Train Energy Management Machine Learning Synchronization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current train control systems lack reliable coordination and synchronization between lead and trailing locomotives, especially during communication degradation, and do not utilize machine learning for energy management to maintain synchronization between centralized and distributed train control models.
Innovation Solution
A train control system employing an energy management machine learning modeling engine with centralized and edge-based computer processing systems, utilizing data acquisition hubs to create and compare models for generating control commands, and adjusting throttle and braking requests based on learning function outputs to mitigate divergence between lead and trailing locomotives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If distributed control systems are used in train consists, then operational flexibility and redundancy are improved, but synchronization and coordination between lead and trailing locomotives deteriorate
Solution Approach 1:
The system continuously monitors operational parameters from both lead and trailing locomotives and uses machine learning models to compare actual performance against predicted performance. The feedback loop adjusts control commands in real-time to maintain synchronization, with the trailing locomotive receiving adjusted throttle and braking commands based on the learned divergence patterns between distributed and centralized control models.
Solution Approach 2:
A machine learning intermediary system acts as a mediator between the distributed control system on the trailing locomotive and the centralized control system on the lead locomotive. This intermediary learns the discrepancies between distributed and centralized control models and translates them into corrective control commands, enabling the trailing locomotive to operate autonomously while maintaining synchronization with the lead locomotive.
2Extent of automation
If machine learning models are deployed on distributed edge-based systems, then operational autonomy is improved, but model synchronization with centralized control deteriorates
Solution Approach 1:
The machine learning model is trained in advance using historical operational data from both centralized and distributed control systems. This preliminary training enables the model to predict divergence patterns between centralized and distributed control models before actual operation begins. The pre-trained model then operates autonomously on the trailing locomotive, applying learned corrections without requiring real-time communication with the centralized system.
3Reliability
If communication bandwidth is increased for real-time data transmission, then control synchronization is improved, but system complexity and cost deteriorate
Solution Approach 1:
The system extracts and processes only the essential operational parameters needed for synchronization control, rather than transmitting all available sensor data. The machine learning model on the trailing locomotive processes local sensor data autonomously, extracting only the necessary control commands and sending them to the lead locomotive. This extraction approach maintains synchronization reliability while minimizing communication bandwidth requirements and system complexity.
Data Source
AI summary
A train control system uses artificial intelligence for maintaining synchronization between centralized and distributed train control models. A machine learning engine receives training data from a data acquisition hub, a first set of output control commands from a centralized virtual system modeling engine, and a second set of output control commands from a distributed virtual system modeling engine. The machine learning engine compares the first set of output control commands and the second set of output control commands, and trains a learning system using the training data to enable the machine learning engine to safely mitigate any difference between the first and second sets of output control commands using a learning function including at least one learning parameter.