Parameter identification of fluid field based on CFD reduced-order model and 3D-Var data assimilation

Abstract

Data assimilation (DA) significantly improves the accuracy of field state and parameter estimation by merging experimental data with predictions from high-fidelity numerical models. However, despite their precision and high resolution, the substantial computational cost associated with these numerical models often hinders their practical application in DA. To overcome this challenge, this study presents a novel three-dimensional variational (3D-Var) DA framework that leverages a reduced-order model (ROM) for boundary parameter estimation in computational fluid dynamics (CFD) models. The framework utilizes Proper Orthogonal Decomposition (POD)-Galerkin projection to construct the ROM, enabling near real-time solutions and significantly enhancing computational efficiency. Furthermore, a nonlinear observation operator is developed within the reduced basis space, which directly connects parameters with observational data. This approach eliminates the necessity for full-state reconstruction, thereby further streamlining the computational process. Benchmark results indicate that the proposed method achieves high accuracy and robustness, offering optimal background information for subsequent state estimation. This advancement not only reduces computational overhead but also maintains the integrity and reliability of the estimations, making it a promising tool for real-time applications in complex fluid dynamics scenarios.