Discretization Error Suppression and Consistency Verification via a 3D Continuous Field Model
This paper analyzes the structural impact of finite spatial grid partitioning (finite difference grid models) on numerical computation and verifies the mathematical consistency of the "3D Continuous Field Model," which processes space as a three-dimensional scalar field. The verification protocol implements the Method of Manufactured Solutions (MMS), an established verification methodology in computational science, alongside the exact Soloviev equilibrium solution from magnetohydrodynamics (MHD). Tracking the behavior of spatial gradient energy with respect to variations in grid resolution (Grid Size N) demonstrated that the conventional finite difference grid model exhibits an 82.51% increase in the total energy integral as resolution increases, indicating a divergence from the constant analytical solution. Conversely, the proposed 3D continuous field model, which incorporates 3D linear interpolation as a pre-processing step, stabilizes at an energy integral value 25.51% lower than that calculated by the finite difference grid model in the high-resolution limit, converging within the vicinity of the analytical solution. Furthermore, in grid convergence index (GCI) evaluations based on the ASME V&V 20 standard, the finite difference grid model exhibits error saturation, whereas the 3D continuous field model demonstrates a monotonic decreasing trend. This report presents the operational characteristics of both models based on pure mathematical numerical data, excluding uncertain factors inherent in physical experimentation.本論文では、空間の有限格子分割(有限差分格子モデル)が数値計算に与える構造的影響を分析し、空間を3次元スカラー場として処理する「3D連続場モデル」の数理的整合性を検証する。検証プロトコルとして、計算科学における検証手法である人為的構成解法(MMS)および磁気流体力学(MHD)のSoloviev平衡厳密解を適用した。格子解像度(Grid Size N)の変更に伴う空間勾配エネルギーの動向を追跡した結果、従来の有限差分格子モデルは解像度の上飾に伴い総エネルギー積分値が82.51%上昇し、解析解である定数からの乖離を示すことが確認された。一方、前処理として3次元線形補間を導入した提案3D連続場モデルは、高解像度極限においてエネルギー積分値が有限差分格子モデルの算出値に対し25.51%低い値で安定し、解析解の近傍に収束した。さらに、ASME V&V 20規格に基づく格子収束性評価(GCI)において、有限差分格子モデルが一定のエラー値で飽和するのに対し、3D連続場モデルは単調な減少傾向を示した。本稿では、物理的実験における不確定要因を排した純粋数学的数値データを基に、両モデルの演算特性を報告する。
Authors
- Akira Tsuyoshi (ORCID: https://orcid.org/0009-0009-4025-0841)
Institutions
- Oldham Council (GB)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-06-09
- DOI
- https://doi.org/10.5281/zenodo.20604431
- Primary Topic
- Advanced Numerical Methods in Computational Mathematics
- Type
- preprint