The Establishment of A Constitutive Model of Sand Under Monotonic Loading by Adopting the Support Vector Machine (SVM)

Autor: Raopin Liao, Yan Zhang, Junbo Qiu, Yong-gang Zhang
Rok vydání: 2021
Předmět:
Zdroj: Arabian Journal for Science and Engineering. 47:4421-4435
ISSN: 2191-4281
2193-567X
Popis: The numerical simulation software is always used to predict the foundation’s response under the function of soil liquefaction, in which the built-in dynamic constitutive model is the key to ensuring the effectiveness; however, sand's deformation mechanism is very complicated, especially for complex stress path, the classical method using the mathematical model to simulate the relationship between the elastoplastic modulus and measurable macroscopic parameters cannot fully simulate the response. As artificial technology develops, because of its strong ability to learn the law behind the complicated phenomenon based on the massive data, it is worthy of applying this technology to the constitutive model. In this paper, the standard Fujian sand was chosen to conduct the monotonic consolidation drained triaxial test under different initial densities and initial confining stress. By recording the stress and strain state data, the current deformation modulus and current physical and mechanical state were calculated, the apparent relationship between dynamic shear modulus $$G_{{\text{s}}}$$ , dynamic volumetric modulus $$K_{{\text{s}}}$$ and shear strain, the volumetric strain was revealed. The collected 333 data were divided into a training set and test set by the optimal ratio of 9:1, the value of $$C$$ and $$g$$ of SVM was determined by using three different optimisation algorithms grid search, particle swarm optimisation (PSO) and grey wolf optimisation (GWO), the corresponding performance of trained models was compared, and the cause of the difference was also analysed. The results showed that the PSO–SVM and GWO–SVM could predict the evolution of the deformation modulus better. The error is caused by insufficient data for the condition of high confining stress; the GWO–SVM can simulate the response of sand under monotonic loading best. In a word, the application of SVM to the establishment of the constitutive model is capable and prospecting.
Databáze: OpenAIRE
Externí odkaz: