A Neural Network Expert System Developed for Predicting the Ground Settlement and the Damage Assessment of Adjacent Structures Due to Tunnel Excavation

Ground movements during tunnel excavation are varied in magnitude and trend depending on several factors including the tunnel geometry, ground conditions, excavation methods, and support measures, etc. There exist several empirical and semi-empirical formulas available for predicting ground movements. These conventional methods, however, do not simultaneously take into consideration all the factors, resulting in inaccurate predictions. In this study, an expert system called NESASS(Neural network Expert System for Adjacent Structure Safety analysis) was developed to predict the ground movements and damages occurring in adjacent structures due to tunnel excavation. In developing NESASS, an artificial neural network technique was incorporated to take into account both measured field data and tunnel construction information. Results of learning and inference of NESASS for the known site conditions and measured ground movements confirmed a high reliability with a very narrow margin of errors. NESASS was found to be compatibly reliable in predicting the ground movements due to tunnel excavation with respect to the field data that are excluded in the learning process. NESASS includes a scheme that allows the prediction of damage parameters and the possible crack patterns. Since evaluating the damage of buildings occurring throughout tunnel excavation steps calls for a 3-dimensional assessment, the conventional 2-dimensional theories were extended into 3-dimensional ones in this study. The building damages predicted from the proposed method were compared with those calculated from the 3-dimensional finite element analysis as well as with the measured field data. Results of the comparison showed that the reliability of the proposed scheme was acceptable.