The effect of drilling rig parameters on rate of penetration using artificial neural networks and multivariate linear regression analysis

The effect of drilling rig parameters on rate of penetration using artificial neural networks and multivariate linear regression analysis

At the present time, the oil industry has a special place in the world. The development of the upstream industries is a major concern for engineers and experts. By predicting drilling conditions in order to increase the drilling operationʹs efficiency, decrease cost of drilling and operation risks, in oil and gas well can reach favorable results. One of the criteria of drilling efficiency is the bitʹs rate of penetration (ROP), which is defined as the ratio of drilled depth over time. This research is carried out based on the information from GACHSARAN oil field. Initially data were collected and then effective parameters on the penetration rate were determined. Finally by choosing the most proper neural network, sensitivity analysis on the parameters was carried out. Furthermore, considering the coefficient of determination and mean square error as the main criteria, a comparison between these two mentioned methods has been done. Results show the used model of artificial neural network with seven input parameters and two hidden layers with ten and sixteen neurons, has markedly higher coefficient of determination with lower values of errors with respect to the linear regression. Therefore the predicted results via neural network show better agreements with actual values