Abstarct: Numerical simulation of tunneling with multiple uncertainties data is generally time consuming, which requires a lot of computational effort with a large number of concepts and factors. In order to make reasonable predictions in the short run time and with maintenance in machine tunneling, alternative models are very suitable for describing the dependence of the response with geotechnical and operational uncertainty of tunneling.In this research, after modeling the components and factors involved in the tunneling process using the finite difference method with FLAC 3D software and validating the model with monitoring data, sensitivity analysis on geotechnical and operational factors of tunnel construction with earth pressure balance machine is applied. The whole research process has been done on two case studies of Tehran Metro Line 7 and Tabriz Metro Line 2. To perform global sensitivity analysis, random samples were generated using Latin Hypercube method and numerical simulations were performed for these samples. Then, using elementary effect Morris method, sensitivity analysis was performed on the input factor and effective factors were selected. The results show that the mexta-model reduces the computational effort of the numerical simulation process and allows real-time prediction of the tunnel construction process in a pre-selected part of the tunnel project. The numerical results of the study confirm that the simulated ANN network can predict the surface settlement with reasonable reliability (more than 98%). Then, the combination of particle swarm optimization algorithm (PSO) for the first case study and genetic optimization algorithm (GA) for the second case study with neural network were used to perform back analysis and identify geotechnical and operational factors. Unlike the standard method for back analysis, in which each new case of factors must be modeled and this process will take several hours, the process of identification and back analysis with the mexta-model was performed in a few mimits. This process was used to predict the operating factors (face pressure and grouting pressure) to achieve the optimal surface settlement rate. Finally, the performance of genetic optimization and particle swarm algorithms for optimizing geotechnical factors was investigated and compared. Dualgorithm analysis showed that each dualgorithm with an error rate of about 6% can be used in the back analysis by mexta model.