A novel physics-based preconditioner for nodal integral method using JFNK for 2D Burgers equation. (April 2021)

• Record Type: Journal Article
• Authors: Kumar, Niteen; Singh, Suneet
• Is Part Of: Progress in nuclear energy. Volume 134(2021)
• Publication Date: 2021-04
• Subjects: Jacobian free Newton Krylov method -- Alternating direction implicit -- Physics based preconditioning -- Burgers' equation -- Nodal integral method; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 333.7924
• DOI: 10.1016/j.pnucene.2021.103668
• Languages: English
• ISSNs: 0149-1970

Abstract

Abstract: Nodal Integral Methods (NIM) are prevalent for solving neutron transport equations. Due to the success of this method for solving neutron transport problems, the method was used for solving fluid flow problems. In NIM, first PDEs are averaged over a node to form the ODEs and final schemes are developed using the analytical solution of these ODEs. Further modifications were made, leading to two versions called modified NIM (MNIM) and modified MNIM (M 2 NIM). These successive improvements were made to make it more robust for fluid flow problems, still it is limited to low Reynold number ( Re ) problems. The shortfall of nonlinear solvers compatible with NIM is the main reason for this limitation. Jacobian Free Newton Krylov (JFNK) in combination with NIM, is one way to improve the capabilities of NIM. However, the number of Krylov iterations can be quite large for high Re case. Here, a physics-based preconditioner is developed using linearized M 2 NIM. The developed algorithm is tested for the simulation of shock wave using 2D Burgers' equation with larger time steps and higher Re and found that the method is very accurate even with a coarse grid. The method is very effective in reducing computational time and Krylov iterations. Highlights: Jacobian Free Newton Krylov method in context of Nodal Integral method. Alternating Direction Implicit based preconditioner in two dimensions. Two dimensional propagating shock wave using Burgers' equation is solved. The algorithm shows a significant reduction in CPU runtime.