A-posteriori error estimates for systems of hyperbolic conservation laws

Jan Giesselmann,Aleksey Sikstel
2023-05-02
Abstract:We provide rigorous and computable a-posteriori error estimates for first order finite-volume approximations of nonlinear systems of hyperbolic conservation laws in one spatial dimension. Our estimators rely on recent stability results by Bressan, Chiri and Shen and a novel method to compute negative order norms of residuals. Numerical experiments show that the error estimator converges with the rate predicted by a-priori error estimates.
Numerical Analysis
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to provide strict a posteriori error estimates for the initial value problem of one - dimensional nonlinear hyperbolic conservation law systems. Specifically, the author focuses on how to provide reliable a posteriori error estimates when using the first - order finite - volume method to approximately solve these systems. These error estimates are crucial for the verification of numerical methods and the generation of adaptive meshes. ### Background and Motivation When dealing with hyperbolic conservation law systems, since the solution may contain discontinuities (such as shocks), traditional a priori error estimates are often not accurate enough. Therefore, a posteriori error estimates become an important tool, which can dynamically evaluate errors during the calculation process and guide the adaptive adjustment of the mesh. ### Main Contributions 1. **Strict A Posteriori Error Estimates**: The author provides strict a posteriori error estimates, which are based on the stability results recently proposed by Bressan, Chiri and Shen et al. 2. **Calculation Method of Negative - Order Norms**: A new method is introduced to calculate the negative - order norms of residuals, which is a key part of a posteriori error estimates. 3. **Numerical Experiments**: The effectiveness of the proposed method is verified through numerical experiments, and the results show that the convergence rate of the error estimator is consistent with that predicted by the a priori error estimates. ### Method Overview - **Localization of Weak Residuals**: Decompose the global weak residual into local weak residuals on each space - time cell and calculate its \( W^{-1,1} \) norm. - **Projection Operator**: Introduce a projection operator to map Lipschitz continuous test functions to a finite - dimensional space, so that the upper bound of local weak residuals can be effectively calculated. - **Estimation of Total Variation and Oscillation**: Combine stability analysis to calculate the total variation and oscillation of the numerical solution, which are crucial for error estimation. ### Mathematical Formulas - **Definition of Weak Residuals**: \[ B_j^n[\hat{u}] : W^{1,\infty}(K_j^n) \to \mathbb{R}^m, \] \[ \phi \mapsto \int_{x_{j-\frac{1}{2}}}^{x_{j+\frac{1}{2}}} \hat{u}_j^n \phi(t_n, x) \, dx - \int_{x_{j-\frac{1}{2}}}^{x_{j+\frac{1}{2}}} \hat{u}_j^{n+1} \phi(t_{n+1}, x) \, dx + \int_{K_j^n} \hat{u}_j^n (\phi_t + f(\hat{u}_j^n) \phi_x) \, dx \, dt \] \[ + \int_{t_n}^{t_{n+1}} \hat{f}(\hat{u}_j^{n - 1}, \hat{u}_j^n) \phi(t, x_{j-\frac{1}{2}}) \, dt - \int_{t_n}^{t_{n+1}} \hat{f}(\hat{u}_j^n, \hat{u}_j^{n+1}) \phi(t, x_{j+\frac{1}{2}}) \, dt. \] - **Properties of Projection Operators**: \[ \|P_j^n \phi\|_{W^{1,\infty}(K_j^n)} \leq \max\{3, \sqrt{8 + 8(c_j^n)^2}\} \|\phi\|_{W^{1,\infty}(K_j^n)}, \] where \( c_j^n=\frac{\Delta t_{n+\frac{1}{2}}}{\Delta x_j} \). - **Final Form of Error Estimates**: \[ \epsilon := \begin{cases} 0 & \text{if } \hat{u} \equiv \text{constant}, \\ C \max\{\beta, \eta\} \sup_{s \in [0, t]} \text{TV}