Abstarct: The rate of areal cutting depends on more recognition of physical, chemical and mechanical characteristics which is needed for stone, specifications of chain saw machine and achieve parameters that are able control of prediction of the rate of areal cutting of carbonate stones. The prediction of the rate of areal cutting of carbonate stones is very important issue and complicated. After taking samples of achieve faces of marble mines of Dehbid and Shayan, the needed physical, chemical and mechanical specifications of 7 samples of decorative stone, analyzed and recorded. Regarding the same condition of cutting the faces The rate of superficial cutting with focus on achieve parameters that are able to control of the chain saw cutting machine in different conditions measurement and recorded. For linear and nonlinear regression models has been presented 6 models so that R2 for the best model which includes 0.96 for linear regression and 0.95 for nonlinear regression. In this model arrival parameters of mechanical and physical specifications of stone includes the uniaxial compressive strength, los angeles and schmite and output parameter of model is the rate of areal cutting. In this model arrival parameters of mechanical and physical specification of stone include the uniaxial compressive strength, Los Angeles and schmite and exit parameter the model of rate of areal cutting with los angeles and relationship of the opposition parameter of the rate of areal cutting another model of regression in both phase linear and nonlinear regression regarding R2 has very good condition. The arrival parameter of the model of physical and mechanical specifications of stone and specifications of active which is able to control of chain saw machine includes the speed of chain, the angel of saw and the speed of machine and arrival parameter model of the rate of areal cutting with the evaluation of presented relationship of the model confined direct line between the rate of areal cutting. A series of making model on the bare of three ways of artificial neural network includes group way of data management (GMDH) multi laxyers perceptron (MLP) artificial neural network (RBF) for the data 98 outputs for prediction of the production rate in three model making for the GMDH style in series of model 18 for the RBF style in series of 9 models and for MLP method in series 10 models baxse of conditions was model. All models on the baxse of functions of algorithm includes (R2) the error average square of (RMSE) and (VAF) have evaluated for GMDH style the amounts of functions of algorithm includes (R2) . Areal cutting rate with the speed of chain the speed of machine and los angeles and the opposite relationship of cutting rate with angle of saw uniaxial compressive strength the hardship schmite cutting rate with the of chain the speed of machine and los angeles and the opposite relationship of cutting rate with angle of saw, uniaxial compressive strength and hardship schmite. For GMDH method numbers of performance of the best models includes The error of the average square (RMSE), the education equals 0.252 the error of the average squares (RMSE) test equals 0.245 VAF education equals 90.3 and VAF test , equals 92.06 for RBF method, the amount of functions of the best model includes R^2 education equals 0.97, R^2 test equals 0.66 the error average square(RMSE)test equals 0.73 VAF education equals 96.44 and VAF test equals 65.94 for the style MLP the certain amounts of function of the best moddle of R^2education equals 0.81 R^2 test equals 0.22, VAF education equals 67.01 and VAF test equals 62.07 a comparison between three models of artificial neural network on the baxse of specification of function was shown that the most suitable way for the prediction of the rate of production for the cutting machine of chain saw GMDH and is the best model of this method which is superficial slice with the speed of chain, the speed of machine and los angeles and the relationship.