Application of ANN-PSO algorithm based on FDM numerical modelling for back analysis of EPB TBM tunneling parameters

Abstract

In this research, three- dimension finite difference method was applied to simulate the tunnel excavation by earth pressure balance (EPB) boring machine of Tehran Metro Line 7 tunnel in Iran. Sensitivity analysis and parameter identification was carried out by Morris Elementary Effect method to evaluate the relative sensitivity of model for each input parameter. Then, artificial neural network (ANN) was used as a meta model to predict model response for the dataset of selected sensitive parameters. For this purpose, 100 numerical simulation has carried that a database including 1500 data was compiled for using in ANN. Thereafter, PSO algorithm was employed as a parameter identification technique to find the optimized values of the parameters according to monitoring surface settlement. Also, after material identification, the same PSO-ANN algorithm was used to optimize operation parameter (face pressure). The results showed that ANN prediction and numerical simulation with initial set of parameters has more than 98% concordance with about 2.4% RMSE and 0.014 MAE for all dataset. The optimized parameters gained via back analysis enable the meta model to well predict the ground settlement in the about 15 seconds.

Keywords:

• EPB tunneling
• Morris elementary effect method
• back analysis
• PSO-ANN algorithm

Acknowledgments

The authors would like to extend appreciation to SCE for providing the necessary data that made the study possible.

Disclosure statement

No potential conflict of interest was reported by the authors.