Session: Session 02-05: Two Phase Cooling - II

Paper Number: 171460

171460 - Integrated Experimental, Numerical, and Machine Learning-Based Optimization of Two-Phase Skived Fin Cold Plates 

This study provides a comprehensive evaluation of two-phase flow and heat transfer performance in skived fin cold plates, integrating both experimental measurements and numerical analysis. The primary aim is to identify the most thermally efficient design by systematically investigating the influence of geometric parameters and operating conditions on cold plate performance.

Nine cold plate samples were fabricated and tested, each featuring distinct geometric configurations. All samples maintained a uniform base footprint of 1 inch × 1 inch, while varying in key design parameters: number of channels (19 to 91), fin height (2.5 to 5 mm), fin thickness (0.008 to 0.4 mm), and channel gap width (0.16 to 1.15 mm). Refrigerant R134a was used as the working fluid. Experiments were conducted at three volumetric flow rates—0.5, 1.0, and 1.5 LPM—under a fixed inlet temperature of 25.5 °C and a subcooling of 4.5 °C.

Key thermal metrics, including Critical Heat Flux (CHF), thermal resistance, and apparent heat transfer coefficient, were used to benchmark the performance of each cold plate. Additionally, parameters such as subcooling length, exit quality, and phase distribution were calculated to distinguish the contributions of single-phase and two-phase heat transfer.

The results reveal that minimum thermal resistance was typically observed near a surface heat flux of 100 W/cm², beyond which thermal resistance remained largely constant. Furthermore, increasing the flow rate had minimal impact on thermal resistance at 1.0 and 1.5 LPM. Geometric design played a critical role in determining performance: plates with narrower channel gaps and taller fins exhibited significantly improved thermal characteristics. Optimal results were achieved for samples with an aspect ratio around 12. However, aspect ratio alone was not a sufficient predictor of performance; a combination of geometric and flow parameters must be considered for effective design.

The second objective of the study was to optimize cold plate geometry specifically for two-phase operation using a hybrid framework combining optimization algorithms and machine learning (ML) techniques. To achieve this, key geometric parameters—number of channels, fin thickness, fin height, and channel gap—were reformulated into three non-dimensional groups. These were then correlated with experimental data to train predictive ML models and guide optimization algorithms toward identifying optimal configurations for enhanced two-phase thermal performance.

Overall, this study not only highlights the critical influence of geometry on two-phase heat transfer in skived fin cold plates but also introduces a data-driven framework for performance prediction and optimization. The insights gained may inform the next generation of high-performance cold plate designs for electronics cooling and other advanced thermal management applications.

Presenting Author: Mohammad Ahmadian Elmi Villanova University

Presenting Author Biography: Dr. Mohammad Ahmadian Elmi is a postdoctoral researcher in the Department of Mechanical Engineering at Villanova University. His research focuses on advanced thermal management systems, particularly two-phase cooling technologies, experimental heat transfer analysis, and data-driven design optimization. Currently, he integrates experimental, numerical, and machine learning techniques to enhance the thermal performance of skived fin cold plates for electronics cooling. His broader interests include multiphase flow modeling and sustainable thermal system design.