Structural health monitoring (SHM) is essential for ensuring the safety and durability of civil infrastructures amid rapid urbanization and aging structures. This paper presents a low-cost, real-time SHM system that integrates IoT-based sensor networks with deep learning-driven crack detection. Using the ESP32 microcontroller, MPU6050 for vibration/tilt sensing, and moisture sensors, data is transmitted wirelessly to the Thinger.io cloud platform for visualization and alerting. Complementing this, a fine-tuned VGG16 convolutional neural network (CNN) achieves pixel-level crack segmentation with F1 scores exceeding 99% on a 40,000-image dataset. The hybrid approach overcomes limitations of traditional manual inspections and prior IoT systems (e.g., Arduino + ThingSpeak), offering faster processing, real-time alerts via MQTT, and scalable multi-modal assessment. Experimental validation in lab and field settings confirms high accuracy, early damage detection, and practical applicability for bridges, buildings, and heritage structures.
BANTEY et al. (Wed,) studied this question.