A hybrid data-driven solution to facilitate safe mud window prediction

Abstract

AbstractSafe mud window (SMW) defines the allowable limits of the mud weights that can be used while drilling O&G wells. Controlling the mud weight within the SMW limits would help avoid many serious problems such as wellbore instability issues, loss of circulation, etc. SMW can be defined by the minimum mud weight below which shear failure (breakout) may occur (MWBO) and the maximum mud weight above which tensile failure (breakdown) may occur (MWBD). These limits can be determined from the geomechanical analysis of downhole formations. However, such analysis is not always accessible for most drilled wells. Therefore, in this study, a new approach is introduced to develop a new data-driven model to estimate the safe mud weight range in no time and without additional cost. New models were developed using an artificial neural network (ANN) to estimate both MWBO and MWBD directly from the logging data that are usually available for most wells. The ANN-based models were trained using actual data from a Middle Eastern field before being tested by an unseen dataset. The models achieved high accuracy exceeding 92% upon comparing the predicted and observed output values. Additionally, new equations were established based on the optimized ANN models’ weights and biases whereby both MWBO and MWBD can be calculated without the need for any complicated codes. Finally, another dataset from the same field was then used to validate the new equations and the results demonstrated the high robustness of the new equations to estimate MWBO and MWBD with a low mean absolute percentage error of 0.60% at maximum. So, unlike the costly conventional approaches, the newly developed equations would facilitate determining the SMW limits in a timely and economically effective way, with high accuracy whenever the logging data are available.