Scholarlink

<p>The wear of bearing bush has a significant impact on the reliability of internal combustion engines. In this study, the wear behavior of PVD-coated bearing bushes under varying operating conditions is systematically investigated. The research not only delves into the underlying wear mechanisms, but also aims to predict wear trends through comprehensive analysis. Gaussian process regression (GPR) models are combined with machine learning techniques to construct the correlation between wear characteristics and performance indicators using limited experimental data, which can reduce trial and error costs and shorten the development cycle from exploration to engineering applications. The results show that with the increase in load and wear time, the wear loss of sliding bearing increases, and shows the opposite trend with the increase of rotational speed. The predicted values of the GPR model show good agreement with the measured values. Meanwhile, the GPR model constructed by the combined kernel function can better capture the relationship between the wear test parameters than the GPR model constructed by the single kernel function, and the prediction accuracy is higher. The BFGS-MR1-GPR model demonstrated a reduction in MAE by 0.0439, MSE by 0.0456, MAPE by 12%, and RMSE by 0.0486. For the BFGS-MR2-GPR model, the MAE was decreased by 0.0440, with other evaluation metrics showing consistent reduction magnitudes relative to the BFGS-RBF-GPR model as observed in the BFGS-MR1-GPR model.</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

Research on Sliding Bearing Wear Prediction Based on Gaussian Process Regression

  • 期刊论文

  • 发表日期：

    2025年12月9日

查看全文
  • Shanshan Liu,
  • Fengming Du,
  • Zanbin Gao,
  • Xiaoguang Han,
  • Yan Shen,
  • Jingsi Wang,
  • Weiwei Wang

摘要

The wear of bearing bush has a significant impact on the reliability of internal combustion engines. In this study, the wear behavior of PVD-coated bearing bushes under varying operating conditions is systematically investigated. The research not only delves into the underlying wear mechanisms, but also aims to predict wear trends through comprehensive analysis. Gaussian process regression (GPR) models are combined with machine learning techniques to construct the correlation between wear characteristics and performance indicators using limited experimental data, which can reduce trial and error costs and shorten the development cycle from exploration to engineering applications. The results show that with the increase in load and wear time, the wear loss of sliding bearing increases, and shows the opposite trend with the increase of rotational speed. The predicted values of the GPR model show good agreement with the measured values. Meanwhile, the GPR model constructed by the combined kernel function can better capture the relationship between the wear test parameters than the GPR model constructed by the single kernel function, and the prediction accuracy is higher. The BFGS-MR1-GPR model demonstrated a reduction in MAE by 0.0439, MSE by 0.0456, MAPE by 12%, and RMSE by 0.0486. For the BFGS-MR2-GPR model, the MAE was decreased by 0.0440, with other evaluation metrics showing consistent reduction magnitudes relative to the BFGS-RBF-GPR model as observed in the BFGS-MR1-GPR model.