Bayesian-optimized machine learning framework for predicting Thermal Contact Resistance
Year: 2025
Type: article
Abstract: Accurate prediction of Thermal Contact Resistance (TCR) is essential for optimizing thermal management—whether enhancing or limiting heat flow—in energy, electronics, and aerospace systems. This study develops a robust data-driven framework that integrates Bayesian Optimization (BO) with machine learning (ML) to predict TCR across diverse interstitial materials and operating conditions, using a curated dataset of 300 experimental points. The analysis revealed a clear physical hierarchy: metal/solder interfaces exhibited the lowest resistance (≈8.3 ×10⁻⁶ K·m²/W), while the bare interface showed the highest (≈4.8 ×10⁻² K·m²/W). After dataset refinement using Cook’s Distance, the optimized ML model achieved a high coefficient of determination (R² = 0.898), outperforming the standard ML model (R² = 0.828). Feature importance analysis identified the interstitial material as the most dominant predictor, with an importance score more than twice that of any other feature, underscoring its pivotal role in governing heat transfer. Contact pressure ranked second, followed by the thermal conductivity of the first material, while other features exerted minor influence. Although the BO process did not further reduce the cross-validated RMSE, the final framework proved robust, stable, and physically consistent, providing a reliable foundation for data-driven prediction of TCR. Besides, the data-driven framework exhibits higher predictive accuracy than the classical model.
Source: Materials Today Communications
Authors: Malik I. Alamayreh, Ahmad Alamayreh, Amal A. Amer
Institutions: American University of Madaba, Al-Ahliyya Amman University
Language: English
Subfield: Materials Chemistry
Field: Materials Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,830
OpenAlex ID: W4417177628