Effect of textural and physical properties of the carbonate rocks on dynamic elastic modulus: application of statistical and intelligent methods

Abstract

This study investigated the effect of physical and textural characteristics of limestone and dolomite samples on dynamic elastic modulus (Ed). First, water absorption, density, carbonate percentage, and hardness were measured, and textural properties using thin sections were determined. Then, the effect of textural and physical characteristics on dynamic characteristics was determined by statistical and machine learning (ML) methods. The results showed that the total textures in these rocks are mudstone, wackestone, and packstone. Wackestone and mudstone showed a negative impact on dynamic characteristics, while packstone has a positive impact on Ed. Regression analysis showed that the impact of physical properties on dynamic properties is less than textural properties. Meanwhile, among textural properties, mudstone showed the least effect on Ed. Support vector regression, Levenberg-Marquardt artificial neural network, and Gaussian process regression were used to validate the statistical methods. The correlation coefficient, root mean square error (RMSE), spider diagram, Nash Sutcliffe efficiency (NSE), A10, and A20 indices were used to evaluate the models. Intelligent methods are more accurate than statistical methods in estimating Ed. The LMANN with a correlation coefficient of 99 % and an RMSE of 0.07 %, A10 = 1.00, A20 = 1.00, and NSE = 0.99 showed the highest accuracy among the used models.