat sink still has high pressure drops due to the small hydraulic diameter channels and both the single plane manifold and the fractal channels have much of the heated surface cooled by lower performance macrochannels instead of microchannels. The solution analyzed in this paper for solving the microchannel problems is the manifold microchannel cooler (MMC) [11] shown in Fig. 1(a), with the dashed box indicating the active cooling area. The fluid volume is represented by the non-transparent grey volume and the solid volume is shown in the figure as partially transparent. The fluid flow path is shown in both Fig. 1(a) and (c). The MMC incorporates a two-stage flow designed to take advantage of both large and small channels. The fluid flows in through large, long channels located away from the heated surface (manifold channels), then is forced through much smaller perpendicular channels in contact with the heated surface (microchannels), and then flows out through another large manifold channel. This design reduces total flow restriction since the majority of the flow occurs in the larger manifold channels. It also maintains the cooling potential of microchannels since they can cover the entire heated surface. The design also improves chip temperature uniformity by reducing the effects of heating along the microchannel length. This can reduce the chance of mechanical failures caused by in-plane spatial temperature gradients. The primary drawback of the MMC desing is the additional complexity and volume associated with the addition of the manifold channels. Full-size image (43 K) Fig. 1. Schematics of the (a) overall MMC structure with the dashed box indicating the actively cooled area, the top arrows showing the fluidic path, the black region showing the modeled portion, and the microchannel length used in the models pointed out (b) MMC structure indicating the critical dimensions with the solid grey area representing the fluid (c) fluidic volume of the modeled structure showing the flow paths and the three pressure drops used to calculate the analytical pressure drop approximation. Figure options The critical dimensions for the MMC are shown in Fig. 1(b) with “m” referring to the microchannel dimensions and “M” referring to the manifold dimensions. The three subscripts: h, f, and w refer to the height, fin width and channel width, respectively. There have been several numerical studies of the MMC design that have verified the benefits of the design. Harpole and Eninger [11] were the first to propose the MMC design through a numerical study which looked to optimize the microchannel dimensions. However, their study did not look to optimize the effect of the manifold channels or consider their effect on the pressure drop. Copeland published a numerical study on the MMC design using a single microchannel unit cell [12] which is highlighted by the boxed area in Fig. 1(c). The unit cell includes a single half-channel of a single microchannel section with the fluid flowing in and out of the bottom, symmetry is assumed on all sidewalls, and a constant velocity and temperature assumption is applied at the inlet. A number of models were run to see the effects of various microchannel parameters on the performance. Poh and Ng [13] and [14] expanded on the same unit cell model as Copeland, showing similar results. Each of the previous models has been limited to the microchannel unit cell, which fails to capture the effects of the manifold portion of the device. Another numerical study by Ryu did include the manifold portion under a single microchannel unit cell [15], but did not look into the effects of pressure drop or flow maldistribution. Additionally, the work by Bejan and Errera [16] presented an analytical solution to optimizing a 2D array of fractal type channels in terms of both thermal performance and pressure drop. While the geometry is different, this is another valid approach to optimizing geometries if it was expanded to 3D. To the authors’ knowledge, no modeling has been done to date looking at the entire length of a manifold to see the effects the manifold has on the plurality of channels. This work presents the first modeling of a full MMC structure to elucidate the complicated relationships between dimensions, thermal performance, and pressure drop. Primary performance metrics remain chip temperature and pressure drop which are typically competing factors due to the fact that smaller hydraulic channels typically have higher thermal performance but increased pressure drop. The following sections first develop analytical expressions to approximate the pressure drop in MMC systems to guide the approximate design space, followed by a complete numerical solution for a single geometry. Finally, a parametric study is performed on various MMC dimensions.

Due to the inherent problems with straight microchannels, the MMC design is a promising alternative to reduce the pressure drop and allow for better device temperature uniformity. This work has shown that in contrast to assumptions made in previous models, the microchannel portion of the MMC cannot be adequately modeled as individual unit cells. Modeling the entire manifold structure is necessary to account for the large deviations in microchannel velocities along the length of the chip which affect the maximum chip temperature location and the thermal gradients. A swirling motion in the microchannels was also shown that has not been seen in modeling before and has been shown to improve the heat transfer coefficient by 50%. The numerical results were compared to analytical equations developed from the Hagan–Poiseuille pressure drop equations showing that the analytical equations, while underpredicting the pressure drop, can be used to give a reasonable approximation of the impact of various dimensions on the pressure drop and can be used for design and comparative purposes. The results of a single geometry were analyzed in depth to show the effect of flow rates on the system pressure drop, thermal performance and flow conditions. A MMC structure was evaluated against a comparable straight microchannel design and showed that for a similar thermal performance, there was a 97% improvement in pressure drop. The numerical sensitivity study has shown that Mh should be designed as large as possible within manufacturable and size limitations, Mf should be minimized, and Mw should be determined based on the intended application. Numerical results have shown that mw has a larger impact on the thermal performance than mf As the chip length increases, both the pressure drop and temperature increase.