Industrial AI for secure and resilient data and model sharing in railways

(AI Factory /RAILWAY)

The purpose of this project is to develop an end-to-end maintenance decision pipeline for Railway Overhead Catenary (ROC). ROC due to its span over large region is exposed to varying weather conditions, is in different states of maintenance, and due to long life, many standards are in use at any given time. Maintenance decisions in such complex scenarios depend on a holistic understanding of the cause-and-effect relationship.

Complex systems comprise numerous independent, interacting entities. They are considered complex due to their behaviour, which is not explicitly designed or intended during integration. Maintenance decisions and maintenance of complex systems depend on initial understanding and the development of knowledge about the system over its lifetime. Improvements in managing the knowledge about the industrial assets and deploying the available knowledge are critical for the maintenance decision process. Advancements in knowledge, system management methods, and technology can support this.

Developing and managing complex systems in science, engineering, business, and other fields reveals a recurring pattern: the complex system is greater than the sum of its parts. Complex systems exhibit emergent behaviour and value. This implies complex systems are operated and maintained using limited methods and knowledge. Therefore, to effectively respond to emergent behaviour, the focus should shift from local optimisation to a holistic outcome of the complex system.

Hence, a system-of-systems approach was developed during this project to address the various artefacts interacting with the ROC.

As a part of the development, the following artefacts have been developed:

• A data processing pipeline for point cloud data to classify and extract the geometric structure of individual cables of ROC
• An architecture to support microservices to support integration of stakeholders providing support services such as data acquisition, point cloud processing, and anomaly visualisation support.
• A hybrid digital twin to integrate information from ROC geometry with

The work resulted in a licentiate thesis and a doctoral thesis:

• Patwardhan, A. (2022). Enablement of digital twins for railway overhead catenary system.
• Patwardhan, A. (2024). A Novel Approach to Developing Digital Twin in Maintenance Utilising Industrial Artificial Intelligence.