AI-DRIVEN CRACK ANALYSIS FOR STRUCTURAL HEALTH MONITORING OF ENGINEERING CEMENT COMPOSITE: A LABORATORY STUDY

Engineering Cement Composite (ECC) is a type of high-ductility concrete due to the presence of fibers in its composition. The specific mechanical behavior of ECC shows that under tensile loads, multiple thin cracks are present instead of a single wide one. This feature is known as ECC’s crack control capability, which limits the crack width and enhances the durability of the concrete. Hence, ECC is particularly suitable for critical infrastructures such as bridges and tunnels, especially in earthquake-prone regions. In this experimental study, we focus on monitoring ECC samples subjected to controlled loads in a laboratory setting to analyze their cracking behavior using artificial intelligence (AI) techniques. For this purpose, we developed a semantic segmentation model using the Unet architecture to identify cracks within photos of the samples and, subsequently, to measure the crack dimensions using computer vision algorithms. In this investigation, we present an AI-based algorithm that measures the crack width at each point and enables the calculation of the applied load based on this identification with a minimum accuracy of 93.7%. Our experiments show that this technique can segment and measure the crack intensity by producing results that align with real quantifications. The proposed approach offers a reliable and cost-effective tool for structural health monitoring (SHM) of ECC structures as an alternative to the traditional monitoring trends. The proposed algorithms can be integrated with automated techniques, such as using drones during the data collection stage.