Background
In the rail transport sector, maintenance directors must solve a difficult equation: how can they reconcile traffic growth, stronger security levels, and budgets (for the long term) under constraints? Faced with this situation, a major railway operator decides to take advantage of promises of predictive maintenance and Machine Learning in order to optimize commercial exploitation time, improve the organization of maintenance work, and modernize budget planning.
Challenges
Under these circumstances, the customer and Wavestone respond to three challenges:
- Determining, with the best level of performance, the equipment most likely to break down within a given time horizon.
- Enriching the preventive (scheduled and routine) and corrective (sustained and targeted) maintenance plans with a strategic predictive maintenance plan based on probability.
- Setting up the foundations of the Artificial Intelligence area in the Maintenance IS.
Responses & key success factors
Wavestone designs, prepares, and implements the 1st version of the predictive maintenance system:
- Current situation of maintenance operations and establishment of a set of un-siloed data made usable for predictive purposes.
- Development of the Machine Learning application for predictive maintenance combining supervised and non-supervised learning, detection of anomalies, and business analyst feedback loops. Supported by Wavestone’s R&D work, the application is approved through blind back-testing.
- Design and preparation of the implementation of the predictive maintenance plan (organization & IS).
