NEXTRAIL GmbH
NEXTRAIL GmbH: Software solution for an ETCS Gamma brake model
NEXTRAIL GmbH: Software solution for an ETCS Gamma brake model
NEXTRAIL GmbH: Software solution for an ETCS Gamma brake model
The goal of this project was to develop a highly scalable, vectorized simulation framework specifically designed to compute the Gamma braking parameters for ETCS-compliant vehicles.
The goal of this project was to develop a highly scalable, vectorized simulation framework specifically designed to compute the Gamma braking parameters for ETCS-compliant vehicles.
The goal of this project was to develop a highly scalable, vectorized simulation framework specifically designed to compute the Gamma braking parameters for ETCS-compliant vehicles.
Logistics
Data Science
Python
Клієнт
Клієнт
NEXTRAIL GmbH
NEXTRAIL GmbH
NEXTRAIL GmbH
Дата
Дата
5 months from concept to full deployment
5 months from concept to full deployment
5 months from concept to full deployment
Роль
Роль
Development partner
Development partner
Development partner
Вебсайт
Вебсайт
nextrail.com
nextrail.com
nextrail.com
NEXTRAIL GmbH
NEXTRAIL GmbH
Client overview
Client overview
Client overview
NEXTRAIL is a key partner in modernizing 21st-century rail transport, dedicated to improving the industry’s capacity, efficiency, safety, and environmental footprint. They offer services that span conception, system engineering, data management, and safety assessment, all the way to supporting the integration of digital technologies onto rolling stock.
NEXTRAIL is a key partner in modernizing 21st-century rail transport, dedicated to improving the industry’s capacity, efficiency, safety, and environmental footprint. They offer services that span conception, system engineering, data management, and safety assessment, all the way to supporting the integration of digital technologies onto rolling stock.
NEXTRAIL is a key partner in modernizing 21st-century rail transport, dedicated to improving the industry’s capacity, efficiency, safety, and environmental footprint. They offer services that span conception, system engineering, data management, and safety assessment, all the way to supporting the integration of digital technologies onto rolling stock.
The challenge
The challenge
The challenge
The digitalization of railway systems is creating new opportunities for vastly more efficient operations. European Train Control System (ETCS) is the core of this transformation. ETCS is the next-generation standard that relies heavily on accurate vehicle braking models to guarantee operational performance and maximize line capacity.
The digitalization of railway systems is creating new opportunities for vastly more efficient operations. European Train Control System (ETCS) is the core of this transformation. ETCS is the next-generation standard that relies heavily on accurate vehicle braking models to guarantee operational performance and maximize line capacity.
The Gamma model serves as a foundational element, being widely adopted across railcars and trains due to its ability to provide a precise physical description of braking characteristics. This precision is essential for implementing Automatic Train Operation (ATO).
The primary goal of this project was to develop a highly scalable, vectorized simulation framework specifically designed to compute these Gamma braking parameters for ETCS-compliant vehicles. We initiated the engagement with a workshop involving the Lead Data Scientist and two Data Engineers to properly estimate and plan the comprehensive project development timeline.
The Gamma model serves as a foundational element, being widely adopted across railcars and trains due to its ability to provide a precise physical description of braking characteristics. This precision is essential for implementing Automatic Train Operation (ATO).
The primary goal of this project was to develop a highly scalable, vectorized simulation framework specifically designed to compute these Gamma braking parameters for ETCS-compliant vehicles. We initiated the engagement with a workshop involving the Lead Data Scientist and two Data Engineers to properly estimate and plan the comprehensive project development timeline.
The Gamma model serves as a foundational element, being widely adopted across railcars and trains due to its ability to provide a precise physical description of braking characteristics. This precision is essential for implementing Automatic Train Operation (ATO).
The primary goal of this project was to develop a highly scalable, vectorized simulation framework specifically designed to compute these Gamma braking parameters for ETCS-compliant vehicles. We initiated the engagement with a workshop involving the Lead Data Scientist and two Data Engineers to properly estimate and plan the comprehensive project development timeline.
The solution was engineered to meet modern compliance and engineering standards:
• It accounts for component deviations and failure probabilities, aligning with EN 17997 standards.
• It supports modern software engineering workflows, including cloud-based execution, CI/CD pipelines, and extensive data management, ensuring robustness and rapid deployment.
The solution was engineered to meet modern compliance and engineering standards:
• It accounts for component deviations and failure probabilities, aligning with EN 17997 standards.
• It supports modern software engineering workflows, including cloud-based execution, CI/CD pipelines, and extensive data management, ensuring robustness and rapid deployment.
Model architecture
Model architecture
Model architecture
To ensure the ETCS operates safely and efficiently, detailed and accurate modeling of vehicle braking behavior is essential. This requires a robust simulation environment capable of handling complex mechanical and electronic interactions, component variability, and failure risks.
The model includes:
• Hierarchical system: braking systems are modeled as blocks → sub-blocks → individual components.
