Implementation Aspects of Galerkin and Least-Squares Petrov-Galerkin Reduced Order Models for Unsteady Flows

Autor(es)

Zucatti, Victor ; Wolf, William R.

• Tipo de documento: Artigo Científico
• Fonte: AIAA SCITECH 2021 FORUM; v. N/A, p. 17-pg., 2021-01-01.

Resumo

Implementation aspects of Galerkin and least-squares Petrov-Galerkin (LSPG) reduced-order models (ROMs) are discussed. An assessment of calibration, hyper-reduction, time accuracy and the choice of time marching schemes is presented. Galerkin and LSPG ROMs are tested on two compressible flow problems. The canonical low Reynolds flow past a cylinder is first analysed to serve as a benchmark for calibrated Galerkin and LSPG ROMs. The second problem consists of the turbulent flow over an airfoil undergoing deep dynamic stall. Results are analysed taking into account the impact of different time-marching schemes, degree of hyper-reduction, time step size, number of POD modes used in reconstruction and a measure calibration intrusiveness. Results show that calibration significantly improves accuracy and stability for both Galerkin and LSPG ROMs. For the first problem investigated, both methods are capable of providing results in excellent agreement with the full-order model. Results demonstrate that a very aggressive hyper-reduction can be used with the LSPG method. On the other hand, ROMs require a fine-tuning for the dynamic stall case in order to reproduce the relevant flow dynamics. Results shows that the calibrated Galerkin models are more robust and computationally efficient than those obtained using the LSPG scheme. Also, Galerkin models benefit from the non-conservative form of the compressible Navier-Stokes equations by not requiring hyper-reduction to reduce complexity. (AU)

• Processo FAPESP: 18/11410-9 - Utilização da análise de componentes principais para a construção de modelos de ordem reduzida
• Beneficiário: Victor Zucatti da Silva
• Modalidade de apoio: Bolsas no Brasil - Mestrado
• Processo FAPESP: 13/08293-7 - CECC - Centro de Engenharia e Ciências Computacionais
• Beneficiário: Munir Salomao Skaf
• Modalidade de apoio: Auxílio à Pesquisa - Centros de Pesquisa, Inovação e Difusão - CEPIDs
• Processo FAPESP: 19/18809-7 - Estratégias de modelagem de ordem reduzida para simulação de escoamentos não estacionários
• Beneficiário: Victor Zucatti da Silva
• Modalidade de apoio: Bolsas no Exterior - Estágio de Pesquisa - Mestrado