Axle Counting Device Using Reinforcement Learning for Track Occupancy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing axle counting methods for determining track section occupancy are unreliable, especially in curved or switch areas, and fail to adapt to changing train usage patterns, leading to potential collisions and inefficiencies.

Innovation Solution

The implementation of a computer-aided axle counting system using reinforcing learning algorithms, specifically neuronal networks, to optimize parameter settings for detecting wheel influences, allowing for adaptive and reliable occupancy determination across various track sections without prior knowledge of individual application situations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional axle counting methods are used, then the system is simple and cost-effective, but the reliability is insufficient especially in curved or switch areas

Engineering Contradiction:
Improveoccupancy determination reliabilityVSAvoidaxle counting system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system changes parameters dynamically by using reinforcement learning to continuously optimize detection thresholds and parameters based on actual operating conditions. The algorithm adjusts parameters such as vibration thresholds, time windows, and pattern recognition criteria to adapt to different track sections, curves, and switch areas, thereby improving reliability without requiring complex hardware modifications

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The axle counting system performs self-optimization through the reinforcement learning algorithm that automatically learns and adapts to specific track conditions without external intervention. The system uses historical data and continuous feedback to self-adjust detection parameters, eliminating the need for manual calibration and reducing dependency on expert knowledge for each track section

Inventive Principle:
Principle #25Self-service

2Adaptability or versatility

If fixed parameter settings are used for axle counting, then the system is simple to operate, but it cannot adapt to changing train usage patterns and individual track section requirements

Engineering Contradiction:
Improveadaptability to track section characteristicsVSAvoidparameter optimization complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system transitions from static fixed parameters to dynamic adaptive parameters through reinforcement learning. The detection parameters are continuously updated based on real-time performance feedback and changing operational conditions, allowing the system to adapt to different train types, speeds, and track conditions while maintaining a unified hardware platform

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The reinforcement learning algorithm implements continuous feedback loops where occupancy determination results are fed back to optimize future detections. The system learns from correct and incorrect detections, adjusting parameters to improve accuracy over time for each specific track section, thereby achieving high adaptability without requiring complex manual configuration

Inventive Principle:
Principle #23Feedback

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach enhances the reliability and adaptability of axle counting, reducing inaccuracies and costs by automatically optimizing the system during operation, ensuring consistent surveillance and adjusting to changing conditions, thereby improving the overall reliability of train operations.

Implementation Method 1

the physical wheel influence of trains traveling on the track section is converted into a reference voltage

Methodology Applied
Scientific EffectWheel influence conversion: Piezoelectric Effect

Data Source

PatentEP3938265B1Method for determining a track occupancy, and axle counting device
Publication Date: 2025.02.12 SIEMENS MOBILITY GMBH
  • EP3938265B1 patent drawingFigure 1
  • EP3938265B1 patent drawingFigure 2

AI summary

The invention relates to a method for computer-aided determination of the occupancy state of a track section (GA), wherein the physical wheel influence of the train (ZUG) travelling along the track section (GA) is converted into a rectified voltage, the rectified voltage is read in the form of raw data into a processor (PRC) via an analogue-to-digital converter (ADU), and an assessment takes place with the aid of at least one algorithm in order to examine whether or not a wheel influence is present, on which basis an influencing state is derived. The influencing state is output in order to determine the occupancy state. The algorithm uses consolidating learning to optimise the parameters taken into consideration for the identification to identify wheel influences within the scope of the assessment. The invention also relates to an axle counting device (AZE) for identifying wheel influences on a track section (GA), and to a computer program product.