Scholarlink

For the purpose of this study, a machine learning (ML) approach is used to address predictive maintenance tasks based on machine failures, facilitating effective maintenance interventions. To develop a machine failure prediction model based on a classification problem, a synthetic dataset containing 10.000 instances with 14 features is employed. Ten ML algorithms are tested: AdaBoost, Bagging Classifier, CatBoost, Decision Tree (DT), K-Nearest Neighbor (KNN), Naïve Bayes (NB), Random Forest (RF), Support Vector Classifier (SVC), Stacking Classifier, and Voting Classifier. The appropriate procedures are implemented in the Python programming language using its libraries, following a methodology conducted through proper activities, with the aim of obtaining a certain number of predicted classes. The 40 classification models are trained on all data, failure cases, the balanced dataset, and the binary balanced dataset. The results, in the form of metrics such as accuracy, precision, recall, and F1 score, are calculated and evaluated according to the criterion of the highest accuracy values. These values are cross-referenced with the confusion matrices, as well as with the F1-scores, if necessary. For all datasets, the CatBoost model proved to be the best, with accuracy rates ranging from 90.52% to 97.90%. To achieve better performance, it is necessary to further expand the dataset by including additional failure cases.

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

Leveraging Machine Learning As a Prerequisite for Effective Maintenance

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Nevena Tufegdžić,
  • Milica Tufegdžić,
  • Raul Turmanidze,
  • Giorgi Popkhadze

摘要

For the purpose of this study, a machine learning (ML) approach is used to address predictive maintenance tasks based on machine failures, facilitating effective maintenance interventions. To develop a machine failure prediction model based on a classification problem, a synthetic dataset containing 10.000 instances with 14 features is employed. Ten ML algorithms are tested: AdaBoost, Bagging Classifier, CatBoost, Decision Tree (DT), K-Nearest Neighbor (KNN), Naïve Bayes (NB), Random Forest (RF), Support Vector Classifier (SVC), Stacking Classifier, and Voting Classifier. The appropriate procedures are implemented in the Python programming language using its libraries, following a methodology conducted through proper activities, with the aim of obtaining a certain number of predicted classes. The 40 classification models are trained on all data, failure cases, the balanced dataset, and the binary balanced dataset. The results, in the form of metrics such as accuracy, precision, recall, and F1 score, are calculated and evaluated according to the criterion of the highest accuracy values. These values are cross-referenced with the confusion matrices, as well as with the F1-scores, if necessary. For all datasets, the CatBoost model proved to be the best, with accuracy rates ranging from 90.52% to 97.90%. To achieve better performance, it is necessary to further expand the dataset by including additional failure cases.