Railway turnouts
Predicting faults of railway turnouts from the power consumption of their drives.
Railway industry
Predicting faults of railway turnouts
In this project the power consumption, in watts, of 300 turnouts is monitored. The turnouts perform periodic motions and are driven by single-phase or three-phase AC electric motors. These complex machines move the heavy turnouts either through friction drives or with the help of hydraulic transmission systems.
What and how to measure
Power consumption of each turnout
You can see the power curve of a three-phase electric motor that moves the switch blade. The motor is connected to a hydraulic system that balances the curve, which requires very sensitive measurement devices.
The large machine and the need for high sensitivity bring significant noise. Without PRECOG it is almost impossible to find the problems, because some fault types stay within the normal curve range.
Current and voltage, indoors
Current is measured with current sensor clamps and voltage with SIL-4 certified voltage measurement devices. Power consumption is current times voltage. The devices protect lives, so the measurements use SIL-4 certified devices.
Power consumption is measured instead of direct force, because the devices can be installed indoors in the relay room and measure continuously. Direct force measurement devices have to stand outdoors and suit only temporary measurement.
The analysis of such waveforms is described on the periodic motion analysis page.
Machine learning and data analytics
Machine learning
The algorithms analyse the data to find failure patterns in assets such as point machines. They help to select the relevant reference curves, which reduces noise and improves the accuracy of the prediction. As more data is collected, the confidence in predicting the behaviour of the asset and the time to failure grows.
Data analytics
Large data sets are processed to find trends and diagnose issues in real time. This helps to prioritise repairs, reduce unplanned failures and optimise maintenance schedules.
Testing
Other physical factors can influence a turnout as well, for example temperature, rain, snow and other environmental effects. They are reflected in the target data, which can be defined by testing the turnout or by analysing long-term data (see the target in the input file description). Testing the possible failures and analysing the data to understand the physical relations are crucial for setting up the right strategy with PRECOG.
Cost reduction and results
- Shorter logistic chains and smaller storage are possible with focused maintenance teams.
- Optimised asset use leads to long-term savings and better reliability.
- Fewer unplanned failures cut operating costs and optimise staff deployment.
- Intelligent maintenance strategies improve the allocation of resources and safety.
- Remote condition monitoring finds failures early and prevents over-maintenance.
- Predictive maintenance with data and machine learning reduces downtime and extends the life of the assets.
- It is a green solution, because less maintenance leaves a smaller carbon footprint.
Processing the telemetry data
- SensorsCurrent clamps and SIL-4 voltage devices
- Relay roomIndoor, continuous measurement
- IoT networkTransfers the measured values
- Data serverCalls the PRECOG server for the analysis
- PRECOG APIOn-premises or in the cloud
PRECOG is available through an API, with or without data storage, and can be deployed on-premises or in the cloud. The advantage of storing the data is that only the latest data has to be sent, which shortens the end-to-end prediction time. PRECOG follows best practices of safety and data protection to keep the connection secure.
Do you have a similar project in mind?