CFD-05. Processor cooling. Boundary layer.

We continue with the posts related to Computational Fluid Dynamics (CFD). In this case we are going to carry out a benchmark of a common example in thermal analysis using CFD, the study of the cooling of a processor. We will use Acusolve with the pre-processor Hyperworks CFD. The case study can be seen in the image below.

Schematic diagram of a processor cooling system

CFD Model

As explained in the tutorial, available in our Downloads section, the problem is going to be solved considering the symmetry conditions in order to study only a practically two-dimensional domain of height h/2. Appropriate inlet and outlet sections are added to allow the stabilisation of the flow. This leaves the computational domain as shown in the following image with a thickness of 5mm to generate a 3D mesh to solve the problem with Acusolve.

Geometry of the CFD model to be simulated

As boundary conditions, symmetries are placed on all faces except the one corresponding to the processor board, which is located in the lower part of the central area, of length L. A constant target temperature of 300K will be modelled on this board. In this way we can calculate the thermal dissipation achieved by the system. Another option to study the cooling of a processor would be to impose a dissipated power on it and study at which temperature the component stabilises.

Mesh and Boundary Layer

In this analysis, in order to capture heat transfer accurately, it is essential to create an adequate boundary layer in the mesh. The tutorial linked to this post shows the process for estimating an initial element size and growth rate to capture the phenomenon. In Acusolve the boundary layer is defined as a mesh control, similar to a refinement zone but with different parameters.

Processor cooling: Results

After simulating the model, the results show the boundary layer in both the velocity and temperature profiles. These are presented in the following image.

Temperature and speed results in chip cooling study

This simulation is compared with the analytical results considering the dissipated power result in the processor of about 91mW. The average outlet temperature obtained from the analytical calculation fits well with that predicted by the simulated model. The tutorial concludes with the calculation of the Nusselt number of the heat exchanger and a brief study of the forces caused by the flow on the plates, a useful value in other types of analysis.

In the following CFD-related post, we will discuss the different turbulence models commonly used in CFD. To do so, we will look at a common benchmark in this type of simulations, the prediction of the reverse flow after a step.


Related entries

Resultados de una simulación CFD con modelos híbridos y transición de flujo

CFD-14. Flow transition with hybrid turbulence models.

02, Jun, 2023
Imágenes de la simulación híbrida LES-RANS con modelo IDDES

CFD-13. LES-RANS Hybrid Turbulence Models

26, May, 2023
Imagenes de una simulación LES con y sin turbulencia artificial en el inlet

CFD-12. LES simulation. Addition of Artificial Turbulence.

19, May, 2023