• Component parameters: each element has a failure probability and deviation coefficient.
• Dependencies: inter-component interactions and alpha parameters define how component deviations propagate through the system.
• Train structure parsing: the simulator automatically reads and structures vehicle configurations from input data, enabling flexible evaluation of different trainsets and configurations.
• Graph representation: braking system hierarchy and interactions are visualized as graphs, aiding debugging, analysis, and documentation.
To ensure the ETCS operates safely and efficiently, detailed and accurate modeling of vehicle braking behavior is essential. This requires a robust simulation environment capable of handling complex mechanical and electronic interactions, component variability, and failure risks.
The model includes:
• Hierarchical system: braking systems are modeled as blocks → sub-blocks → individual components.
• Component parameters: each element has a failure probability and deviation coefficient.
• Dependencies: inter-component interactions and alpha parameters define how component deviations propagate through the system.
• Train structure parsing: the simulator automatically reads and structures vehicle configurations from input data, enabling flexible evaluation of different trainsets and configurations.
• Graph representation: braking system hierarchy and interactions are visualized as graphs, aiding debugging, analysis, and documentation.
To ensure the ETCS operates safely and efficiently, detailed and accurate modeling of vehicle braking behavior is essential. This requires a robust simulation environment capable of handling complex mechanical and electronic interactions, component variability, and failure risks.
The model includes:
• Hierarchical system: braking systems are modeled as blocks → sub-blocks → individual components.
• Component parameters: each element has a failure probability and deviation coefficient.
• Dependencies: inter-component interactions and alpha parameters define how component deviations propagate through the system.
• Train structure parsing: the simulator automatically reads and structures vehicle configurations from input data, enabling flexible evaluation of different trainsets and configurations.
• Graph representation: braking system hierarchy and interactions are visualized as graphs, aiding debugging, analysis, and documentation.
Methodology
Methodology
Methodology
To successfully implement the system and calculate accurate Gamma braking parameters, the team defined a tailored approach centered on the following methodology:
• Monte Carlo simulation: Gamma braking parameters are derived from measurement data, statistical data, error rates, and brake architecture.
• Vectorized and parallelized computation: all simulations are performed using JAX with vectorized operations, reducing loops and conditionals. Multiprocessing allows distribution across CPU cores for additional speed-up.
• Database integration: simulation results are stored in PostgreSQL or SQLite databases, enabling structured storage, retrieval, and further analysis.
• Automated workflows: simulations can be triggered via GitHub Actions, with automated server setup, configuration selection, and execution of specific trainset simulations.
• Test coverage: core modules, including simulation engine and graph transformations, are fully covered by automated tests, ensuring reliability and reproducibility.
• Documentation: mathematical formulas, model architecture, and assumptions are documented in a structured manner, supporting reproducibility and knowledge transfer.
To successfully implement the system and calculate accurate Gamma braking parameters, the team defined a tailored approach centered on the following methodology:
• Monte Carlo simulation: Gamma braking parameters are derived from measurement data, statistical data, error rates, and brake architecture.
• Vectorized and parallelized computation: all simulations are performed using JAX with vectorized operations, reducing loops and conditionals. Multiprocessing allows distribution across CPU cores for additional speed-up.
• Database integration: simulation results are stored in PostgreSQL or SQLite databases, enabling structured storage, retrieval, and further analysis.
• Automated workflows: simulations can be triggered via GitHub Actions, with automated server setup, configuration selection, and execution of specific trainset simulations.
• Test coverage: core modules, including simulation engine and graph transformations, are fully covered by automated tests, ensuring reliability and reproducibility.
• Documentation: mathematical formulas, model architecture, and assumptions are documented in a structured manner, supporting reproducibility and knowledge transfer.
To successfully implement the system and calculate accurate Gamma braking parameters, the team defined a tailored approach centered on the following methodology:
• Monte Carlo simulation: Gamma braking parameters are derived from measurement data, statistical data, error rates, and brake architecture.
• Vectorized and parallelized computation: all simulations are performed using JAX with vectorized operations, reducing loops and conditionals. Multiprocessing allows distribution across CPU cores for additional speed-up.
• Database integration: simulation results are stored in PostgreSQL or SQLite databases, enabling structured storage, retrieval, and further analysis.
• Automated workflows: simulations can be triggered via GitHub Actions, with automated server setup, configuration selection, and execution of specific trainset simulations.
• Test coverage: core modules, including simulation engine and graph transformations, are fully covered by automated tests, ensuring reliability and reproducibility.
• Documentation: mathematical formulas, model architecture, and assumptions are documented in a structured manner, supporting reproducibility and knowledge transfer.
Optimization
Optimization
To guarantee the simulation framework delivers the stringent speed and scalability required for high-throughput engineering analysis, we implemented a set of core strategic optimizations:
To guarantee the simulation framework delivers the stringent speed and scalability required for high-throughput engineering analysis, we implemented a set of core strategic optimizations:
• Matrix transformations for fast propagation of component deviations.
