Points Predict is Thales’s new, smart maintenance service for rail network operators. As it monitors trackside equipment, it also learns equipment performance norms, offering operators a range of new benefits:

• Accounts for seasonal drifts and environmental changes, automatically adjusting alert detection points for each individual asset;
•  Predicts likely faults by building a library of deviation profiles, enabling the right maintenance team with the right equipment to be sent on site;
• Removes complexity, reducing the levels of training and experience required of monitoring staff;
• The Cloud-based system makes more real-time data accessible to more people, supporting analysis, fault trend identification and decision making.

Switching from dates to data

Scheduled maintenance programmes are safe but expensive, imposing work and cost, even when it’s not necessary. In recent years, operators have been moving towards a condition monitoring approach, using systems such as Thales’s smart maintenance platform. Linked to the power supplies of electro-mechanical equipment such as point machines, these systems monitor ampage and derive performance parameters, such as duration, average and peak power. An alarm is triggered if levels hit a threshold point, leaving staff to deduce a likely cause.

However, an asset’s power profile can drift, depending on factors such as seasonal temperature change and level of usage; adjusting trigger thresholds for an entire rail system of anything over 100 networked devices is a fulltime job. In addition, a human being must set threshold parameters carefully: too tight and an increase in false alerts leads to wasted call-outs; too loose and a fault may be missed.

Predicting and classifying problems

The launch of Points Predict represents a step change in digital capability. Current systems work at the “discover” level, capturing asset knowledge while enabling analysis and ad-hoc investigation into large volumes of data. The new algorithm introduces a “predict” level of functionality, applying machine learning algorithms to multiple streams of data to automate diagnosis and prognosis of asset behaviour.

Unlike any other package on the market, Points Predict removes the need to adjust thresholds or set balanced warning parameters manually. Using an anomaly detection and fault classification algorithm and taking account of ambient conditions and workload, Points Predict learns what a typical baseline behaviour looks like for every asset within a network.

The algorithm requires adjustment of one single sensitivity parameter, freeing the operator from having to balance and manage thresholds across multiple points types in order to ensure an optimal outcome.

The system is also busy building a library of fault profiles so that when an anomaly is detected it can classify the fault to a high degree of accuracy, allowing maintenance crews to respond quickly and appropriately. Tests with operational track data showed:

• A 50% improvement in detection rates
• 75% fewer false alerts
• 90% fault classification accuracy after training with a single example

A bolt-on digital capability

Points Predict also adds a layer of predict intelligence capability to existing digital monitoring systems.

Operators can leverage its digital technology to gain more value from their incumbent condition monitoring systems rather than having to replace them.

Increasing efficiency through intelligent maintenance

The benefits of a more intelligent, predictive fault monitoring and classification system are felt by all stakeholders; passengers experience fewer delays and cancellations due to equipment failures; operators minimize fines from rail regulators for delays; and less trackside work reduces risk of injury while operations staff are released from repetitive monitoring work to focus on new tasks.

Points Predict represents a fast track to a more efficient and higher quality rail transport service.

You can read more about Thales’s smart maintenance platform here.