Uniaxial compressive strength(UCS)is a significant mechanical measure in rock engineering,key for classifying rock masses,conducting stability assessments,and guiding design processes.Over the past five decades,numero...Uniaxial compressive strength(UCS)is a significant mechanical measure in rock engineering,key for classifying rock masses,conducting stability assessments,and guiding design processes.Over the past five decades,numerous techniques and devices have emerged for UCS measurement.This paper presents a comprehensive literature review of existing methodologies and advancements in UCS testing,examining the theoretical foundations,testing equipment,data processing techniques,and criteria for selecting appropriate UCS testing methods.Additionally,the study highlights a shift toward automated,precise,and computational approaches with multiple inputs(i.e.multiple regression,machine learning,ML)for UCS prediction.Approximately 221 articles published by various researchers between 2000 and 2024 related to ML were reviewed,focusing on the application of ML models,including artificial neural networks(ANNs),adaptive-network-based fuzzy inference system(ANFIS),random forest(RF),support vector machine(SVM),and extreme gradient boosting(XGBoost),in UCS prediction.The review also observed the growing use of hybrid models integrating ML with optimization techniques,significantly improving UCS estimation.Numerous empirical correlations,both direct and indirect,have been established in the past several years based on the physical properties of rocks.However,utilizing these proposed equations to reliably estimate UCS remains challenging due to the variability in lithology,rock origin,and other factors.This study systematically presents the established correlation expressions,considering their lithology,the number of samples used to establish the expressions,the coefficient of determination(R2),and the locations.This allows geologists and engineers to easily apply these established expressions in situations where direct estimation is not possible,enabling them to approximate UCS results.展开更多
文摘Uniaxial compressive strength(UCS)is a significant mechanical measure in rock engineering,key for classifying rock masses,conducting stability assessments,and guiding design processes.Over the past five decades,numerous techniques and devices have emerged for UCS measurement.This paper presents a comprehensive literature review of existing methodologies and advancements in UCS testing,examining the theoretical foundations,testing equipment,data processing techniques,and criteria for selecting appropriate UCS testing methods.Additionally,the study highlights a shift toward automated,precise,and computational approaches with multiple inputs(i.e.multiple regression,machine learning,ML)for UCS prediction.Approximately 221 articles published by various researchers between 2000 and 2024 related to ML were reviewed,focusing on the application of ML models,including artificial neural networks(ANNs),adaptive-network-based fuzzy inference system(ANFIS),random forest(RF),support vector machine(SVM),and extreme gradient boosting(XGBoost),in UCS prediction.The review also observed the growing use of hybrid models integrating ML with optimization techniques,significantly improving UCS estimation.Numerous empirical correlations,both direct and indirect,have been established in the past several years based on the physical properties of rocks.However,utilizing these proposed equations to reliably estimate UCS remains challenging due to the variability in lithology,rock origin,and other factors.This study systematically presents the established correlation expressions,considering their lithology,the number of samples used to establish the expressions,the coefficient of determination(R2),and the locations.This allows geologists and engineers to easily apply these established expressions in situations where direct estimation is not possible,enabling them to approximate UCS results.
