Numerical simulation of ultrasonic tests for generation of tomograms of concrete elements using artificial neural networks

Abstract

Concrete structures are constantly deteriorating naturally. In this sense, the use of Non-Destructive Testing (NDT) constitutes a promising strategy for monitoring and evaluating the state of these structures without damaging their appearance or performance. Ultrasound is one of the most commonly used NDT in the field of civil engineering and the structural evaluation through this method is advantageous due to its low cost, simplicity of execution while not actually causing any type of damage to the structure. The development of research on this topic meets the growing concern with the evaluation of durability conditions and the rehabilitation of structures. The use of numerical tools to simulate the propagation of mechanical waves allows the evaluation of behavior hypotheses comparing numerical and experimental results. Confirming the evaluated hypotheses, numerical simulation allows predicting behavior and studying phenomena that are difficult to analyze experimentally. Therefore, this research aims to propose a tomographic reconstruction method coupling the concepts used in amplitude tomography, travel-time tomography and backprojection techniques using inverse analysis and artificial neural networks. Amplitude tomography coupled with backprojection concepts will be implemented at software TUSom, which is a tool for travel time tomography developed at FAPESP research 2011/22179-7. Computational inverse techniques will be applied to the tomographic problem using artificial neural networks. Development of a reliable tool and methodology using numerical simulation of the ultrasonic wave propagation (UWP) to be applied to the study of the concrete heterogeneity, detection of voids and damages is expected as a result of this research. (AU)