A Prototype of a Lightweight Structural Health Monitoring System Based on Edge Computing

Area of Science:

• Engineering
• Computer Science
• Materials Science

Background:

• Traditional wired Bridge Structural Health Monitoring (BSHM) systems are costly and complex.
• Existing wireless BSHM systems face challenges in cost, synchronization, and reliability.
• Cloud-based extreme event detection methods are unsuitable for edge environments due to real-time and bandwidth limitations.

Purpose of the Study:

• To develop a lightweight wireless BSHM system utilizing edge computing for local data acquisition and real-time extreme event detection.
• To enhance the efficiency and reliability of BSHM for small-to-medium-sized bridges.

Main Methods:

• A wireless BSHM system with sensor nodes for acceleration data collection and an intelligent hub for data processing.
• Conversion of acceleration data into time-frequency images for training a MobileNetV2 model.
• Implementation of model quantization and Neural Processing Unit (NPU) acceleration for efficient on-device inference.

Main Results:

• The system demonstrated high acquisition accuracy, precise clock synchronization, and strong anti-interference capabilities in laboratory tests.
• The quantized model on the NPU achieved a 26x inference speedup, 35% power reduction, and <1% accuracy loss compared to an unquantized model on an ARM CPU.
• Successful verification of the system's performance on a laboratory steel bridge.

Conclusions:

• The proposed lightweight wireless BSHM system offers a cost-effective and reliable solution for structural health monitoring.
• Edge computing enables local intelligence and rapid response for extreme event detection in bridges.
• The system shows strong potential for real-world applications in bridge safety and maintenance.