Uso de técnicas de inteligência artificial na estimativa de deslocamentos e danos induzidos por escavações subterrâneas

Abstract

Estimating displacements induced by tunneling and the possible damage it may cause structures adjacent to the tunnel is one of the challenges the project engineer team faces in the tunnel design phase. The main methods used to estimate displacements are empirical, analytical and numerical simulations. Each of the aforementioned methods has advantages and disadvantages. For example, the main advantage of empirical methods is they are easy to use and their main disadvantage is they do not consider mass deformability and resistance parameters. The main advantage of analytical methods is that they incorporate the 3D effect (displacements that may occur before tunneling) and the ground deformability and resistance parameters. The main disadvantage of this method is that the 3D effect and the ground deformability and resistance parameters are in a single parameter called gap. With numerical simulations, the main advantage is the possibility to simulate the various construction phases for the work, and the disadvantage is the use of a realistic constitutive model while being in possession of a testing program that can provide the necessary parameters for using the model. In this study, we use two Artificial Intelligence (AI) techniques, one called Artificial Neural Networks (ANNs) which were used to estimate the displacements induced by tunneling, and the other called Fuzzy Logic, which was used to estimate the damage in structures adjacent to the tunnel. These techniques were used in a real situation that occurred during the excavations for the Federal District’s subway. The results reveal that the ANNs technique is extremely efficient when we have a set of data for training and that encompass the entire universe of the problem. Estimated damage using the Fuzzy Logic Technique of structures adjacent to the Federal District subway excavations were in keeping with what happened in the field.