Armin Rastbood - School of Mining Engineering, College of Engineering, University of Tehran, Tehran, Iran.
Abbas Majdi - School of Mining Engineering, College of Engineering, University of Tehran, Tehran, Iran, .Post Code: ۱۴۳۹۹۵۷۱۳۱
Yaghoob Gholipour - School of Civil Engineering, College of Engineering, University of Tehran, Tehran, Iran. Post Code: ۱۴۳۹۹۵۷۱۳۱
Thirapong Pipatpongsa - Department of Urban Management, Kyoto University, Japan

The critical parameters in investigating the performance of designed support system of tunnels are the structural forces i.e. peak values of axial and shear forces, and moments. In this research, a complete database was firstly prepared using finite element method. Using finite element models, we modeled the segmental tunnel lining that was composed of 5+1 concrete segments in one ring. Then, an artificial neural network (ANN) model of multi-layer perceptron was developed to estimate the lining structural forces. To do this, the number of neurons and their arrangement were optimized based on the obtained minimum values from the root mean square error (RMSE). To prove the efficiency of the developed ANN model, we calculated the coefficient of efficiency (CE), determination coefficient (R2), variance account for (VAF), and RMSE values. The results demonstrated a promising precision and high efficiency of the presented ANN method for estimating the structural forces of tunnel lining composed of concrete segments instead of alternative costly and tedious solutions. Finally, the sensitivity analysis showed that among the input variables, the tunnel cover is the most influencing variable on the lining structural forces. However, other input variables, i.e. lateral earth pressure and key segment position were the second important variables affecting the induced stresses on tunnel lining.

Artificial neural network, lining, Multi-layer perceptron, segment, tunnel