Estimation of Surface Fault Displacement by High Performance Computing

Abstract

Reliable estimation of surface fault displacements is crucial to the safety of nuclear power plant facilities. It is necessary to develop a numerical method for the estimation. In the study, we develop a finite element method in which the following two functions are implemented: (1) a symplectic time integration of an explicit scheme to properly conserve the energy of the system; and (2) rigorously formulated joint elements of high order. The finite element method is enhanced with parallel computing capability. We apply the developed method to solve simple three-dimensional models of faults embedded in a rock mass. It includes a comparison of results from quasi-static and dynamic simulations and investigation of the sensitivity of results to the shear stiffness on faults. In the study, we propose capacity computing with a quasi-static simulation for uncertainty quantification.

Predictive Simulation for Surface Fault Occurrence Using High-Performance Computing

“…This is extremely important for the estimation of ∆ c . Sawada et al [13] demonstrated that the calculated ∆ c and δ from quasi-static simulations were close to those calculated using dynamic simulations with damping and proposed the use of quasi-static simulations for capacity computing. Therefore, quasi-static simulations were adopted in this study.…”

“…We implemented joint elements using a nonlinear spring-type constitutive equation to represent fault movement [12,13]. The constitutive equation for the shear movement of the joint element is defined as follows (see Figure 4):…”

“…We assumed a cohesion of 0.025 MPa and friction angle of 25 • on the fault planes by reference to the experimental value of the crushed zone in NPP sites [14]. The ratio between the initial and final shear stiffness κ d /κ 0 was assumed to be 0.01 in reference to the analyses conducted using a simple ground/fault model and a real earthquake ground model [13,14]. The critical slip u cr was 0.1 m in reference to the value used in the dynamic rupture simulation of the fault [4].…”

“…We developed an FEM strengthened with the following two functions: (1) a symplectic time integration explicit scheme to conserve the energy of the system and (2) rigorously formulated high-order joint elements. The FEM was also enhanced with parallel computing capabilities [12,13]. Another use of HPC is to perform multi-case analyses with different input conditions in parallel.…”