An Investigation Into the Advancements of Edge-AI Capabilities for Structural Health Monitoring

Abstract

The aim of this study is to investigate the capabilities of edge-AI in the field of structural health monitoring, with a particular emphasis on detecting cracks in concrete bridges. Comprehensive literature suggests that edge-AI approaches have not been utilized in the structural health monitoring domain (SHM). This article proposed two novel frameworks: an edge-AI framework for the SHM domain and a cloud-edge adaptive intelligence for crack detection (CEAIC) to utilize edge-AI approaches for real-time scenarios. The framework incorporates a novel edge-AI framework, the crowd intelligence approach, and the CrANET framework to perform weakly supervised crack segmentation. Quantization approaches are used to transform the deep learning model into an optimized model to be compatible with edge devices. The edge-AI experiments for the cracks detection task are conducted using Kneron KL520 and Google Coral development board. A responsive website has been developed to demonstrate the real-world implications of the CEAIC framework. This study has the potential to provide a cost-effective and reliable solution for real-time monitoring and assessment of concrete bridge cracks, thereby improving the safety and longevity of bridges. The outcomes of this research endeavor will furnish us with invaluable insights regarding the feasibility and potential advantages of incorporating edge AI into the SHM domain.