Cost-Sensitive Deep Learning for Scania Component X: Minimising Operational Cost via Asymmetric Threshold Optimisation
##plugins.themes.bootstrap3.article.main##
##plugins.themes.bootstrap3.article.sidebar##
Abstract
Maintenance decisions in industrial fleets must balance the cost of unnecessary interventions against the higher cost of missed failures. This paper presents a cost-sensitive deep learning approach for the Scania Component X benchmark. Rather than predicting the original five degradation classes directly, the problem is reformulated as a binary task that identifies whether a vehicle is healthy or at risk based on recent operational data. The predicted risk score is then converted into a maintenance decision through asymmetric threshold optimisation under the official Scania 5×5 cost matrix. Three temporal deep learning models—CNN, Transformer, and Temporal Convolutional Networks—are evaluated under the same cost-aware training and decision setting. Results on real data from thousands of heavy-duty trucks show how cost-sensitive learning affects the trade-off between failure avoidance, maintenance workload, and total operational cost.
How to Cite
##plugins.themes.bootstrap3.article.details##
Cost-sensitive learning, Deep learning, Maintenance decision-making, Predictive maintenance, Scania Component X, Threshold optimisation, Time-series classification, Heavy-duty trucks
Chehade, A., et al. (2022). Conditional Gaussian mixture model for warranty claims forecasting. Reliability Engineering & System Safety, 218, 108180. doi: 10.1016/j.ress.2021.108180
Deldari, S., Loke, S. W., & Zaslavsky, A. (2023). Predictive maintenance using deep learning: Review, challenges, and industrial applications. Sensors, 23(13), 6058. doi: 10.3390/s23136058
Ding, B., & Fang, H. (2017). Fault prediction for nonlinear stochastic systems with incipient faults. ISA Transactions, 68, 327–334.
Dintén, R., Zorrilla, M., Veloso, B., & Gama, J. (2025). Building of transformer-based RUL predictors supported by explainability techniques: Application on real industrial datasets. Information Fusion, 103892.
Johnson, J. M., & Khoshgoftaar, T. M. (2021). Thresholding strategies for deep learning with highly imbalanced big data. In Advances in Intelligent Systems and Computing (Vol. 1232, pp. 123–136). Springer. doi: 10.1007/978-981-15-6759-9_9
Kharazian, Z., Lindgren, T., Magnússon, S., Steinert, O., & Andersson Reyna, O. (2025). Scania Component X dataset: A real-world multivariate time-series dataset for predictive maintenance. arXiv preprint arXiv:2401.15199.
Khoshkangini, R., Mashhadi, P. S., Tegnered, D., & Rognvaldsson, T. (2022). Vehicles behavioral prediction using multi-task ensemble learning. SSRN Electronic Journal. doi: 10.2139/ssrn.4087639
Khoshkangini, R., Pashami, S., & Nowaczyk, S. (2019). Warranty claim rate prediction using logged vehicle data. In Progress in Artificial Intelligence (Vol. 11804, pp. 663–674). Springer. doi: 10.1007/978-3-030-30241-2_55
Lei, Y., et al. (2020). Applications of machine learning to machine fault diagnosis: A review. Mechanical Systems and Signal Processing, 138, 106587.
Man, J., & Zhou, Q. (2018). Prediction of hard failures using stochastic degradation signals. Computers & Industrial Engineering, 125, 480–489.
Mashhadi, P. S., et al. (2019). Stacked ensemble of recurrent neural networks for RUL prediction. Applied Sciences.
Parton, M., Fois, A., Vegliò, M., Metta, C., & Gregnanin, M. (2024). Predicting the failure of Component X in the Scania dataset with graph neural networks. In International Symposium on Intelligent Data Analysis (pp. 251–259).
Rengasamy, D., Jafari, M., & Rothwell, B. (2020). Deep learning with dynamically weighted loss function for prognostics and health management. Sensors, 20(3), 723. doi: 10.3390/s20030723
Shah, S., Wilder, B., Perrault, A., & Tambe, M. (2024). Leaving the nest: Going beyond local loss functions for predict-then-optimize. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 38, pp. 14902–14909).
Xu, D., & Zhang, J. (2026). Risk-aware decision-making for predictive maintenance using distributional reinforcement learning. International Journal of Systems Assurance Engineering and Management. doi: 10.1007/s40747-025-02127-w
Yang, Y., & Iqbal, M. Z. (2025). Cost-optimised machine learning model comparison for predictive maintenance. Electronics, 14(12), 2497.
Zhang, Q., Wang, J., & Li, T. (2023). A comprehensive review of cost-sensitive learning in predictive maintenance. Reliability Engineering & System Safety, 238, 109444. doi: 10.1016/j.ress.2023.109444
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
The Prognostic and Health Management Society advocates open-access to scientific data and uses a Creative Commons license for publishing and distributing any papers. A Creative Commons license does not relinquish the author’s copyright; rather it allows them to share some of their rights with any member of the public under certain conditions whilst enjoying full legal protection. By submitting an article to the International Conference of the Prognostics and Health Management Society, the authors agree to be bound by the associated terms and conditions including the following:
As the author, you retain the copyright to your Work. By submitting your Work, you are granting anybody the right to copy, distribute and transmit your Work and to adapt your Work with proper attribution under the terms of the Creative Commons Attribution 3.0 United States license. You assign rights to the Prognostics and Health Management Society to publish and disseminate your Work through electronic and print media if it is accepted for publication. A license note citing the Creative Commons Attribution 3.0 United States License as shown below needs to be placed in the footnote on the first page of the article.
First Author et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 United States License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.