Real-time Detection and Correlation of Infrastructure Damage Events

AI Quick Summary

The proposed project aims to tackle the challenge of detecting and correlating damage events on railway infrastructure caused by electric arcs at pantographs, which typically go unnoticed until residue inspection on locomotives. We propose to evaluate a solution that provides real-time monitoring and pinpoint the exact location and timing of such events by developing a compact, energy-efficient edge computing solution equipped with GPS and various sensors—including visual, audio, and electromagnetic field detectors. This system, which could be mounted on locomotives or stationary infrastructure and powered by solar cells and batteries, promises to enhance maintenance efficiency and reduce downtime. Collaboration with the Institute of Sensors and Electronics at the University of Applied Sciences Northwestern Switzerland (FHNW) will ensure robust design, validation, and solution implementation. The proposed idea addresses the SBB challenge of "Maintenance and Repair."Image courtesy of publication "Study on electric spark discharge between pantograph and catenary in electrified railway." (http://dx.doi.org/10.1049/els2.12043)Vision and JustificationDisclaimer: The idea is based on previous exchanges with SBB personnel outside the ideation workshop from a while ago. Thus, it must be (re-)validated by maintenance and repair specialists.The proposed project addresses the challenge of detecting and correlating damage events on railway infrastructure, specifically focusing on (but not necessarily limited to) issues caused by electric arcs at pantographs that can damage overhead lines. Typically, these events are detected only retrospectively through residue inspection on locomotives, making it challenging to pinpoint the exact location and time of occurrence. Since the locomotive travel path is known and can be reconstructed, narrowing down the infrastructure track such an event could have occurred is possible. But, still, this leaves open many possibilities.We propose developing an edge computing solution with a small embedded board equipped with GPS and sensors to monitor pantographs in real time to detect damage events and correlate them with precise time and location.Proposed Idea: Innovative Real-Time Monitoring and DetectionOur solution aims to provide continuous, real-time monitoring of pantographs to detect damage events as they occur, using visual data and/or other sensor inputs. This would enable immediate identification of the precise location and timing of electric arcs, facilitating quicker repairs and reducing potential downtime. The sensor suite may include a video feed to observe electric arcs visually, microphones to observe them through sound, capacitive or inductive sensors to measure field irregularities associated with electric arcs, and other novel data sources that could enhance detection accuracy. The actual feasibility of the sensors is TBD, and so are the required training data and sensor data analysis methods.We propose encapsulating an embedded device with high-processing capabilities and low power needs (e.g., Xilinx System-on-Chip) along with the sensors described below into a compact, all-in-one analysis lab. This lab can be easily mounted on top of locomotives or at stationary locations such as power lines, train stations, etc., to observe the infrastructure constantly. Such small devices could likely be powered by solar cells and batteries, making them energy-efficient and sustainable for long-term use. For a proof-of-concept, these autonomous units do not require integration with the locomotive or other existing systems, streamlining deployment and testing processes.CollaborationWe propose collaborating with the Institute of Sensors and Electronics (ISE) at the University of Applied Sciences Northwestern Switzerland (FHNW), leveraging their extensive experience in sensor design and validation. The ISE team includes a Master's student working on embedding machine learning models with embedded hardware, who can contribute to implementing this project. Our contribution would encompass hardware and sensor integration, support for detection processes, and data analysis, ensuring robust and validated real-time monitoring solutions.BackgroundIn previous projects, we have successfully applied machine learning models on edge devices like Xilinx FPGA SoC to detect and track objects autonomously in real-time video feeds. Our expertise in real-time edge computing solutions will be leveraged to design and implement the proposed monitoring system, ensuring high efficiency and reliability at the source of data collection.There are several interesting papers published over the last few years on detecting such electric arc events that can serve as valuable input to the project.Deep learning-based Arc detection in pantograph-catenary systems (https://ieeexplore.ieee.org/document/8266412)A Novel Arc Detection Method for DC Railway Systems (https://www.mdpi.com/1996-1073/14/2/444)A novel arc detection and identification method in pantograph-catenary system based on deep learning (https://www.nature.com/articles/s41598-025-88109-x)

Project Idea Metadata

• Project Idea Name: Real-time Detection and Correlation of Infrastructure Damage Events
• Date: 3/16/2025 8:22:42 PM
• Administrators:
  • Laszlo Etesi

Project Idea Description

The proposed idea addresses the SBB challenge of "Maintenance and Repair."

