ANN Model for Predicting Mud Loss Rate from Unconfined Compressive Strength and Drilling Data

Abstract

Lost circulation is a major issue that increases the cost of petroleum exploration operations. During the well planning period, consideration of the degree of severity of mud loss may lead to significant technical and financial benefits. This will assist in the prevention of losses by putting preventative measures in place before running into lost circulation region. This study aimed to predict the amount of mud loss rate (MLR) by using new models with artificial neural networks (ANNs). This model was built in order to obtain a knowledge of the relationship between the amount of loss and the drilling parameters that can be controlled, such as (the rate of penetration (ROP), flow rate (FLW), standpipe pressure (SPP), weight on bit (WOB), nozzle area (TFA), rotation per minute (RPM), and torque (TRQ)), the drilling fluid properties and geomechanical properties like unconfined compressive strength (UCS). Gaining information about UCS along the wellbore is essential for dealing with drilling problems like lost circulation. The new model was developed using a dataset of 209 loss events that were collected from 21 oil wells in the Rumaila oil field’s Dammam and Hartha formations that encountered loss circulation events. Apart from other controllable drilling parameters, it was demonstrated that the rate of losses was also sensitive to UCS values. The amount of mud losses rate constantly rises with increasing UCS. The suggested artificial neural networks (ANN) model was employed to forecast the rate of losses for 21 wells. A comparison plot depicting the actual rate of lost circulation versus the predicted rate was generated as a function of depth. The results indicate that the new model is able to precisely forecast the lost circulation function of controllable drilling variables, drilling mud properties, and UCS with a correlation coefficient of 0.995.