Determination of safe mud weight window based on well logging data using artificial intelligence. Issue 4 (4th July 2019)

• Record Type: Journal Article
• Authors: Zahiri, Javad; Abdideh, Mohammad; Ghaleh Golab, Elias
• Is Part Of: Geosystem engineering. Volume 22:Issue 4(2019)
• Page Start: 193
• Page End: 205
• Publication Date: 2019-07-04
• Subjects: Mechanical earth model -- safe mud weight window -- well logging data -- machine learning -- support vector machine; Mining engineering -- Periodicals; Petroleum engineering -- Periodicals; Gas engineering -- Periodicals; Geology, Economic -- Periodicals; 620
• DOI: 10.1080/12269328.2018.1504697
• Languages: English
• ISSNs: 1226-9328

Abstract

ABSTRACT: Identification of different stresses applied to the environment surrounding wellbore via different processes, and combining these data with mechanical parameters of common formations in hydrocarbon reservoirs comprise a key for addressing a wide range of costly problems and issues in the oil industry. In the present research, first, an attempt was made to construct mechanical earth model based on well logging data, elastic moduli of rock, and appropriate failure criteria for the final purpose of calculating and determining safe mud weight window (SMWW). Finally, appropriate artificial intelligence and machine-learning algorithms were used to establish a relationship between well logging data and SMWW, which could be used to calculate and predict SMWW without using associated relationships with the mechanical earth model. This might end up with a decreased number of required parameters for calculating SMWW, including uniaxial compressive strength. In the present research, the learning process was conducted using datasets from three wells, two of which provided training data, with the other one used as testing data. The prepared model was finally used to predict corresponding pressures to SMWW and baseline pressures for hydraulic fracturing operation. The model gave a coefficient of determination of 0.93 when applied to the testing data using support vector regression algorithm with radial basis function kernel, indicating large capabilities of this algorithm in predicting non-foreseen data.