• Vectorized computations eliminate iterative loops and maximize GPU usage.
• Multiprocessing and GPU acceleration enable the simultaneous simulation of hundreds of scenarios.
• Matrix transformations for fast propagation of component deviations.
• Vectorized computations eliminate iterative loops and maximize GPU usage.
• Multiprocessing and GPU acceleration enable the simultaneous simulation of hundreds of scenarios.
• Graph node reduction: consecutive components with independent distributions can be merged into a “supercomponent” with a combined distribution, reducing the number of nodes in the graph and therefore the number of computations.
• Graph node reduction: consecutive components with independent distributions can be merged into a “supercomponent” with a combined distribution, reducing the number of nodes in the graph and therefore the number of computations.
Results
Results
Results
The framework’s approach and optimizations resulted in:
The framework’s approach and optimizations resulted in:
Accurate estimation of the train braking coefficient based on vehicle-specific parameters and system configuration.
Accurate estimation of the train braking coefficient based on vehicle-specific parameters and system configuration.
Accurate estimation of the train braking coefficient based on vehicle-specific parameters and system configuration.
Significant speed-up due to node reduction in the component graph.
Significant speed-up due to node reduction in the component graph.
Significant speed-up due to node reduction in the component graph.
Fast sensitivity analysis of individual components on overall braking performance.
Fast sensitivity analysis of individual components on overall braking performance.
Fast sensitivity analysis of individual components on overall braking performance.
Performance improvements: tens of thousands of times faster than classical approaches.
Performance improvements: tens of thousands of times faster than classical approaches.
Performance improvements: tens of thousands of times faster than classical approaches.
Supports large-scale scenario analyses under varying operational conditions (temperature, load, wear), improving operational stability and performance.
Supports large-scale scenario analyses under varying operational conditions (temperature, load, wear), improving operational stability and performance.
Supports large-scale scenario analyses under varying operational conditions (temperature, load, wear), improving operational stability and performance.
Database storage and visualization allow for easy inspection, reporting, and verification of simulation results.
Database storage and visualization allow for easy inspection, reporting, and verification of simulation results.
Database storage and visualization allow for easy inspection, reporting, and verification of simulation results.
Conclusions
Conclusions
Conclusions
Key conclusions from the project include:
Complex hierarchical braking systems can be modeled efficiently and accurately using vectorized probabilistic simulations to compute the train braking coefficient.
• Graph node reduction by merging consecutive components with independent distributions significantly decreases computation time without sacrificing accuracy.
• Integration of vectorization, multiprocess parallelism, and GPU acceleration ensures extremely high computational throughput.
• The framework supports modern software engineering practices, including automated CI/ CD workflows via GitHub Actions, database storage (PostgreSQL/SQLite), visualization of braking system graphs, and reproducible documentation with mathematical formulas.
• Comprehensive test coverage of core modules ensures reliability and reproducibility of simulation results.
Overall, the solution provides actionable insights for vehicle design, optimization, and operational planning, demonstrating the combination of vehicle technology expertise and software engineering in scalable, cloudready solutions.
Key conclusions from the project include:
Complex hierarchical braking systems can be modeled efficiently and accurately using vectorized probabilistic simulations to compute the train braking coefficient.
• Graph node reduction by merging consecutive components with independent distributions significantly decreases computation time without sacrificing accuracy.
• Integration of vectorization, multiprocess parallelism, and GPU acceleration ensures extremely high computational throughput.
• The framework supports modern software engineering practices, including automated CI/ CD workflows via GitHub Actions, database storage (PostgreSQL/SQLite), visualization of braking system graphs, and reproducible documentation with mathematical formulas.
• Comprehensive test coverage of core modules ensures reliability and reproducibility of simulation results.
Overall, the solution provides actionable insights for vehicle design, optimization, and operational planning, demonstrating the combination of vehicle technology expertise and software engineering in scalable, cloudready solutions.
Key conclusions from the project include:
Complex hierarchical braking systems can be modeled efficiently and accurately using vectorized probabilistic simulations to compute the train braking coefficient.
• Graph node reduction by merging consecutive components with independent distributions significantly decreases computation time without sacrificing accuracy.
• Integration of vectorization, multiprocess parallelism, and GPU acceleration ensures extremely high computational throughput.
• The framework supports modern software engineering practices, including automated CI/ CD workflows via GitHub Actions, database storage (PostgreSQL/SQLite), visualization of braking system graphs, and reproducible documentation with mathematical formulas.
• Comprehensive test coverage of core modules ensures reliability and reproducibility of simulation results.
Overall, the solution provides actionable insights for vehicle design, optimization, and operational planning, demonstrating the combination of vehicle technology expertise and software engineering in scalable, cloudready solutions.
Download extended case study in .pdf
© 2025 Brightgrove. Всі права захищені.
© 2025 Brightgrove. Всі права захищені.
© 2025 Brightgrove. Всі права захищені.
© 2025 Brightgrove. Всі права захищені.