Overview

This project uses structured log files from signalling systems to remotely monitor asset performance and detect early signs of failure. It enables smarter, condition-based maintenance without adding new sensors or modifying existing systems.

Why Log Files?

Logs are already there — rich with information. By parsing them in real time, we can extract metrics like motor drive times and state transition times to identify patterns, precursors, and faults.

A live example

I have an AWS server currently replaying previous logs files (5-days worth) in a loop. The graphs below show the transition times for a particular set of railway points.

Loading chart…

A points machine is an electromechanical device that forms part of the signalling system's interlocking. It moves and locks railway points (switches) into position, ensuring that train wheels are safely guided onto the correct track. Control systems communicate with the interlocking to verify that all elements are in a safe state before a route is set. Establishing a route typically involves positioning and locking points in a specific configuration to create a safe passage through a junction.

The points machine is made up of several subsystems that work in series to perform its core function: setting the points to either the normal or reverse position. When instructed to change state, the machine follows a defined sequence of operations: unlocking, driving, detecting, and finally locking. This process ensures that the points are fully moved and securely held in the commanded position before a train is permitted to traverse the junction.

Each component within the points machine plays a critical role in the transition process. Electrical parts such as contacts and relays, together with driver motors, clutches, and the mechanical assemblies, all interact to move and secure the points. Proper setup and lubrication are also essential. Any degradation in these elements can affect the quality of the movement — often visible in changes measured in mere milliseconds — and may signal emerging faults before outright failure occurs.

By analysing transition time data, standard deviation can be applied to identify whether movements — both normal and reverse — are drifting outside the acceptable performance window. Even small increases in variability can indicate early signs of wear, misalignment, or loss of efficiency in the points machine, allowing a restorative task to be scheduled before a failure occurs.

Architecture

• Systems: Relay-based computer-controlled signalling assets
• Log Files: Raw data captured passively
• Data Acquisition: Custom software parses logs and sends events to a common database
• Processing: Rule-based filters and anomaly detection
• Output: Visualisation dashboards and automated work order triggers

Tools

Node.js, MySQL, JSON/XML streaming parsers, dashboard frontends, RESTful APIs, with potential integrations with enterprise maintenance systems.

Scope for more

In the log files I'm processing, there is a vast amount of information. It would be straightforward to extract and display the status of assets such as signals, axle counters, points position indicators, and transponder loops. In fact, I found using transponder loop data to track train movements particularly interesting — so I built the example below.

Train movements are recorded as loop detection events and mapped onto a structured grid of loops and landmarks. A dedicated web service streams these events through an API, continuously updating and rendering train positions in real time.

Final Thoughts

This has been a rewarding project to work on, and I can see significant value in implementing systems like this across the railway. The real challenge, however, lies not in the technology but in the culture of the industry. Rail organisations are often resistant to change, and with good reason: safety is paramount, and maintenance practices have been refined over decades. As a result, people and organisations remain comfortable with the status quo.

Leaders frequently express a desire to improve performance and are willing to fund development projects like this one. Yet, too often the investment stops short of change management — which, in my experience, represents half the effort required for successful implementation. Without addressing this human and organisational dimension, even the best technical solutions struggle to gain traction. I expand on this further in my note: Great Solutions Can Fail →