Inzu Group – Predictive Maintenance of Railway Diversion Elements (SMART DIVERTER)
Inzu Group
Sector: Mobility
Business Case
Railway diversions are crucial components of the infrastructure and are significantly impacted in their regular operation. Additionally, the construction material, manganese steel, while noble, tends to develop cracks in hidden areas and is resistant to traditional inspection methods, such as ultrasound. New detection techniques and AI are applied to detect these problems.
Objectives
Develop an automatic detection and identification system based on self-learning using decentralised architectures with real-time information, enabling continuous assessment for predictive maintenance and ultrasonic inspection, with federated IoT monitoring and machine learning.
Use case
Advanced self-learning ML techniques are used for accurate detection and identification of railway crossing faults in a multi-sensor environment, with a scalable and flexible decentralised architecture and rapid deployment, optimising data processing efficiency and enabling real-time updates for proactive maintenance strategies.
Infrastructure
Edge Computing
Technology
Machine learning and deep learning
Data
Private time series datasets
Resources
It was necessary to involve experts in HW, FW, SW development, mechanical and structural engineers, and experts in algorithms and artificial intelligence. Virtually all Aingura staff were involved at some stage of development.
Difficulties and learning
Among the many difficulties encountered, the ones that caused the most problems were related to the installation and power supply, and to the adjustment of susceptance changes with temperature.
KPIs (business impact and metrics of the model)
45% reduction in the current cost of railroad crossing maintenance. 30% reduction in urgent level crossing replacements. More than 95% of the cracks detected in early, non-hazardous stages.
Funding
The project was funded with a combination of 20% own investment, 30% customer projects, and 50% public funding.
Collaborators, Partners
Multiple: Amurrio Ferrocarril y Equipos, Tecnalia, Dasel, UPM, BSC, Inalia, Titanium.
