Application of artificial neural network for determining elastic constants of a transversely isotropic rock from a single-orientation core. (December 2022)

• Record Type: Journal Article
• Authors: Lee, Yoonsung; Yim, Juhyi; Hong, Seungki; Min, Ki-Bok
• Is Part Of: International journal of rock mechanics and mining sciences. Volume 160(2022)
• Publication Date: 2022-12
• Subjects: Transversely isotropy -- Elastic constant -- Artificial neural network -- Strip load test -- Single-orientation core -- Deformability; Rock mechanics -- Periodicals; Soil mechanics -- Periodicals; Mining engineering -- Periodicals; Roches, Mécanique des -- Périodiques; Sols, Mécanique des -- Périodiques; Technique minière -- Périodiques; 624.151305
• DOI: 10.1016/j.ijrmms.2022.105277
• Languages: English
• ISSNs: 1365-1609

Abstract

Abstract: Numerous efforts have been made to determine the five independent elastic constants of transversely isotropic (TI) rocks. Recently, the novel strip load test method combined with the strain inversion method, offering the advantage of requiring only a single-orientation core, was presented (Yim J, Hong S, Lee Y, Min K–B. A novel method to determine five elastic constants of a transversely isotropic rock using a single-orientation core by strip load test and strain inversion. Int J Rock Mech Min Sci . 2022; 154:105115. 1 ). As a follow-up study, this paper suggests artificial neural networks (ANNs) to replace the strain inversion for determining five elastic constants of TI rocks with a strip load test method. The method comprises three main parts; the first was the strip load test experiment, the second part involves training ANNs using numerous datasets, and the final part is the application of trained ANNs to determine the elastic constants. The proposed method was numerically validated based on homogeneous and heterogeneous TI rocks. Experimental validation using Asan gneiss showed that the elastic constants determined from ANNs are in good agreement with those determined by using strain inversion and conventional method. The ANNs suggested in this study can significantly reduce the computing time required for strain inversion by numerical modelling and can be potentially used for other stress analyses of anisotropic rock. Highlights: Artificial neural networks were used to determine anisotropic elastic constants. Suggested method uses a single core with strip load test. Artificial neural networks can replace strain inversion method. Suggested method was validated by numerical and laboratory experiments.