Trains are not only environmentally friendly but also one of the safest ways to travel. Increasingly sophisticated safety technology has been installed on trains, tracks, and in stations to prevent accidents and human error.

Customer experience has been improved by easy-to-use ticket portals and passenger information systems providing latest travel information in apps, trains and stations. In addition, many trains also offer Internet access.

Train applications have always been a driving force behind technological advancements. First attempts to automate train operation (ATO) were made as early as 1920. Today, a number of railway lines around the world are already served by driverless trains.

Digital systems, modern communication, sensors, and monitoring technology are installed in most railway systems today. State-of-the-art technology is being used to control trains, supervise critical components and send information to the central control center in order to optimize traffic flow and increase safety. Diagnostic data helps the maintenance and service centers to detect anomalies at an early stage and schedule maintenance work without affecting train traffic. 

The Railway environment is highly complex. As shown in Figure 1, the system consists of the train itself, the rail network with switches and crossings, signaling systems and video surveillance that are in constant communication with the control center. Maintenance and repairs are based on intervals specified by the manufacturer and monitoring data from the train.
 

Together with other information such as the number of passengers, reasons for delays, cancellations or faults, weather conditions and energy consumption, a vast amount of data is available, which in the past was only analyzed partially and not as a whole.

Key Objectives of Railway AI: Efficiency, Safety, and Predictive Maintenance

Evaluating large amounts of data and discovering patterns is the strength of artificial intelligence which makes railway an ideal application. By using AI and own algorithms developed for their specific needs, railway operators expect to achieve:

• Optimization of train traffic, also by using simulation techniques with digital twins
• Energy savings by optimizing acceleration and braking
• Improved asset management matching actual needs
• Increased safety and prevention of human errors based on improved monitoring and sensor technologies
• Improved passenger experience by better flow management und information
• Predictive maintenance based on the analysis of monitoring data from the rail network and trains

As shown in figure 2 there are two different areas of application for AI. One is the analysis of data from trains and trackside systems in combination with information from, for example, the train ticketing system, weather services, service workshops, video cameras, and manufacturers. Actual scenarios and operating conditions can be analyzed and compared with historical data, which enables better traffic flow. Service companies can identify signs of wear and tear or malfunctions faster and intervene at an early stage.  

AI can be used on trains to optimize monitoring systems, increase safety and enable driverless operation. In this article, we will focus on the complexity of powering systems that usually have to be retrofitted in locomotives or railcars and enable local AI data processing or prepare data for evaluation in a central high-performance AI computer.

Analog circuits often require a variety of supply voltages and a power supply tailored to the device. Most digital circuits operate with one standard input voltage whether 3.3V, 5V, 12V or in industrial applications also 24V. Other voltages needed inside the system are generated by small DC/DC converters on the PCB. This standardization of supply voltages allows retrofits based on a few standard power supplies.

Nevertheless, train applications continue to face the challenge of widely varying train voltages, not only with large tolerances (60%–140% of the nominal voltage) but also with longer interruptions of up to 20 ms (Figure 3).
 

To reduce the number of different units for worldwide retrofits, power supplies should operate across the entire range, and mounting options should be as flexible as possible. The following examples show how P-Duke achieves this with new families designed for rail applications.

Real-World Applications: P-DUKE’s Railway-Certified Power Solutions

Obstacle detection
Important for safety and finally driverless operation is the detection of obstacles like people, animals, cars, or landslides. The environment is constantly monitored by cameras (video and thermal), radar and long-range LiDAR. Considering the long braking distance of trains, systems should cover distances of 1000m or more and detect critical situations in milliseconds rather than seconds 

This is only possible by integrating AI into the systems on the train with algorithms constantly refined by data sent from other trains to a central hub and analyzed by AI servers. This continuously improves object detection of each individual system.  

LiDAR systems (optical radar), send short laser pulses, measure transit time and angle of the reflected signal and calculate distance and position. A manufacturer of a long-range LiDAR family was looking for a plug-and play solution operating from all train voltages and backing up 20ms supply voltage drops. Input voltage of the system itself was 9V – 36V and power levels either 15 or 35W 

The URCD and URED families from P-Duke shown in figure 4 had been selected for this application

They cover the full input voltage range of 14 – 160V and provide output voltages of 3.3 to 24V with power levels of 10, 20 and 40W. With 20ms hold up time, integrated EMI filters and protection circuits no external components are needed. The Din Rail mounting option and pushpin connectors enable installation in a rack within a few minutes.

Track control
Do you remember the workers walking along the tracks and checking rails and track beds? A dangerous job that often led to accidents. Today systems mounted under the train can take 3 D measurements while the train is moving. Data is sent to the control center for further analysis by AI for safety, maintenance and repair planning purposes.     

For this application a customer wanted to use a radar system originally designed for harsh industrial environments and powered by the standard 24V industrial bus. Systems are mounted on hi-rail trucks or inspection trains. Therefore, they have to work from train voltages as well as 24 truck batteries (figure 5). Devices are mounted under a wide variety of vehicles and trains, quantities are relatively small, and therefore an easy-to-install solution was needed. Due to the short distances less power is needed and the URCD10U, a 10W converter, was the ideal solution as either the DIN Rail or the wall mount option can be used.
 

Tracking Systems
Tracking systems have been in place for decades to ensure that two trains never occupy the same section. Transmitters in the track bed automatically throttled the maximum speed or sent the current position to the train driver and the control center. Dead man circuits automatically stopped the train if the driver became incapacitated or unresponsive.
  
Modern systems can handle many more tasks like counting the axles of trains or collecting safety-relevant data such as vibrations in the bogie, condition of brakes, or temperatures of bearings and axles.

This again generates a wealth of data that is also suitable for AI analysis to improve algorithms in systems on the train. However, more data must be collected and transmitted which requires upgrades and retrofits of the system. The legacy system used a rugged embedded computer requiring 12V/30W and a power supply family was used with either 9V – 36V, 18V – 75V or 43 – 160V input ranges depending on the train it was installed initially.   

However, retrofitting different trains must be carried out during individual service intervals. Therefore, one single power supply for all systems means far less logistical effort. Again, with the URED40U family P-Duke was able to offer a solution not only working over the full 14 – 160V train voltage range but also providing 40W of output power in the same housing.  

Conclusion: Accelerating Autonomous Rail Operations with P-DUKE

These are just a few examples of modern systems in the railway ecosystem whose functionality can be improved through the use of AI. With modern sensor technology and appropriate evaluation software, trains, tracks, level crossings, signaling systems, switches, tunnels, marshalling yards, and platforms can be monitored fully automatically. With the help of AI millions of data points obtained from thousands of devices can be used to improve the algorithms in the respective systems and thus optimize the detection of obstacles, critical situations, or signs of wear and tear, for example.
 
By combining this with weather data, information from the ticket booking system, actual passenger flows, real travel times, and results from maintenance and repair operations, the goals of greater safety, better customer satisfaction, and expanded autonomous operation can be achieved.

The necessary upgrades to trains require uniform and easy-to-install solutions, especially for the power supplies. P-Duke offers an optimal platform that covers all train input voltages, provides many standard output voltages, is approved according to rail standards, requires no external components, and is easy to install.