Intelligent identification of soil and operation parameters in mechanised tunnelling by a hybrid model of artificial neural network-genetic algorithm (case study: Tabriz Metro Line 2)

ABSTRACT In this article, the ability of the artificial neural network-genetic algorithm (ANN-GA) to perform back analysis and predict maximum surface settlement in mechanised tunnelling is investigated. The required data of the ANN meta-model was generated using 150 three-dimensional finite-difference simulations. The global sensitivity analysis was performed on 19 parameters, including 17 geotechnical parameters of soil layers and 2 operational parameters of face pressure and grouting pressure. The predicted results using ANN were in good agreement with the numerical simulations so that R = 99% and rRMSE = 1.5% are obtained. Then, back analysis was performed using the ANN-GA hybrid algorithm and the geotechnical data of the monitoring point were updated using the maximum surface settlement monitored at this point. Also, for the geotechnical parameters considered in the design phase, using the same algorithm, the number of operational parameters required for optimal settlement was predicted.