In-Transit Railcar Defect Detection with Machine Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing railcar inspection methods are labor-intensive and limited to stationary inspections, failing to effectively detect defects during transport due to extreme conditions, which can lead to component wear and dislodgment.

Innovation Solution

A machine-learning framework that utilizes field cameras and multiple machine-learning algorithms to analyze railcar images in real-time, identifying defects such as missing, broken, or displaced components, and initiates remediation operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If visual inspection of railcar components is performed manually, then inspection can be conducted, but it is labor intensive and only available while the train is stopped at a station

Engineering Contradiction:

Improveinspection availabilityVSAvoidinspection downtime

Core Design Contradiction:

Ease of operationVSLoss of time

Solution Approach 1:

The patent replaces manual visual inspection with an automated machine-learning-based image analysis system. Field cameras capture images of railcar components during transport, and machine-learning algorithms automatically analyze these images to detect defects, eliminating the need for manual inspection and enabling continuous monitoring without stopping the train.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If manual inspection is performed only when the train is stopped, then labor costs are reduced, but defect detection is delayed and components may become dislodged due to vibration and temperature changes

Engineering Contradiction:

Improvecomponent integrityVSAvoiddetection delay

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system enables continuous inspection of railcar components during transport by capturing images with field cameras as the train moves through the field of view. The machine-learning algorithms continuously analyze these images in real-time, providing ongoing monitoring of component integrity without interruption, thereby detecting defects immediately rather than delaying detection until the train stops.

Inventive Principle:

Principle #20Continuity of useful action

3Measurement precision

If automated machine-learning inspection is implemented, then inspection accuracy and continuity are improved, but system complexity increases

Engineering Contradiction:

Improvedefect detection accuracyVSAvoidinspection system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces field cameras as an intermediary device to capture images of railcar components during transport. These cameras serve as the interface between the moving train and the stationary analysis system, allowing images to be captured and transmitted for machine-learning analysis without requiring complex moving inspection equipment, thereby managing system complexity while achieving automated inspection.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12462362B2Machine-learning framework for detecting defects or conditions of railcar systems

Publication Date: 2025.11.04 NORFOLK SOUTHERN CORP

AI summary

A computer-implemented method in which one or more processing devices perform operations may include obtaining a field image of a railcar collected from a field camera system and applying a machine-learning algorithm to the field image to generate a machine-learning algorithm output. The method may also include performing a post-processing operation on the machine-learning algorithm output to generate a filtered machine-learning algorithm output. Further, the method may include detecting a defect of the railcar using the filtered machine-learning algorithm output.