Artificial Neural Network Model for Predicting Wellbore Instability

Abstract Drilling activities have progressed to deep and ultra deep seas in recent times and with it comes more challenges. Due to the difficulty of directly obtaining important parameters like in-situ stress and fracture gradient, simple models have been evolved. This study is a novel attempt to make up for the gap inherent in such models namely that they neglect chemical and thermal effects, settling for only effective stress and a time-dependent analysis. The study applied the Neural Network (NN) technology to predict geomechanical parameters. Neural Network (NN) as a branch of Artificial intelligence (AI) possesses the ability of training available parameters to replace data that cannot be immediately or easily acquired. Data of a well drilled in the Niger Delta Region of Nigeria was used as the case study. A training set of input data was used to train the network and a validation set ensured a completely independent measure of network accuracy. A Neural Network model was developed in Neuroph Studio, Java neural network platform and the Netbeans IDE. The model has the advantage of being easy to use, open source, cross-platform and generally designed to save the cost associated with wellbore instability.