In 2022 alone, there were 1,164 train derailments in the U.S. according to the Federal Railroad Administration, resulting in over $100 million in damages. This equates to an average of three derailments per day. While Europe has some of the safest railways in the world, rail accidents have been in the news globally. For example, three accidents made international headlines this year: a fifty-car freight train pileup in the U.S., a passenger train crash in the Netherlands that caused one death and multiple injuries, and another major crash in Odisha, India. The U.S. crash caused catastrophic environmental and community destruction and cost 387 million dollars in damages. Train derailments present a threat to global rail safety and have resulted in substantial financial losses.
What causes train derailments?
Ageing infrastructure in the rail system is often the culprit. It has created challenges in detecting mechanical issues and has led to accidents. For example, overheated wheel bearings were found to be the cause of the U.S. crash this year and have contributed to numerous other derailments. While the U.S. rail network incorporates approximately 6,000 hot box detectors to monitor the temperatures of wheel bearings as trains pass by, these detectors have not been able to effectively predict and prevent imminent failures. Current technology does not correlate data from multiple detectors and incorporate other relevant information, such as recent service issues with the bearings. Having all this information would enable predictive analytics to accurately identify failing wheel bearings and alert train engineers before an accident occurs.
Likewise, current techniques for monitoring rail switches and intersections often do not integrate all the information necessary to detect and prevent accidents in real time. Instead, they often rely on more rudimentary signalling systems, which can fail and lead to catastrophic accidents. For example, in June 2023 in India, three trains collided due to a signalling error that routed an inbound passenger train onto a siding and into the path of a stationary freight train. This accident killed 296 people and injured 1,200.
Although most train derailments are not as catastrophic as the crash in India, they pose a safety threat and cost the industry millions of dollars annually. The rail system’s ageing infrastructure needs help. Luckily, recent technology advancements can reduce their alarming frequency, increase safety, and lower operational costs.
Digital Twins with Machine Learning offer a compelling solution
New software technology called Digital Twins can improve rail safety by adding intelligent real-time monitoring to the rail system and creating timely alerts when problems arise. Digital twins can simultaneously monitor thousands of assets, such as rail cars and switches, to continuously look for emerging issues. For example, Digital Twins can track all rolling stock to detect impending wheel bearing failures. Because each Digital Twin can independently correlate data from multiple hot box detectors and combine this with other relevant information, such as the wheel bearing’s service issues and the ambient temperature, it can precisely identify the conditions which indicate an impending wheel bearing failure.
To accurately decipher what is or isn’t an issue and signal an alert only when a real problem is occurring, Digital Twins can incorporate machine learning algorithms that identify patterns in the data indicating an emerging problem, such as an overheated wheel bearing. To analyse wheel-bearing temperatures, machine learning algorithms use training data from thousands of detectors to learn which patterns of temperature measurements need alerting. They enable Digital Twins to provide fast, powerful analytics.
Digital twins can also monitor switches, crossings, and train positions to look for conflicts that could lead to collisions. For example, in the Dutch accident, a train unexpectedly collided with a crane. A Digital Twin of the crane that was blocking the tracks could exchange information with other Digital Twins of approaching trains to detect a conflict and generate an alert. Having Digital Twins monitor key assets provides another layer of safety that can help avoid accidents.
Using thousands of Digital Twins to provide real-time monitoring requires a fast, scalable computing platform. In-memory computing technology can host Digital Twins and enable fast access to data. This technology automatically distributes the processing workload across multiple servers to reduce analysis times down to milliseconds.
Model rail networks with Digital Twin simulations
The flexibility of Digital Twins also enables to implement simulations that test their monitoring capabilities and accuracy before deployment in a live system. Simulations allow system designers to model complex rail networks and ensure that monitoring techniques will successfully detect issues and generate alerts when needed.
For example, to demonstrate how Digital Twins can detect wheel bearing failures, ScaleOut Software created a simulation of the U.S. freight rail system using its ScaleOut Digital Twins technology. This simulation modelled thousands of trains and injected random cases of deteriorating wheel bearings to validate that Digital Twins can track wheel-bearing temperatures from multiple hot box detectors and alert engineers before failures occur.
Keeping railways safe
Railways are the backbone of the modern supply chain and a cornerstone of public transportation. However, today’s rail monitoring infrastructure often falls short of detecting emerging issues, leading to costly accidents. New technologies, like Digital Twins, machine learning, and in-memory computing, can build on existing rail infrastructure to provide continuous monitoring, identify problems, and boost situational awareness for personnel. They have the potential to significantly reduce the current frequency of accidents and pave the way for a safer, more efficient, and reliable transportation system.
Source: www.railtech.com