Abstarct: Predicting penetration rates in hard rock TBM's Tunnelling projects has a real impact on the estimating project's cost and Time. Many models have been introduced to predict the penetration rate of TBM each one of them has their own advantages and disadvantages .Reviews shows that rating models are the best models to predict the penetration rate .past reaserchese have shown that correct parameters of rock mass classification alone is not a good index to predict the penetration rate .so introduction of a new model that includes machine's parameters as well as the Geomechanic's parameters is necessary .in this study we tried to use neural network toolbox in MATLAB software to choose TBM's parameters affecting the penetration rate with maximum effect and minimum overlap. these parameters are the rotational speed of Cutter head, trust force per cutter, RMRadjusted, tensile strength. then baxsed on expert's opinion and Fuzzy Delphi Analytic Hierarchy the parameters have been wighted so that the new model is more comprehensive and less data dependent. then given to the relative importance of parameters used a maximum rate of 100 was considered. also in The new classification all four parameters were classified separately and rated nonlinearly, in the next step for validation of the method, the new Rating system was applied on data obtained from18 zone of Zagros tunnel with different geomechanical characteristics. after that penetration rate with a correlation coefficient of R ² = 0.9191 obtained using regression analysis of the mathematical equation y = 0.0621x - 0.8098 between the rates obtained from the new model and the actual penetration rate. then the values predicted by the new model, the values predicted by the other models and the actual peneteration rate values were analyzed and compared. values predicted by the new model are closest to the actual penetration rate compared to other models.