Predicting the Risk of Lost Circulation using Support Vector Machine Model
Mud loss is one of the most common problems in drilling operations. Millions of dollars are spent every year to control or mitigate this issue. The aim of this work is to use mud weight (MW), equivalent circulation density (ECD), yield point (YP), flow rate (Q), plastic viscosity (PV), nozzles total flow area (TFA), weight on bit (WOB), and revolutions per minute (RPM) to predict and assess the risk of lost circulation (partial or complete loss). Data of more than 3000 wells collected from multiple sources. Support vector machine (SVM) with six Kernel functions (linear, quadratic, cubic, fine Gaussian, medium Gaussian, and coarse Gaussian) were tested and the highest accuracy model was selected. 5-fold cross-validation was conducted to ensure a sufficient representation of all data points in the training process. The fine Gaussian Kernel showed the best accuracy (99.3%) among the other models and it was selected to train the final model. The model created in this study can be utilized to assess the risk of lost circulation and help the drilling personnel to mitigate/reduce the risk, or prepare the required remedies to stop the risk. Machine learning and other data-driven methods have a great potential to revolutionize the oil and gas and save time and money.
H. H. Alkinani et al., "Predicting the Risk of Lost Circulation using Support Vector Machine Model," 54th U.S. Rock Mechanics/Geomechanics Symposium, American Rock Mechanics Association (ARMA), Jul 2020.
54th U.S. Rock Mechanics/Geomechanics Symposium (2020: Jun. 28-Jul. 1, Virtual)
Geosciences and Geological and Petroleum Engineering
Article - Conference proceedings
© 2020 American Rock Mechanics Association (ARMA), All rights reserved.
01 Jul 2020