Automatic Detection of Surface Thermal Cracks in Structural Concrete with Numerical Correlation Analysis

Abstract

Artificial Intelligence (AI) based automatic methods are emerging in the field of structural health monitoring. These methods are useful for the identification and detection of damage. The aim of this study is to develop an automated method for identifying cracks in fire affected concrete structures. This chapter presents a customized convolutional neural network (CNN) to identify the damage and quantify the extent of cracking. The five layer CNN deep learning technique is used. The outcome of the proposed deep learning technique is validated through optical observation. It is shown that crack quantification through optical observation provides good correlation with the calculated results of the proposed deep learning method. The numerical correlation analysis is performed with respect to the temperature variation and crack length. Analyses highlight that increased temperatures lead to more extensive damage in the concrete specimens, as would be expected. The work provides a basis for using AI to determine the extent to which a structure has been damaged and potentially estimate the thermal exposure based on observations.