Prediction of penetration rate (ROP) is important process in optimization of drilling due to its crucial role in lowering drilling operation costs. This process has complex nature due to too many interrelated factors that affected the rate of penetration, which make difficult predicting process. This paper shows a new technique of rate of penetration prediction by using artificial neural network technique. A three layers model composed of two hidden layers and output layer has built by using drilling parameters data extracted from mud logging and wire line log for Alhalfaya oil field. These drilling parameters includes mechanical (WOB, RPM), hydraulic (HIS), and travel transit time (DT). Five data set represented five formations gathered from five drilled wells were involved in modeling process.Approximatlly,85 % of these data were used for training the ANN models, and 15% to assess their accuracy and direction of stability. The results of the simulation showed good matching between the raw data and the predicted values of ROP by Artificial Neural Network (ANN) model. In addition, a good fitness was obtained in the estimation of drilling cost from ANN method when compared to the raw data.

This paper presents a detailed formulation of a rate of penetration (ROP) model, consideringmany drilling parameters and conditions for obtaining maximum drilling rate as well asminimizing the drilling cost.A regression analysis technique has been usedfor ROP modeling in Mishref formation.The datawere extracted from routinely available mud and wirelinelogs. These data includes weight onbit ,rotary speed,horse per square inch,and transit time.For ROP modeling, data of five wellsinHalfaya oil field in south Iraq were extracted.Statistical software called SPSS was used forimproving the modeling data and to perform linear and nonlinear multiple regression analysis.This improving approach included detection the outliers of modeling parameters, grouping themodeling data, moving average and finally applying the regression analysis.Results of modeling showed that the grouping of modeling data exhibited good convergencewith actual data and the overall model of oil field could produce good fitness with the actualdata in both cases of linear and nonlinear models.Also,a good estimation of drilling cost couldbe obtained when using this model.