Real-time prediction of tensile and uniaxial compressive strength from artificial intelligence-based correlations

The study of the geomechanical parameters is necessary for field planning and development. Two of the most critical parameters used to describe the rock strength are the tensile (T s ) and the uniaxial compressive strength (UCS). Measuring these two parameters in the lab is time-consuming. Consequently, non-destructive methods have been developed to predict these parameters fast and reliable. Field drilling data can be reliable, continuous, and rapid technology in predicting UCS and T s . Herein, an artificial neural intelligence network (ANN) predicts T s and UCS from actual drilling data collected from two fields in the Middle East. The data include rate of penetration (ROP), weight on bit (WOB), torque (T), drilling fluid injection rate (Q), and the standpipe pressure (SPP). Several sensitivity analyses were conducted to optimize the models’ parameters and inputs, followed by extracting the weights and biases for developing ANN-based relations for T s and UCS. The results showed that the ANN was highly accurate during the training phase in predicting UCS with an AAPE of 0.28%, and T s with an AAPE of 0.28%. The developed correlation effectively predicted Ts and UCS for an average AAPE of 0.59 % during the testing phase and only 0.65 % for the validation data set for both parameters. This method provides a real-time effective tool for predicting the strength parameters in continuous, fast, and reliable measurements from the drilling field data.