Data-driven and experimental optimization of thermal performance in ZrO₂ Nanofluid oscillating heat pipes

The growing thermal load of small electronic equipment and electric vehicle (EV) batteries has put a strain on the need to develop sophisticated thermal control systems. In this study, the thermal performance of an oscillating heat pipe (OHP) charged with ZrO₂–acetone nanofluid was experimentally investigated and optimized using a combination of Response Surface Methodology (RSM) and Machine Learning (ML) techniques. The input parameters, namely ZrO₂ nanoparticle concentration (1–3 wt.%), fill ratio (40–80%), and heat input (15–35 W), were systematically varied according to a Box–Behnken experimental design to evaluate their effects on thermal resistance (TR) and heat transfer coefficient (HTC). Analysis of variance (ANOVA) and response surface analysis were employed to identify the significance of process parameters and their interactions, while K-nearest neighbor (KNN), support vector machine (SVM), and extreme gradient boosting (XGBoost) models were developed to predict thermal performance. The obtained results indicate that heat input and fill ratio are the dominant factors influencing TR and HTC, whereas the effect of nanoparticle concentration is comparatively less significant within the investigated range. The optimum operating condition was obtained at approximately 2 wt.% ZrO₂ concentration, 80% fill ratio, and 35 W heat input, resulting in a minimum thermal resistance of 1.03 °C/W and a maximum heat transfer coefficient of approximately 222 W/m 2 ·°C. The novelty of this work lies in integrating experimental investigation with RSM-based optimization and comparative ML prediction models for ZrO₂ nanofluid-based OHPs, providing an effective data-driven framework for thermal performance prediction and optimization in advanced thermal management applications.