Scholarlink

<p>In this study, the application of machine learning (ML) approaches to predict solid-particle erosion (SPE) wear in aluminum metal matrix composites produced by powder metallurgy is explored. Four composite materials made of an aluminum substrate reinforced with different types of reinforcements, such as Al<sub>2</sub>O<sub>3</sub>, ZrO<sub>2</sub>, SiC, and WC, were investigated by varying the reinforcement content (0-10&#xa0;wt.%). The experiments were performed in compliance with ASTM G76 standards and included four impingement angles (30, 45, 60, 90°), velocities ranging from 30 to 90&#xa0;m/s, and types of reinforcements. A dataset comprising 121 experimental data points was collected. Various models, including SVM, SVR, ANN, decision tree, GBM, multiple linear regression, and random forest, were built in Python. The evaluation of the model’s performance relied on calculating <i>R</i><sup>2</sup>, RMSE, MAE, and the scattering index (SI). GBM appeared to be the most accurate model in the analysis (<i>R</i><sup>2</sup> = 0.95, SI = 0.06) along with SVR and SVM_PUK models. Predictive models exhibited low prediction errors from 2.70 to 9.05%. These results confirm that ML models can effectively capture nonlinear erosion behavior and provide a reliable predictive framework for PM-based AMMCs, thereby reducing experimental effort and improving material design optimization.</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-Based Prediction of Solid Particle Erosion Wear of Aluminum Metal Matrix Composites through Powder Metallurgy Route

  • 期刊论文

  • 发表日期：

    2026年7月16日

查看全文
  • Manvendra Yadav,
  • Virendra Pratap Singh,
  • L. A. Kumaraswamidhas,
  • Akant Kumar Singh,
  • Ashish Kumar

摘要

In this study, the application of machine learning (ML) approaches to predict solid-particle erosion (SPE) wear in aluminum metal matrix composites produced by powder metallurgy is explored. Four composite materials made of an aluminum substrate reinforced with different types of reinforcements, such as Al2O3, ZrO2, SiC, and WC, were investigated by varying the reinforcement content (0-10 wt.%). The experiments were performed in compliance with ASTM G76 standards and included four impingement angles (30, 45, 60, 90°), velocities ranging from 30 to 90 m/s, and types of reinforcements. A dataset comprising 121 experimental data points was collected. Various models, including SVM, SVR, ANN, decision tree, GBM, multiple linear regression, and random forest, were built in Python. The evaluation of the model’s performance relied on calculating R2, RMSE, MAE, and the scattering index (SI). GBM appeared to be the most accurate model in the analysis (R2 = 0.95, SI = 0.06) along with SVR and SVM_PUK models. Predictive models exhibited low prediction errors from 2.70 to 9.05%. These results confirm that ML models can effectively capture nonlinear erosion behavior and provide a reliable predictive framework for PM-based AMMCs, thereby reducing experimental effort and improving material design optimization.