Estimation of the Static Young's Modulus for Sandstone Reservoirs Using Support Vector Regression

The static Young's Modulus (E_static) is an important parameter affecting the design of different aspects related to oil and gas producing wells. It is significantly changing based on the type of the formation, and hence, an accurate method of identifying E_static is required. This study evaluates the performance of support vector regression (SVR) for prediction of the E_static. The SVR model was learned to evaluate the E_static from the well logs of the bulk formation density in addition to compressional and shear transit time. It was learned and tested on 592 training datasets of the inputs and their corresponding E_static, these datasets were obtained from a sandstone formation in Well-A. The learned SVR model was then validated on 38 data points from Well-B, the performance of the optimized SVR on predicting the E_static for the validation data was also compared with these of the early optimized artificial neural networks (ANN) and functional neural networks (FNN). As a result, all machine learning models showed high precision in predicting the E_static for the validation data where E_static was estimated with average absolute percentage errors of 3.80%, 2.54, and 2.03% and correlation coefficients of 0.991, 0.997, and 0.999 using the optimized ANN, FNN, and SVR models, respectively.