Prediction of Mode-I rock fracture toughness using support vector regression with metaheuristic optimization algorithms. (1st April 2022)

• Record Type: Journal Article
• Authors: Mahmoodzadeh, Arsalan; Nejati, Hamid Reza; Mohammadi, Mokhtar; Hashim Ibrahim, Hawkar; Khishe, Mohammad; Rashidi, Shima; Farid Hama Ali, Hunar
• Is Part Of: Engineering fracture mechanics. Volume 264(2022)
• Publication Date: 2022-04-01
• Subjects: Mode-I rock fracture toughness -- Machine learning -- Metaheuristic optimization -- cracked Chevron notched Brazilian disc test -- Mutual information test; Fracture mechanics -- Periodicals; Rupture, Mécanique de la -- Périodiques; Fracture mechanics; Periodicals; 620.112605
• DOI: 10.1016/j.engfracmech.2022.108334
• Languages: English
• ISSNs: 0013-7944

Abstract

Highlights: Developing hybrid algorithms to predict Mode-I rock fracture toughness. Using 250 datasets in the models. Applying both holdout and 5-fold cross-validation methods to validate the results. Sensitivity analysis of the input parameters using mutual information test. Recognition of the most robust model to predict the Mode-I rock fracture toughness. Abstract: In this work, the support vector regression method is combined with six metaheuristic optimization models of particle swarm optimization, grey wolf optimization, multiverse optimization, moth flame optimization, sine cosine algorithm, and social spider optimization to predict Mode-I rock fracture toughness. In addition, four other models of random regression forest, extra regression tree, decision regression tree, and fully-connected neural network that previously were used to predict the Mode-I rock fracture toughness by other researchers, was applied by this study. 250 datasets, including six input parameters and one output parameter (Mode-I rock fracture toughness) were utilized in the models obtained through the cracked Chevron notched Brazilian disc testing specimens suggested by the ISRM in the laboratory. Finally, the hybrid model of support vector regression-particle swarm optimization produced the most accurate results and it was recommended to predict the Mode-I rock fracture toughness. Also, the mutual information test was used to examine the impact of each input parameter on the Mode-I rock fracture toughness. Finally, the uniaxial tensile strength was identified as the most effective parameter on the Mode-I rock fracture toughness.