This work develops a hybrid design and analysis strategy to improve the dry sliding wear resistance of ultrasonically processed AA7075 composites reinforced with in-situ formed Al₃Ti and TiB₂ particulates. Experimental measurements are combined with Response Surface Methodology (RSM) to optimise reinforcement levels and sliding conditions for minimum wear rate and friction coefficient. A machine-learning prediction model is also established to estimate the tribological behaviour with high generalisation accuracy, verified by a prediction deviation of < 5% for both wear and friction responses. Furthermore, Shapley Additive Explanations (SHAP) are employed to identify the dominant process–response relationships, consistently revealing temperature as the most influential parameter governing wear behaviour, followed by applied load, while sliding speed exhibits a secondary effect. The integrated approach validates the robustness of the reinforcement route, offers reduced experimental cost, and provides deeper mechanistic insight into the wear behaviour of AA7075-based hybrid composites.Highlights<UnorderedList Mark="Bullet"> <ItemContent> AA7075 hybrid nanocomposites fabricated with fixed 3 wt% AIN and variable TiB₂ (2–8 wt%) via ultrasonic-assisted stir–squeeze casting. </ItemContent> <ItemContent> Uniform reinforcement dispersion and refined microstructure confirmed through SEM, TEM, and EDS. </ItemContent> <ItemContent> High-temperature dry sliding tests (50–250 °C) revealed optimal wear resistance and COF at 6 wt% TiB₂ + 3 wt% AIN. </ItemContent> <ItemContent> Response surface methodology and multi-model machine learning (RF, XGB, SVR, DNN) accurately predicted wear and frictional responses. </ItemContent> <ItemContent> Explainable AI (SHAP) identified temperature as the most influential factor governing wear rate, followed by applied load, while reinforcement content contributed to wear suppression. </ItemContent> <ItemContent> Experimental validation confirmed prediction accuracy with < 5% error, demonstrating the framework’s reliability for material design. </ItemContent> </UnorderedList>

High-temperature tribology of AA7075-TiB₂/AlN composites: experiments, response-surface optimisation, and interpretable machine learning

期刊论文
出版模式：

开放获取
发表日期：

2026年3月8日

查看全文

Namrata Mishra,
Annasamy Baradeswaran

摘要

This work develops a hybrid design and analysis strategy to improve the dry sliding wear resistance of ultrasonically processed AA7075 composites reinforced with in-situ formed Al₃Ti and TiB₂ particulates. Experimental measurements are combined with Response Surface Methodology (RSM) to optimise reinforcement levels and sliding conditions for minimum wear rate and friction coefficient. A machine-learning prediction model is also established to estimate the tribological behaviour with high generalisation accuracy, verified by a prediction deviation of < 5% for both wear and friction responses. Furthermore, Shapley Additive Explanations (SHAP) are employed to identify the dominant process–response relationships, consistently revealing temperature as the most influential parameter governing wear behaviour, followed by applied load, while sliding speed exhibits a secondary effect. The integrated approach validates the robustness of the reinforcement route, offers reduced experimental cost, and provides deeper mechanistic insight into the wear behaviour of AA7075-based hybrid composites.

Highlights

AA7075 hybrid nanocomposites fabricated with fixed 3 wt% AIN and variable TiB₂ (2–8 wt%) via ultrasonic-assisted stir–squeeze casting.

Uniform reinforcement dispersion and refined microstructure confirmed through SEM, TEM, and EDS.

High-temperature dry sliding tests (50–250 °C) revealed optimal wear resistance and COF at 6 wt% TiB₂ + 3 wt% AIN.

Response surface methodology and multi-model machine learning (RF, XGB, SVR, DNN) accurately predicted wear and frictional responses.

Explainable AI (SHAP) identified temperature as the most influential factor governing wear rate, followed by applied load, while reinforcement content contributed to wear suppression.

Experimental validation confirmed prediction accuracy with < 5% error, demonstrating the framework’s reliability for material design.