Image courtesy of publication "Study on electric spark discharge between pantograph and catenary in electrified railway." (http://dx.doi.org/10.1049/els2.12043)

Vision and Justification

Disclaimer: The idea is based on previous exchanges with SBB personnel outside the ideation workshop from a while ago. Thus, it must be (re-)validated by maintenance and repair specialists.

The proposed project addresses the challenge of detecting and correlating damage events on railway infrastructure, specifically focusing on (but not necessarily limited to) issues caused by electric arcs at pantographs that can damage overhead lines. Typically, these events are detected only retrospectively through residue inspection on locomotives, making it challenging to pinpoint the exact location and time of occurrence. Since the locomotive travel path is known and can be reconstructed, narrowing down the infrastructure track such an event could have occurred is possible. But, still, this leaves open many possibilities.

We propose developing an edge computing solution with a small embedded board equipped with GPS and sensors to monitor pantographs in real time to detect damage events and correlate them with precise time and location.

Proposed Idea: Innovative Real-Time Monitoring and Detection

Our solution aims to provide continuous, real-time monitoring of pantographs to detect damage events as they occur, using visual data and/or other sensor inputs. This would enable immediate identification of the precise location and timing of electric arcs, facilitating quicker repairs and reducing potential downtime. The sensor suite may include a video feed to observe electric arcs visually, microphones to observe them through sound, capacitive or inductive sensors to measure field irregularities associated with electric arcs, and other novel data sources that could enhance detection accuracy. The actual feasibility of the sensors is TBD, and so are the required training data and sensor data analysis methods.

We propose encapsulating an embedded device with high-processing capabilities and low power needs (e.g., Xilinx System-on-Chip) along with the sensors described below into a compact, all-in-one analysis lab. This lab can be easily mounted on top of locomotives or at stationary locations such as power lines, train stations, etc., to observe the infrastructure constantly. Such small devices could likely be powered by solar cells and batteries, making them energy-efficient and sustainable for long-term use. For a proof-of-concept, these autonomous units do not require integration with the locomotive or other existing systems, streamlining deployment and testing processes.

Collaboration

We propose collaborating with the Institute of Sensors and Electronics (ISE) at the University of Applied Sciences Northwestern Switzerland (FHNW), leveraging their extensive experience in sensor design and validation. The ISE team includes a Master's student working on embedding machine learning models with embedded hardware, who can contribute to implementing this project. Our contribution would encompass hardware and sensor integration, support for detection processes, and data analysis, ensuring robust and validated real-time monitoring solutions.

Background

In previous projects, we have successfully applied machine learning models on edge devices like Xilinx FPGA SoC to detect and track objects autonomously in real-time video feeds. Our expertise in real-time edge computing solutions will be leveraged to design and implement the proposed monitoring system, ensuring high efficiency and reliability at the source of data collection.

There are several interesting papers published over the last few years on detecting such electric arc events that can serve as valuable input to the project.

• Deep learning-based Arc detection in pantograph-catenary systems (https://ieeexplore.ieee.org/document/8266412)
• A Novel Arc Detection Method for DC Railway Systems (https://www.mdpi.com/1996-1073/14/2/444)
• A novel arc detection and identification method in pantograph-catenary system based on deep learning (https://www.nature.com/articles/s41598-025-88109-x)

The proposed project aims to tackle the challenge of detecting and correlating damage events on railway infrastructure caused by electric arcs at pantographs, which typically go unnoticed until residue inspection on locomotives. We propose to evaluate a solution that provides real-time monitoring and pinpoint the exact location and timing of such events by developing a compact, energy-efficient edge computing solution equipped with GPS and various sensors—including visual, audio, and electromagnetic field detectors. This system, which could be mounted on locomotives or stationary infrastructure and powered by solar cells and batteries, promises to enhance maintenance efficiency and reduce downtime. Collaboration with the Institute of Sensors and Electronics at the University of Applied Sciences Northwestern Switzerland (FHNW) will ensure robust design, validation, and solution implementation.