Edge Integration of Artificial Intelligence into Wireless Smart Sensor Platforms for Railroad Bridge Impact Detection

Area of Science:

• Structural Engineering
• Artificial Intelligence
• Wireless Sensor Networks

Background:

• Many US railroad bridges exceed 100 years old, posing risks due to insufficient vertical clearance.
• Over-height vehicle impacts can cause significant structural damage and service disruptions.
• Rapid detection and notification are critical for ensuring bridge safety and facilitating repairs.

Purpose of the Study:

• To develop an edge computing framework for real-time impact detection on railroad bridges.
• To implement machine learning predictions directly on wireless smart sensors.
• To overcome the limitations of traditional centralized data processing for time-sensitive structural health monitoring.

Main Methods:

• Developed a framework for edge implementation of machine learning (ML) on wireless smart sensors.
• Utilized the Xnode wireless smart sensor platform for AI model deployment.
• Tested the framework using sensor data from impact events on a railroad bridge.

Main Results:

• The proposed edge computing framework effectively processes sensor data directly on the sensor nodes.
• Eliminated the need for data transmission to a central location for impact detection.
• Demonstrated the framework's efficacy for time-sensitive structural health monitoring applications.

Conclusions:

• Edge implementation of ML on wireless smart sensors is a viable solution for rapid impact detection.
• This approach enhances the efficiency and timeliness of structural health monitoring for critical infrastructure.
• The framework offers a cost-effective and scalable method for ensuring railroad bridge safety.