الفهرس | Only 14 pages are availabe for public view |
Abstract Predictive maintenance techniques are designed to determine the state of equipment in action to help know when we can intervene to perform maintenance on it. Predictive maintenance design applies artificial intelligence techniques such as machine learning to analyze data and monitor efficiency. In this thesis, we predict the maintenance of any equipment before it stops to reduce unplanned equipment maintenance by means of machine learning algorithms. We adopted in our work the following supervised machine learning algorithms: Random Forest, Support Vector Machine, KNN, Decision Tree, Logistic Regression, Naïve Bayes and XGBOOST. Simulations and results show that the Random Forest and XGBOOST have almost the same performance. However, results indicate that Random Forest algorithm has value of F1-sore with 0.9973. Followed by the XGBOOST algorithm has a F1-score relative value with 0.9884. The XGBOOST machine learning algorithm is preferred for bigger dataset compared to Random Forest and it works more effectively in the case of small dataset. In addition, we adopted the anomaly detection by using Isolation Forest algorithm to detect the anomaly data that used for predictive maintenance. Finally, the proposed predictive maintenance system is successful in identifying possible failure indicators and reducing some production stoppages as shown by our simulation results. |