Ali Ranjbar

Predicting critical depths where drilling problems are more likely than other depths based on the experience of drilling previous wells in the field can be an important step in reducing drilling time and costs. In this research, by applying machine learning methods and using drilling parameters, the prediction of critical drilling depths has been investigated through the classification of drilling problems. 411 data sets obtained from the drilling of 17 wells in one of Iran's offshore fields were used to build machine learning models. To test the models, 30% of the data were used to prove their generalization. Using MATLAB R2021b software, multi-class classification models including support vector machines and ensemble classifiers were implemented. After optimization, the best validation accuracy belonged to the ensemble classifier with an accuracy of 89.2%. Also, the most accurate models in the testing phase were the ensemble classifier with 87.4% and the support vector machine with 82.1%. Models with appropriate accuracy can be used to predict critical drilling depths, and pay more attention while drilling new wells in the field under study.