Artificial intelligence
Intelligent planning and Digital Twins – Industrie 4.0
Publiée le June 13, 2025
Publiée le June 13, 2025
On the shop floor, planning is often carried out manually or using load plan tables. This approach does not take into account the variability of operations (breakdowns, delays, absences), nor does it adapt to real-time events. Intelligent planning, coupled with digital twins, aims to represent the workshop virtually in order to simulate scenarios, optimize scheduling and react to contingencies.
A digital twin is a virtual replica of a physical system (production line, locomotive, infrastructure). It collects data in real time and can be used to simulate operations, breakdowns or interventions. The StartUs Insights article mentions the startup Quadrical AI, which provides a digital twin system for solar farms to monitor production and guide maintenance. Transposed to rail, this technology can create a twin of each locomotive and workshop to test scenarios (e.g.: what happens if a motor fails? what is the impact on planning?).
Real-time data collection: IoT sensors, embedded systems, HR information. Digital twins aggregate this data to reflect the state of the system.
Optimization engine: planning algorithms (e.g. genetic algorithms, heuristics or linear solvers) generate schedules that respect constraints (available resources, promised dates) and optimize a target (total time, cost, equipment utilization rate).
Scenario simulation: the user can run simulations (“What-If”). For example, if a locomotive breaks down, the tool proposes a new schedule and measures the impact on delivery dates.
Continuous adjustment: the system updates the schedule when real events occur (breakdowns detected by predictive maintenance, absence of a technician, late delivery of parts). AI recommends adjustments in real time.
upBus: in addition to its hybrid vehicles, the German startup offers real-time schedule optimization using AI and IoT. It combines analysis of data from sensors, cameras and IoT to predict breakdowns and adjust schedules.
4AI Systems: this Australian company is developing HORUS, a vision system that detects and classifies hazards around rail corridors. The data collected feeds a digital twin to analyze ideal and abnormal conditions and thus optimize operations.
AXO Track and Apital: their predictive maintenance solutions generate data that can be used for planning purposes. By detecting weaknesses in infrastructure or trains, they enable interventions to be planned without disrupting traffic.
RailVision Analytics: this Canadian startup combines fleet data with AI algorithms to provide driving recommendations (braking, acceleration) and reduce wear and tear and emissions. By adjusting driving behavior, it extends component life and modifies maintenance requirements.
CAMM/CMMS integration: the digital twin connects to production and maintenance systems. When a customer order arrives, the engine calculates the manufacturing (or maintenance) plan, taking into account human resources (HR schedules), tools and available parts (via OCR and stock optimization). APS (Advanced Planning & Scheduling) proposes an initial schedule.
Synchronization with predictive maintenance: if the predictive maintenance engine detects that a motor is likely to break down in two days’ time, the scheduling system brings the intervention forward and recalculates the workload plan.
Shared decision-making: managers can accept, modify or reject proposals. Human interaction remains essential to arbitrate between mathematical optimization and qualitative constraints (customer contracts, strategic priorities).
Monitoring and continuous improvement: feedback (e.g. past real time, unforeseen anomalies) feeds the system. AI models are re-trained to improve forecast quality.
Rail operators are already adopting intelligent planning solutions:
DB and ÖBB: the use of part identification tools (Partium) combined with intelligent planning has drastically reduced search and planning times. Digital twins are also being tested to simulate the impact of breakdowns on schedules.
Anonymized industrial: an engine plant has integrated an APS motor linked to a digital twin. When predictive maintenance signals a risk of breakdown on a line, scheduling automatically recomputes production orders and reschedules teams. The result: 15% higher productivity and reduced delivery delays.
The creation of a digital workshop twin would make it possible to simulate locomotive flows, plan interventions (maintenance and production) and visualize the impact of a breakdown or emergency on planning. Combined with predictive maintenance (article 1) and part recognition (article 2), it would constitute a global planning system capable of adapting in real time to unforeseen events.