Modeling maximum surface settlement due to EPBM tunneling by various soft computing techniques

There are various methods to predict the settlement caused by shallow tunneling, with each method having particular strengths and weaknesses. However, the most important weakness of common methods is the failure to consider all parameters contributing into the settlement. Nowadays, earth pressure balance machines (EPBMs) are commonly applied for tunneling into soft grounds. In this tunneling method, many parameters affect resultant surface settlement, it difficult to estimate the settlement by traditional methods. Soft computing, however, can be devised to cope with such engineering limitations. The aim of this study is to evaluate the ability of the soft computing methods of neuro-genetic system (NGS), adaptive neuro-fuzzy inference system (ANFIS), and gene expression programming (GEP) to predict maximum surface settlement (Smax) caused by tunneling in Shanghai Subway LRT Line 2 project. For this purpose, Smax is considered as a function of geometric, strength and operational factors, with the factors combined using different methods to reconstruct different models. The results showed that the models with operational factors outperformed other models. Among tested methods, ANFIS and NGS presented the best and the worst forecasts, respectively. With respect to the results of this research, it can be said that, despite the fact that GEP had lower accuracy in comparison to ANFIS, it represented the most suitable method to estimate Smax, because of providing useful mathematical equations.