AI-ML for optimized rail maintenance

With railway infrastructure assets across the globe now leveraged to their maximum, organizations are forced to work round the clock to maintain them.

Most discrete maintenance activities require disruption of services. This has a direct economic impact, and results in a significant number of idle hours for the workforce and inefficient use of maintenance equipment. 

The expensive and complex process of scheduling maintenance jobs only makes matters worse. Autonomous scheduling of maintenance jobs, therefore, is essential to minimize disruption and ease the scheduling process. It also ensures seamless and smooth operations for rail infrastructure managers, enabling safer journeys for passenger and freight transport with minimal interruptions. 

For efficient maintenance, railway organizations need to undertake an integrated analysis of the service timetable and access modeling. Efficient asset management requires sufficient ‘access’ (train-free) hours on the track to carry out maintenance and renewal activities. Also, it is important to reduce infrastructure maintenance and renewal costs without compromising safety and customer satisfaction. 

The challenges in asset management can be overcome with an intelligent, AI-ML-based network access planning solution framework that ensures optimized network maintenance and informed decision-making by combining the service timetable with cost-efficiency modelling.  

Timely, efficient rail network maintenance must for resilient operations 

Railways will continue to be a sustainable mode of transport for people and cargo well into the future.

The European Commission marked the year 2021 as the European Year of Rail, an initiative proposed as part of its efforts to achieve climate neutrality by 2050 under the European Green Deal. A robust railway infrastructure maintenance system is critical to achieve this goal. The current methods and systems for railway maintenance have limitations as they are resource-heavy, and resources remain unutilized in idle time during maintenance. Lack of appropriate dataprocessing techniques to perform relative analytics remains a key challenge to railway maintenance activities. Over and above, without proper maintenance, railways, which function mainly as public sector undertakings, face the risk of running afoul of regulatory compliance and user acceptance.

One way to overcome these limitations is to do maintenance activities in multiple phases over several days whenever the rail infrastructure is free (ie, not occupied by a train). But this approach may not be cost-efficient, especially for long maintenance activities as they incur set-up and wind-up costs in addition to labor and equipment costs. Another approach is to make deviations in the timetable (delay some trains, cancel some) to make room for longer free slots to carry out maintenance work. This approach, however, results in service operations cost associated with delay or cancellation of trains. Rail infrastructure managers need to carefully analyze and balance costs associated with each of the approaches and identify the optimum level of train service for the lowest overall cost (considering both value of train services and asset management cost).

Harnessing AI, a deep reinforcement learning-powered solution, and integrating third-party data streams will improve the cost efficiency of maintenance activities. An access optimization framework linking train operations (the timetable) to access network blocks will enable transparent optimization of train services, enabling higher efficiency and lower maintenance cost.

A maintenance framework that strikes right balance between operations, costs 

Figure 1 illustrates an AI-ML-based framework that combines timetable and cost efficiency modelling to provide an optimum maintenance schedule. 

Harnessing AI, a deep reinforcement learning-powered solution, and integrating third-party data streams will improve the cost efficiency of maintenance activities. 

Figure 1: Intelligent solution for rail access management

The key components of the framework are: 

• Operations: This identifies additional access hours from the existing timetable while minimizing the impact to train operations 

• Cost: This will model the optimal allocation of maintenance and renewal activities to longer duration access windows, including sharing of access windows to reduce overhead costs.

The framework focuses on having a trade-off between the incremental value of running trains and the incremental impact of those trains on asset management costs. The core differentiator of the framework is that it enables effective railway asset management at efficient cost. The optimization engine in the framework makes a comparison of costs involved in train delays or cancellations versus labor, equipment, and other charges for multiple days. It, thereby, takes an informed decision by prioritizing jobs that yield better cost savings. The framework helps improve the life cycle cost of track assets by 15-40% by providing longer access window for maintenance activities

Framework analyzes multiple factors to provide options 

The framework does the complex analysis of the scope and duration of maintenance work required, value of operating an individual train, the location of the maintenance work, and the time of day when it is to be done.

Such an analysis requires a combination of timetable and infrastructure cost modelling, and also should take into consideration the fact that railway networks cover dozens of access blocks with hundreds of different activities (tasks) to be carried out over a 24*7 timetable. 

Often, such an analysis is done manually. For instance, if there are no trains running from 12 AM to 4 AM, a 20-hour maintenance job will have a window of only four hours per day (12 AM to 4 AM), requiring five days to accomplish the task. The operations team has to pay labor charges and equipment rental for the full five days for just four hours of work per day. However, delaying or cancelling trains running just before 12 AM or just after 4 AM will give a wider maintenance green zone, making it more cost-effective. The framework will, therefore, consider incremental cost of running late night or early morning services before giving its recommendations. 

An access optimization framework linking train operations (the timetable) to access network blocks will enable transparent optimization of train services, enabling higher efficiency and lower maintenance cost. Such robust solutions can illustrate ways to reduce infrastructure costs with minimum disruption to train services, help achieve incremental cost benefits by running late night or early morning services, and facilitate normalization of outputs between routes. Normalization (comparing the various metrics) of routes’ outputs is required to comply with rail maintenance procedures and carry out an ‘apple-to-apple’ comparison between routes. 

Infrastructure managers can leverage the framework as a ready reckoner or a guide to schedule and execute maintenance work efficiently using the additional access hours. The framework will reduce infrastructure maintenance and renewal cost by enabling integrated work (multiple activities) in fewer, longer possessions with minimal impact on operations. On top of it, a more effective use of access to infrastructure means less maintenance backlog and, hence, better asset reliability. Further, it reduces red zone (when track is occupied by a train and is not available for maintenance activities) working by enabling more activities to be undertaken during the green zone (safe zone, when track is not occupied by a train). This framework helps enhance safety and information accuracy while reducing the idle time for train maintenance operations. The framework can be leveraged for other asset-heavy industries where there is a need of a trade-off between production and maintenance efficiency.

Automated maintenance schedules will soon become industry norm 

With rail traffic increasing, there is bound to be greater demand for rail infrastructure maintenance, necessitating automation of the complex maintenance process. Railroads and their ecosystems would leverage robust frameworks to enable seamless operations and drive resilience for partners.

The framework discussed above would be put to test in scenarios where there is a higher impact on train operations. In any scenario, the savings in maintenance cost should be more than the value lost due to the impact on operations during those additional access hours. The impact may include cancelling, delaying, short terminating, or re-routing train services. Possible disagreement on data-sharing between the railroad ecosystem partners or regulatory approvals on data sharing (between government authorities and railroad maintenance organizations) could be some of the possible roadblocks in seamless implementation of the solution.

However, as the market for automated rail maintenance grows, we may see new revenue-sharing business models emerge that will leverage the power of data and digital solutions. IT companies will, therefore, play a key role in defining the growth and digital transformation strategy for railways.