Overview
This project performed a detailed steady-state conjugate heat transfer (CHT) analysis of a pin-fin aluminium heat sink for a 125 W CPU. The aim was to optimise fin geometry to minimise junction temperature while maintaining an acceptable pressure drop across the heat sink.
Problem Statement
The baseline heat sink design was running the CPU at 91°C junction temperature under full load — 16°C above the 75°C design target. The task was to redesign the fin geometry and spacing to bring the temperature within spec without significantly increasing fan power requirements.
Simulation Setup
| Parameter | Value |
|---|---|
| Software | ANSYS Fluent 2023 R1 |
| Heat dissipation | 125 W |
| Inlet air velocity | 2.5 m/s (forced convection) |
| Ambient temperature | 25°C |
| Fin material | 6061-T6 Aluminium |
| Base material | 6061-T6 Aluminium |
| Mesh elements | ~2.1 million (polyhedral) |
Methodology
CFD — ANSYS Fluent
- Used the k-ε realizable turbulence model with enhanced wall treatment
- Applied a coupled pressure-velocity scheme for steady-state convergence
- Modelled conjugate heat transfer between solid fins and airflow
- Mesh independence study across 3 refinement levels to ensure solution accuracy
Parametric Study
Swept three key geometric parameters:
- Fin height: 20–40 mm
- Fin pitch: 2–5 mm
- Fin thickness: 1–2 mm
A full-factorial DOE with 27 configurations was run and results were response-surface fitted in MATLAB.
Results
| Design | T_junction (°C) | ΔP (Pa) |
|---|---|---|
| Baseline | 91 | 12 |
| Optimised | 73 | 18 |
| Improvement | −18°C | +6 Pa |
The optimised design used:
- Fin height: 35 mm
- Fin pitch: 2.5 mm (staggered arrangement)
- Fin thickness: 1.2 mm
The staggered pin-fin arrangement disrupts the thermal boundary layer between fins, significantly boosting convective heat transfer.
Structural Validation
A brief thermal-structural FEA confirmed that thermal stresses in the fins remained below yield at steady-state operating temperatures.
Key Takeaways
- Staggered fin arrangements consistently outperformed inline layouts by 8–12% in heat transfer at comparable pressure drops
- Mesh quality (especially y⁺ near walls) had a significant impact on predicted Nusselt numbers
- Response surface methods made the large design space tractable with limited compute time