Scholarlink

<p>The detection and prediction of the condition of machinery and transportation systems in the mining industry are of critical importance due to their direct impact on productivity, maintenance cost reduction, and operational safety. Given the high costs and suboptimal performance of this sector, the present study aimed to develop an efficient framework for monitoring and predicting fleet conditions. Raw data were collected and preprocessed through cleaning, normalization, and imputation of missing values. A dataset consisting of 2,950 oil samples and 27 operational and chemical attributes collected between 2020 and 2023 was used in this study. To address the challenge of data imbalance, resampling techniques and class weighting were applied. A set of machine learning algorithms—including Gradient Boosting, Random Forest, Artificial Neural Networks, and Support Vector Machines—were implemented, and their performance was evaluated using multidimensional metrics such as overall accuracy, balanced accuracy, and F1 score and ROC-AUC under a five-fold cross-validation framework. The results indicate that, after hyperparameter optimization, the Support Vector Machine achieved the best balance between precision and recall, with an overall accuracy of 98.19% and an F1 score of 87.43%. Statistical validation results further confirmed the robustness and reliability of the proposed framework for predictive maintenance applications. Overall, the findings suggest that combining systematic data preprocessing, robust algorithms, and multidimensional evaluation metrics provides a solid foundation for intelligent decision-support systems in the mining industry, can support the development of intelligent predictive maintenance systems in mining operations enhancing prediction accuracy, reducing costs, and preventing unexpected mechanical downtimes.</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

Comparative analysis of support vector machines, artificial neural network, random forest and gradient boosting for predictive maintenance in mining machinery and equipment: a case study of Chadormalu Iron Ore Mine

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月25日

查看全文
  • Omid Afzali Alvars,
  • Sajjad Afraei,
  • Majid Ataee-pour

摘要

The detection and prediction of the condition of machinery and transportation systems in the mining industry are of critical importance due to their direct impact on productivity, maintenance cost reduction, and operational safety. Given the high costs and suboptimal performance of this sector, the present study aimed to develop an efficient framework for monitoring and predicting fleet conditions. Raw data were collected and preprocessed through cleaning, normalization, and imputation of missing values. A dataset consisting of 2,950 oil samples and 27 operational and chemical attributes collected between 2020 and 2023 was used in this study. To address the challenge of data imbalance, resampling techniques and class weighting were applied. A set of machine learning algorithms—including Gradient Boosting, Random Forest, Artificial Neural Networks, and Support Vector Machines—were implemented, and their performance was evaluated using multidimensional metrics such as overall accuracy, balanced accuracy, and F1 score and ROC-AUC under a five-fold cross-validation framework. The results indicate that, after hyperparameter optimization, the Support Vector Machine achieved the best balance between precision and recall, with an overall accuracy of 98.19% and an F1 score of 87.43%. Statistical validation results further confirmed the robustness and reliability of the proposed framework for predictive maintenance applications. Overall, the findings suggest that combining systematic data preprocessing, robust algorithms, and multidimensional evaluation metrics provides a solid foundation for intelligent decision-support systems in the mining industry, can support the development of intelligent predictive maintenance systems in mining operations enhancing prediction accuracy, reducing costs, and preventing unexpected mechanical downtimes.