Application of artificial neural network in optimizing the drilling rate of penetration of western desert Egyptian wells

Abstract

This study presents an intensive artificial neural network ANN model designed using FORTRAN language to predict the drilling rate of penetration ROP with high accuracy. The developed ANN model gives a new correlation to be used in ROP prediction as a function of drilling parameters: bit size, mud circulation rate, drilling interval, wight on bit WOB and rotary speed RPM. The back-propagation algorithm is used as learning procedure to supervise neural nets. In the training process More than 3000 actual data points are collected to be used as inputs for the neural network box. The optimization process of ANN model shows that the optimum number of neurons that gives the lowest average error between the actual and predicted ROP is 20 neurons. The resultant average square error estimated for ANN model was less than 3% with R of 0.958. A comparison between the developed ANN-ROP model and the number of selected published ROP models are performed. The results show that ANN model able to estimate ROP with high accuracy compared with other models. The novel ANN model in this study gives empirical correlation that can be used to estimate the ROP without need for artificial intelligence software.