The implementation of a broad class of boundary conditions for non-linear hyperbolic systems

TL;DR

This paper introduces new boundary condition implementation methods for non-linear hyperbolic systems, improving stability and convergence in numerical simulations, especially for complex algebraic conditions.

Contribution

It extends existing boundary condition techniques by combining characteristic equations with extrapolation and introduces projective time-stepping algorithms for better convergence.

Findings

Enhanced boundary condition methods improve simulation accuracy.

Proposed algorithms reduce drift-off errors in algebraic conditions.

Test cases demonstrate the effectiveness of the new methods.

Abstract

We propose methods that augment existing numerical schemes for the simulation of hyperbolic balance laws with Dirichlet boundary conditions to allow for the simulation of a broad class of differential algebraic conditions. Our approach is similar to that of Thompson (1987), where the boundary values were simulated by combining characteristic equations with the time derivative of the algebraic conditions, but differs in two important regards. Firstly, when the boundary is a characteristic of one of the fields Thompson's method can fail to produce reasonable values. We propose a method of combining the characteristic equations with extrapolation which ensures convergence. Secondly, the application of algebraic conditions can suffer from $O(1)$ drift-off error, and we discuss projective time-stepping algorithms designed to converge for this type of system. Test problems for the shallow water equations are presented to demonstrate the result of simulating with and without the modifications discussed, illustrating their necessity for certain problems.