Prediction of Differential Pipe Sticking by Using Probabilistic Artificial Neural Network in Offshore Persian Gulf Oil Fields

Differential pipe sticking is one of the usual and hazardous problems during drilling operation that leads to increasing the total cost. Nowadays minimizing the risk of stuck occurrence is one of the priorities and main goals in petroleum industry. In the past، statistical methods were applied to investigate differential pipe sticking، but these methods cannot remarkably predict the non-linear behaviors. Artificial neural network is a novel method for solving engineering problems. This method is capable of considering the effective parameters at the same time and has the ability of direct generalization and learning from the field data (due to the errors and uncertainties). In this paper، the data from 63 wells of the offshore Persian Gulf oil fields were applied and by using a probabilistic neural network، a predictive model has been developed. High accuracy of this model in predicting differential pipe sticking allows it to be applied in well planning as well as real time drilling operations. Analyzing the result of neural network، associated with engineering viewpoint leads to preventing differential pipe sticking by optimizing the effective parameters.