RESEARCH

Simulation that couples multiple physical models is called as multi-physics simulation. Nuclear reactors involve many distinct physical models: neutron transport, thermal-hydraulics, fuel performance, etc. With increased computational power, multi-physics simulation of nuclear reactor has attracted much attention in recent years. However several instability, inaccuracy can be caused by coupling multiple physical models. Also, additional computational burden is occurred from data communication between independent codes. SNURPL established neutron transport & thermal-hydraulics coupled systems such as nTRACER/MATRA or nTRACER/ESCOT which is coupled with enhanced efficiency and stability.

The most simple and straightforward way to resolve the nonlinear coupling of individual physical models is Picard fixed point iteration which alternates each calculation. The traditional fixed point iteration (Jacobi approach) solves each calculation sequentially. For the enhancement of parallel computing efficiency, the tandem solution approach (Gauss-Seidel approach) in which both calculations are solved simultaneously also has been employed. These two approaches are simple to implement since they don’t require any modification of the solver. But these methods don’t provide either unconditional stability and optimized convergence.

To enhance both stability and performance, more sophisticated coupling methods are developed. The Jacobian Free Newton Krylov method (JFNK), which employs the newton method to minimize the residual to yield faster and more robust convergence, is one of the popular coupling method. The Anderson acceleration, which minimizes the square norm of the residual vector, is also used widely to resolve the nonlinear coupling.

Though each solvers solves the same system, the discretization grids of each solver can be different. In this case, appropriate mapping method for physical variables is required. Mapping methods should guarantee the accuracy, consistency of target physical variables.