Artificial Intelligence Across Rail Networks
Loading results...
Deployment date records the documented start of use, including pilots. Source published is the article date. Source last accessed records a successful editorial retrieval, not confirmation of current deployment. Unknown dates are shown explicitly; dates use YYYY-MM-DD, YYYY-MM or YYYY.
| Country | Organization | Category | AI Application | Source & dates |
|---|---|---|---|---|
| China | Shanghai Mission Information Technology (Mingxiao) | General Office Work |
Reported maintenance-project use: a DeepSeek-based knowledge assistant draws on specialist documentation and historical maintenance records for passenger information, public-address, CCTV and power systems. Conversational retrieval supports technicians in rolling-stock and communications maintenance projects; the supplier does not identify the operators.
|
View source
|
| Australia | Downer XDNA / Sydney Trains | Predictive Maintenance |
Reported fleet validation: wayside temperature, acoustic and wheel-impact data identify at-risk bearings on A- and B-set trains before failure. XDNA describes these models within its rail AI capabilities.
|
View source
|
| Australia | Downer XDNA / Sydney Trains / HCMT | Predictive Maintenance |
Reported application: algorithms forecast pantograph wear on A- and B-sets and detect defective HCMT pantographs before failure. The brochure does not identify the model architecture.
|
View source
|
| Australia | Downer XDNA | Predictive Maintenance |
Reported application: Neuroverse combines train telemetry with historical data to anticipate HVAC and auxiliary-power failures and flag thermal issues. Individual operator deployments are not named.
|
View source
|
| Australia | Downer XDNA | General Office Work |
Reported internal use: TrainBrain retrieves and summarizes engineering and maintenance documentation for technical questions, requirements validation and staff onboarding.
|
View source
|
| Australia | Downer XDNA | Asset and Infrastructure Monitoring |
Developed models: machine learning and computer vision identify trains and their depot roads, replacing manual sighting. The source provides no named deployment site or commissioning date.
|
View source
|
| Australia | Downer XDNA / Future Maintenance Technologies (FMT) | Asset and Infrastructure Monitoring |
Pilot inspections: Downer and FMT tested autonomous rolling-stock inspection tools at Auburn and Pakenham on Waratah and HCMT fleets. Downer identifies TRES as using AI and machine learning with LiDAR, lasers and optical sensors. The sources do not establish fleet-wide regular operation.
|
View source
|
| New Zealand | Downer XDNA / City Rail Link Auckland | Engineering and Design |
Reported project use: Neuroverse Cognition applies agentic AI to connect engineering requirements and assurance evidence. Downer reports its use on Auckland City Rail Link to support testing, commissioning and handover. The source does not document the deployment start.
|
View source
|
| China | China Railway Tunnel Group / National Key Laboratory for Tunnel Boring Machines and Intelligent Maintenance | Engineering and Design |
Reported field validation: Pioneer Tunnel Large Model combines general and specialist models for tunnel engineering, including construction assessment and the Tunnel Hero knowledge assistant. China Railway reports validation at plateau railway tunnels, Chongtai Yangtze River Tunnel and Shenzhen-Jiangmen Railway Pearl River Estuary Tunnel. Function-level coverage at each site is unspecified.
|
View source
|
| Belgium | Railnova | Maintenance |
AI-based defect classification in Railfleet using Large Language Models and multi-agent AI systems to analyse defect descriptions and suggest the five most relevant maintenance categories
|
View source
|
| Belgium | Railnova | Maintenance |
AI-based duplicate corrective defect detection in Railfleet using Large Language Models to identify potentially duplicated maintenance defect reports while leaving the final decision to the user
|
View source
|
| Belgium | Railnova | Predictive Maintenance |
Custom machine-learning-based alerts in Railgenius analyse railway vehicle telemetry and detect complex patterns that cannot be captured through standard thresholds or rule-based monitoring
|
View source
|
| Belgium | Railnova | Predictive Maintenance |
Data-science and machine-learning services for rail fleets including custom predictive analytics anomaly detection failure-pattern analysis and ML-based monitoring solutions
|
View source
|
| France | Europorte / Railnova | Predictive Maintenance |
Predictive maintenance algorithms for locomotives developed using fleet operational and telemetry data in cooperation between Europorte Railnova and École Polytechnique to identify technical issues before failures occur
|
View source
|