Scholarlink

<p>Friction stir processing (FSP) was used to reinforce praseodymium oxide (Pr₆O₁₁) into AA1050 to improve near-surface tribological performance. An experimental investigation based on three parameters (tool rotational speed (TRS), traverse speed (TTS), and shoulder diameter (SD)) using central composite design was designed to systematically evaluate how variations in these parameters affect the microstructure and, subsequently, their correlation with wear performance. Among the wear results obtained for different parameter combinations, the lowest wear rate of 0.0729 × 10<sup>–7</sup>(gm/Nm) occurred for a TRS = 1000&#xa0;rpm, TTS = 50&#xa0;mm/min, and SD = 16&#xa0;mm, whereas TRS = 1200&#xa0;rpm, TTS = 70&#xa0;mm/min, and SD = 16&#xa0;mm resulted in the highest wear rate. EBSD analysis of the specimen with minimum wear showed a fully recrystallized stir zone with weak texture and a fine grain distribution, consistent with particle-stimulated nucleation and Zener pinning by Pr₆O₁₁. The obtained experimental results were subsequently analyzed using six machine learning techniques (KNN, SVR, XGBoost, Gradient Boosting, LightGBM, and CatBoost) to model the wear rate, explore the parameter-wear rate relationship, and predict the best parameter combination for improved wear performance. The best model (XGBoost) achieved <i>R</i><sup>2</sup> = 0.903 with MAPE = 6.4% (MSE = 0.100, MAE = 0.175), which was employed to find the best combination of TRS–TTS–SD. Collectively, the results define a process window where moderate heat input and uniform Pr₆O₁₁ dispersion yield sub-micron grains, stable wear films, and markedly reduced wear, providing a data-informed route for designing surface 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

Machine learning approach to predict the wear rate of Al/Pr₆O₁₁ surface composites fabricated by friction stir processing

  • 期刊论文

  • 发表日期：

    2025年11月21日

查看全文
  • Subi B.,
  • Abhiram Vijay,
  • Padmanaban Ramasamy,
  • VairaVignesh R.

摘要

Friction stir processing (FSP) was used to reinforce praseodymium oxide (Pr₆O₁₁) into AA1050 to improve near-surface tribological performance. An experimental investigation based on three parameters (tool rotational speed (TRS), traverse speed (TTS), and shoulder diameter (SD)) using central composite design was designed to systematically evaluate how variations in these parameters affect the microstructure and, subsequently, their correlation with wear performance. Among the wear results obtained for different parameter combinations, the lowest wear rate of 0.0729 × 10–7(gm/Nm) occurred for a TRS = 1000 rpm, TTS = 50 mm/min, and SD = 16 mm, whereas TRS = 1200 rpm, TTS = 70 mm/min, and SD = 16 mm resulted in the highest wear rate. EBSD analysis of the specimen with minimum wear showed a fully recrystallized stir zone with weak texture and a fine grain distribution, consistent with particle-stimulated nucleation and Zener pinning by Pr₆O₁₁. The obtained experimental results were subsequently analyzed using six machine learning techniques (KNN, SVR, XGBoost, Gradient Boosting, LightGBM, and CatBoost) to model the wear rate, explore the parameter-wear rate relationship, and predict the best parameter combination for improved wear performance. The best model (XGBoost) achieved R2 = 0.903 with MAPE = 6.4% (MSE = 0.100, MAE = 0.175), which was employed to find the best combination of TRS–TTS–SD. Collectively, the results define a process window where moderate heat input and uniform Pr₆O₁₁ dispersion yield sub-micron grains, stable wear films, and markedly reduced wear, providing a data-informed route for designing surface composites.