High order compact fully-discrete scheme for hyperbolic conversation laws

TL;DR

This paper introduces a new high-order compact fully-discrete scheme for hyperbolic conservation laws, combining efficiency and accuracy through innovative interpolation and entropy strategies, outperforming traditional methods in computational cost and resolution.

Contribution

A novel one-step compact fully-discrete scheme utilizing stored by-products for interpolation, with enhanced robustness and multi-dimensional extension, offering significant efficiency improvements over existing methods.

Findings

Achieves comparable or better resolution with 1/10 to 1/13 of WENO+RK3 cost in 2D.

Increases computational cost by only 20-40% over FWENO in 1D.

Demonstrates efficiency and accuracy advantages over semi-discrete schemes.

Abstract

Based on the solution formula method, a series of one-step fully-discrete schemes, such as FWENO/Full-WENO has been proposed. Storing the by-products conservative variables at the half points (grid center) and using them as interpolation information to construct high-order schemes, we obtain a new class of one-step compact fully-discrete schemes. The new scheme can be associate with various non-oscillatory strategies. This paper takes state-of-the-art WENO-JS method as an example and proposes a family of compact fully-discrete WENO scheme. Detailed analysis is conducted on accuracy, errors, computational cost, efficiency and their connection with Hermite interpolation. Meanwhile, we design a new entropy flux linearization strategy for Euler equations to enhance its robustness, and also develop a multi-dimensional method for this compact fully-discrete framework. Due to the new scheme is one-step and utilizes stored by-products information for interpolation, it has a significant advantage in efficiency. For one-dimensional Euler equations, compared to the original FWENO, the computational cost only increases by 20-40%, while is approximately one-third of WENO+RK3. For two-dimensional case, a new special dimension-by-dimension strategy is applied. Although there is an additional computing cost, numerical experiments show that the new scheme only needs about 1/10 to 1/13 cost of that for WENO+RK3 when obtaining similar or even better resolution results, indicating that the new scheme is more efficient than semi-discrete schemes based on RK methods.