Scholarlink

<p>The present study investigates the prediction of wear behavior of AA7075/SiC composites using supervised machine learning (ML) algorithms, emphasizing the role of hyperparameter tuning in enhancing model accuracy. Experimental wear data were generated through pin-on-disc tests under varying SiC reinforcement contents (0–9 wt.%), applied loads, sliding velocities, and sliding distances. Two ensemble-based ML algorithms—Random Forest (RF) and Extreme Gradient Boosting (XGBoost)—have been developed to model the nonlinear relationship between input process parameters and wear rate. Hyperparameter optimization using Grid Search significantly improved predictive performance, yielding <i>R</i><sup>2</sup> ≈ 0.89 and <i>RMSE</i> ≈ 0.0009 for both models. Statistical and residual analyses confirmed that the tuned models provided unbiased and normally distributed errors, ensuring strong generalization capability. The results indicated that SiC reinforcement was the most influential parameter in reducing wear rate, followed by load and sliding distance. The study demonstrates that integrating experimental tribological data with optimized ML algorithms offers a&#xa0;robust and efficient framework for data-driven prediction and process optimization in aluminum matrix composites.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Influence of hyperparameter optimization on the predictive accuracy of supervised machine learning models for the wear behavior of AA7075/SiC microwave sintered composites

  • 期刊论文

  • 发表日期：

    2026年6月19日

查看全文
  • Venkateswara Reddy Kunduru

摘要

The present study investigates the prediction of wear behavior of AA7075/SiC composites using supervised machine learning (ML) algorithms, emphasizing the role of hyperparameter tuning in enhancing model accuracy. Experimental wear data were generated through pin-on-disc tests under varying SiC reinforcement contents (0–9 wt.%), applied loads, sliding velocities, and sliding distances. Two ensemble-based ML algorithms—Random Forest (RF) and Extreme Gradient Boosting (XGBoost)—have been developed to model the nonlinear relationship between input process parameters and wear rate. Hyperparameter optimization using Grid Search significantly improved predictive performance, yielding R2 ≈ 0.89 and RMSE ≈ 0.0009 for both models. Statistical and residual analyses confirmed that the tuned models provided unbiased and normally distributed errors, ensuring strong generalization capability. The results indicated that SiC reinforcement was the most influential parameter in reducing wear rate, followed by load and sliding distance. The study demonstrates that integrating experimental tribological data with optimized ML algorithms offers a robust and efficient framework for data-driven prediction and process optimization in aluminum matrix composites.