Automated Quality Control and Predictive Maintenance of Steel Rail Infrastructure via Federated Learning and Multi-Modal Data Fusion
Year: 2025
Type: article
Abstract: **Abstract:** This paper proposes a novel framework for automated quality control and predictive maintenance of steel rail infrastructure utilizing federated learning and multi-modal data fusion. Existing rail inspection methods are often reactive, expensive, and require significant downtime. Our system addresses these limitations by leveraging a decentralized approach that allows multiple railway operators to collaboratively train a robust predictive model without sharing sensitive raw data. We integrate data from diverse sources, including ultrasonic rail flaw detection (URFD), track geometry measurements, environmental sensors (temperature, humidity, precipitation), and historical maintenance records. This multi-modal data is fused using a transformer-based architecture and fed into a federated learning framework, enabling accurate defect prediction, optimized maintenance scheduling, and reduced operational costs. The system demonstrates a potential 15-20% reduction in maintenance costs and a 10% increase in rail availability within a 5-year implementation horizon. --- *This document is a part of the Freederia Research Archive. Explore our complete collection of advanced research at [en.freederia.com](https://en.freederia.com), or visit our main portal at [freederia.com](https://freederia.com) to learn more about our mission and other initiatives.*
Source: Figshare
Authors: KYUNGJUN, LIM, Researcher, Freederia AI
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6902268436
FWCI:
Cites:
Cited by:
Related to: