Monitoring and Analysis of Surface Cracks in Concrete Using Convolutional Neural Network

Reinforced concrete is widely used for construction of infrastructure projects like bridges, buildings, highways, dams, and power plants etc. Monitoring structural health of infrastructure projects is essential for their uninterrupted functioning. Generally physical inspections are carried out to detect defects in structures for further rectification. With recent advancements in computational algorithms, machine vision based inspection is emerging as an efficient technique for monitoring structural health of structures. Surface cracks in concrete structures are one of the important indicators of its health and enable the assessment of serviceability of the structures. The present work aims to address the issue by developing a novel system using machine vision and deep learning. A convolutional neural network-based methodology is developed to detect surface cracks in concrete cube images. The implementation of proposed methodology is demonstrated through monitoring of crack development in concrete cubes of size 150 × 150 × 150 mm with different compression strengths. The concrete cubes are subjected to compression loading in a standard compression testing machine. The analysis results in location, area, length, and number of cracks in synchronization with the applied compression load. The number of cracks, area of cracks, and length of cracks with respect to compression load are acquired using the developed system for different grades of concrete cubes. Results show that cracks detection and monitoring have been accurately performed with the developed system. The observations obtained from the crack load analysis can be very useful for improved understanding of concrete behaviour. The data acquired and observations can help professionals for improved structural health monitoring.