Keywords: random forest, artificial intelligence, rolling stock, machine learning, predictive model, reliability.

This paper explores the application of machine learning techniques to enhance the reliability of sleeper train critical systems contributing to service operation disruptions. The primary objective of the paper is to develop predictive models capable of identifying and predicting faults to facilitate proactive maintenance. An attempt is made to develop and use two machine learning algorithms, Random Forest (RF) and Support Vector Machine (SVM), to analyse condition monitoring data and predict system failures. An exploratory data analysis was conducted, and some limitations and imbalance in the dataset were observed. The Synthetic Minority Oversampling Technique (SMOTE) was applied to effectively balance the class distribution and improve the model performance. The proposed models were evaluated using precision, recall, F1-score, and the overall accuracy metrics. The results demonstrated that the Random Forest (RF) model significantly outperformed the Support Vector Machine (SVM) model, thereby achieving a well-balanced trade-off between precision and recall. After addressing data imbalance, the RF model achieved an overall accuracy of 75%, compared to 65% accuracy with imbalanced data. The precision and recall scores for the RF model indicated reliable performance in both fault detection and prediction. In contrast, the SVM model exhibited lower performance metrics, especially in identifying faulty incidents, it achieved perfect recall but low precision for one class, also indicating many false positives. The SVM model on the other hand achieved an overall accuracy of 48% before addressing data imbalance, which improved to 70% with balanced data. This paper contribution emphasis on the importance of data quality, feature selection, and model ability in handling data imbalance to support decision making.

download the full-text of this paper (PDF, 14 pages, 1178 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description