Determination of the stress tensor of a triaxial strain cell in a three-layer model using the genetic algorithm and support vector machine

The accurate acquisition of the stress state plays an important role in engineering practice. The over-coring stress relief method calculates stress tensor by measuring strain with a triaxial strain cell. However, the current analytical solution of stress calculation is based on the two-layer model, which cannot calculate cases involving three-layer media. Based on the genetic algorithm (Ga) and support vector machine (SVM), a method for calculating the stress tensor of a triaxial strain cell in a three-layer model is proposed. Based on numerical simulation, Ga is used to search for the four equivalent elastic parameters of the three-layer model. Then the prediction model of Ga inversion parameters is established by employing SVM. Furthermore, using the equivalent elastic parameters and analytical solution of the double-layer model, the correction factor of the stress observation equation, that is, the K value of the three-layer model can be calculated. The proposed method is effective in accurately calculating the stress tensor of a triaxial strain cell of the three-layer model. While the optimal equivalent elastic parameters are adopted, the stress error is minimal and zero. Based on the proposed Ga-SVM model, the stress change of the Yingliangbao (YLB) hydropower station for a three-layer model is calculated. The stress calculation error caused by the difference between the coupling and hollow inclusion materials is corrected.