An IoT-Integrated Deep Learning Framework for Automated Crack Detection in Marine Concrete Structures

Abstract

Marine concrete structures such as piers, harbors, and offshore platforms are susceptible to micro-cracking and degradation due to seawater corrosion, repetitive wave action, and temperature fluctuations. Undetected structural damage can compromise load-bearing capacity, accelerate structural failure, and increase maintenance costs. This study proposes an automated real-time crack detection and monitoring system integrating Internet of Things  sensors with Machine Learning techniques for marine environments. The methodology involves deploying multi-sensor nodes comprising vibration sensors, ultrasonic pulse velocity transducers, and high-resolution optical cameras at key stress points on marine concrete surfaces. Sensor data and surface images are continuously transmitted through a low-power wide-area network to a centralized cloud server. A Convolutional Neural Network is trained to analyze structural vibrations and image features for crack detection, classification, and localisation. Experimental results demonstrate a detection accuracy exceeding 96%, with real-time alerts and minimal false positives under variable weather conditions. The system captures early-stage micro-cracks before visible structural failure, supporting proactive maintenance and reducing inspection labor costs. Future work should focus on energy-harvesting technologies for autonomous sensor operation and expanded assessment of long-term structural fatigue under extreme oceanic events.