Since the first Tunnel Boring Machine (TBM) was built, the performance analysis and the development of accurate prediction models of the machines have been the ultimate goals of many research works. In the literature, there are many prediction models of PR. In most developed empirical methods, they only considered one or two intact rock properties or just rock mass classification systems which are not all deciding factors in PR of TBM. Therefore a method which could take into accounts all intact rock – discontinuity properties - ground condition- operational parameters of machine should be developed and used for TBM performance in rock masses. The Artificial Neural Networks (A) are facilitated all important parameters for predicting the PR of TBM. The fracture is one of the most important rock properties, which has a great influence on PR of TBM. In previous research works, the influence of joint properties such as spacing and orientation on PR has been investigated separately or just a regular joint set was modeled. Since the actual rock mass is geometrically complex, the Discrete Fracture Network (DFN) models which is based on stochastic representation of fracture systems, using the probability density function (PDF) of fracture parameters (e.g. orientation, length) formulated according to field mapping results and generated using Monte Carlo method; may use for creating conceptual model close to the reality. The Distinct Element Method (DEM) is a very useful method that simulates very complex geometrical models of fracture systems, such as DFN models. Using a number of DFN-DEM models for predicting the penetration rate of TBM, a probabilistic numerical modeling is established. In this study the obtained data from 10KM of excavated Zagros tunnel project in Iran were subjected to statistical analyses using MATLAB for prediction of PR using ANN modeling and geometric and mechanical parameters of this tunnel are used as the input data for probabilistic numerical approach. The results of predicting using ANN show that the developed ANN method is efficient for predicting the PR in Zagros tunnel. The ANN model for next 0.5 KM which is recently excavated is well compatible with the real calculated PR of TBM in the field with 79% confident level. Sensibility analysis results on the effect of the values of Thrust and Torque of TBM on the PR of a rock formation in this tunnel shows that the maximum PR in the mentioned formation occurred in the optimum limits of Thrust and Torque. Numerical modeling results also show that the PDF of positive incremental chipping area follows the normal distribution with the mean value of 46%. The Probabilistic Cumulative Distribution Function (PCDF) of results shows that the positive incremental chipping areas are greater than 20% and 33% when 95% and 80% are respectively considered.