Singapore – Public transportation SMRT is artificial intelligence Improving service standards may take years, but if implemented on a large scale, it could reduce train delays and reduce operating costs.
Through a new AI-powered platform called Jarvis, SMRT leverages data from across its systems to predict failures before they occur.
A simple chatbot-style interface gives maintenance teams instant access to insights and predictions of when equipment is likely to fail, allowing them to act before failures occur, reducing train delays, improving reliability, and making passenger journeys safer.
Developed by SMRT’s technology arm Strides Technologies in collaboration with software and cloud computing company Oracle, Jarvis is currently being piloted with about 50 users.
Announced during the period Oracle’s AI World Tour to be held in Singapore on April 14th The multi-year effort aims to apply AI to rail engineering while ensuring the rigorous safety standards required for rail operations are met.
This move will see SMRT expand and improve its use of AI to improve its operations. The company already uses AI systems to proactively identify potential problems and congestion on trains.
It also reflects how AI is pervasive beyond sectors such as banking. and bring healthcare into areas such as public transportation. but, Its effectiveness still depends on sufficient data and supporting infrastructure. At least for now, they are most often used to support human decision-making rather than replace it.
Jarvis’ primary goal is to allow engineers to move away from traditional time-based maintenance (where parts and equipment are checked at regular intervals regardless of their condition) and monitor equipment performance in real time.
This allows maintenance to be performed only when there are signs of wear or potential failure.
“If you can assess conditions and predict failures, you don’t need to do maintenance every three to six months,” said Albert Soh, Head of Business Operations and Analytics at Strides. “I’ll do it when I need to.”
First, the new system focuses on mechanical failures that follow predictable wear patterns. One example is a rolling machine. This is a critical track component that exchanges trains between tracks. Although failures are rare, So explained that when they occur, they can cause significant disruption.
Jarvis is trained to use historical and sensor data to detect early signs of deterioration days in advance, allowing maintenance staff to resolve issues within a scheduled time frame rather than reacting at the time of failure.
Other applications include platform doors across the north-south and east-west lines, with more than 2,000 doors monitored. By analyzing the time it takes for a door to open and close, the system can detect subtle performance degradation and prioritize maintenance before failure occurs.
This will allow engineers to more accurately deploy resources while reducing passenger inconvenience, So said. “For commuters, the result will be less disruption and a smoother journey,” he added.
This transition is expected to help SMRT reduce operational costs, particularly increases in labor costs, by allowing maintenance teams to prioritize work more efficiently.
“It’s about empowering our employees to do more and supporting our growing rail network without the associated increase in headcount.”
The Jarvis pilot will run until the end of 2026 and is expected to be rolled out more widely expected SMRT will take several years to realize as it integrates decades-old systems, improves data quality, and installs sensors where they are needed.
Soh said training an AI model can take six months to a year. Depending on the amount and complexity of data collected, Systems are built, tested, and validated before deployment.
“AI can help us pre-empt problems rather than react to them,” he said, adding that with more than 2 million passenger journeys supported every day, even small improvements in reliability can have a big impact on passengers’ daily lives.
Still, Jarvis cannot predict all problems, especially sudden electrical failures and unexpected events.
Lack of data remains one of the key constraints, and organizations are still grappling with it. collect Train more accurate models using information from different systems.
Chris Sheria, senior vice president of technology for Oracle Japan and Asia Pacific, said this reflects a situation where larger and more significant uses of AI are being rolled out. Rather than a one-time transition from pilot to full implementation, it functions as a system that evolves and learns over time, gaining progressively more complex capabilities.
For SMRT, early versions of Jarvis were used to help engineers search technical manuals for instructions faster before proceeding to analyze data, flag potential failures, and proactively recommend actions.
“Eventually, Jarvis will learn enough and evolve to enhance its decision-making,” Sheria said.
