Scholarlink

<p>Tribological phenomena, encompassing friction, wear, and lubrication, significantly impact the performance and efficiency of mechanical systems across various industries. This research introduces a novel approach by applying multiple machine learning algorithms to predict and optimize tribological behaviors, addressing the current gap in quantitative modeling of coating material effects. The work uniquely investigates the effect of applied load, sliding speed, and coating composition—specifically, the incorporation of graphitic carbon nitride (g-C<sub>3</sub>N<sub>4</sub>) in molybdenum disulfide (MoS₂) coatings—on tribological responses. Several machine learning algorithms, including linear regression, decision trees, random forests, support vector regression, k-nearest neighbors, and neural networks, are employed to capture the complex relationships between the input parameters and the responses. The neural network model achieved the best performance with a low mean squared error (MSE) of 0.0023 and high R-squared of 0.9977 for predicting wear depth, along with an MSE of 567.89 and R-squared of 0.9654 for coefficient of friction predictions. Random forests also exhibited strong performance with testing MSE of 0.0298 and R-squared of 0.9671 for wear depth. Feature importance analysis identifies coating material as the most influential factor for wear depth, while sliding speed is the dominant factor (F-ratio of 8.589) affecting the coefficient of friction. Statistical significance testing confirms a substantial difference between the means (t-statistic of -14.544, p-value of 5.41e-31). Multiple linear regression analysis reveals an optimized input parameter combination of 10 kN applied load, 0.5&#xa0;m/s sliding speed, and 9 Wt% of gCN, minimizing both tribological responses with low MSEs of 0.0214 (wear depth) and 12.8745 (coefficient of friction) on the training set. The study provides valuable insights into tribological optimization, highlighting the potential of machine learning techniques for enhancing system performance, efficiency, and durability. The findings contribute to a deeper understanding of tribological processes and pave the way for future research in developing advanced coatings, lubricants, and condition monitoring systems for tribological applications. </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

Tribological analysis and data-driven modeling of graphitic carbon nitride/molybdenum disulfide nanocomposite coatings: experimental insights and predictive optimization

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Mukul Saxena,
  • Anuj Kumar Sharma,
  • Monika Singh

摘要

Tribological phenomena, encompassing friction, wear, and lubrication, significantly impact the performance and efficiency of mechanical systems across various industries. This research introduces a novel approach by applying multiple machine learning algorithms to predict and optimize tribological behaviors, addressing the current gap in quantitative modeling of coating material effects. The work uniquely investigates the effect of applied load, sliding speed, and coating composition—specifically, the incorporation of graphitic carbon nitride (g-C3N4) in molybdenum disulfide (MoS₂) coatings—on tribological responses. Several machine learning algorithms, including linear regression, decision trees, random forests, support vector regression, k-nearest neighbors, and neural networks, are employed to capture the complex relationships between the input parameters and the responses. The neural network model achieved the best performance with a low mean squared error (MSE) of 0.0023 and high R-squared of 0.9977 for predicting wear depth, along with an MSE of 567.89 and R-squared of 0.9654 for coefficient of friction predictions. Random forests also exhibited strong performance with testing MSE of 0.0298 and R-squared of 0.9671 for wear depth. Feature importance analysis identifies coating material as the most influential factor for wear depth, while sliding speed is the dominant factor (F-ratio of 8.589) affecting the coefficient of friction. Statistical significance testing confirms a substantial difference between the means (t-statistic of -14.544, p-value of 5.41e-31). Multiple linear regression analysis reveals an optimized input parameter combination of 10 kN applied load, 0.5 m/s sliding speed, and 9 Wt% of gCN, minimizing both tribological responses with low MSEs of 0.0214 (wear depth) and 12.8745 (coefficient of friction) on the training set. The study provides valuable insights into tribological optimization, highlighting the potential of machine learning techniques for enhancing system performance, efficiency, and durability. The findings contribute to a deeper understanding of tribological processes and pave the way for future research in developing advanced coatings, lubricants, and condition monitoring systems for tribological applications.