Assessment of the uniaxial compressive strength of intact rocks: an extended comparison between machine and advanced machine learning models

Rock strength is the most deterministic parameter for studying geological disasters in resource development and underground engineering construction. However, the experimental procedure for finding rock strength is arduous and lengthy. Therefore, this investigation introduces an optimal computationa...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Multiscale and Multidisciplinary Modeling, Experiments and Design Jg. 7; H. 4; S. 3301 - 3325
Hauptverfasser: Khatti, Jitendra, Grover, Kamaldeep Singh
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Cham Springer International Publishing 01.09.2024
Schlagworte:
Characterization and Evaluation of Materials
Engineering
Mathematical Applications in the Physical Sciences
Mechanical Engineering
Numerical and Computational Physics
Original Paper
Simulation
Solid Mechanics
Gaussian process regression
Uniaxial compressive strength
Multicollinearity impact
Intact rock
ISSN:2520-8160, 2520-8179
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Rock strength is the most deterministic parameter for studying geological disasters in resource development and underground engineering construction. However, the experimental procedure for finding rock strength is arduous and lengthy. Therefore, this investigation introduces an optimal computational model for predicting the rock uniaxial compressive strength (UCS) by comparing eight machine learning approaches. For developing the predictive models, the selection of the most significant independent variables is essential. Hence, this investigation reveals the most suitable independent variable by developing three cases of input variables, i.e., (i) area, density, wave velocity, and Young's modulus; (ii) mass, density, wave velocity, and Young's modulus; and (iii) density, wave velocity, and Young's modulus. Sixteen performance metrics have analyzed machine learning models' prediction capabilities and reported that the Gaussian process regression (GPR) model has predicted rock UCS with a correlation coefficient (R) of 0.9788, root mean square error (RMSE) of 14.0804 MPa, performance index (PI) of 1.8821, variance accounted for (VAF) of 95.79, index of scatter (IOS) of 0.1167, and index of agreement (IOA) of 0.9063, close to the ideal values and higher than those of other computational models, in case 1. However, the impact of weak multicollinearity has been observed in the performance of the support vector machine model than GPR and ensemble tree models. The score analysis, error characteristics curve, and Anderson–Darling test confirm the robustness of assessing the rock UCS.
ISSN:2520-8160
2520-8179
DOI:10.1007/s41939-024-00408-4