Mean wind flow reconstruction of a high-rise building based on variational data assimilation using sparse pressure measurements
Résumé
Data assimilation consists in combining a model and measurements in order to estimate the state of the system. In wind engineering, the interest to take advantage of both sources of information is extremely high. As a matter of fact, on the one hand, measurements in wind tunnel experiments are in practice extremely sparse and, on the other hand, stationary numerical models based on Reynolds averaged Navier-Stokes (RANS) fail to predict the time-averaged wake accurately. Variational data assimilation (VDA) techniques are defined as high-dimensional optimisation problems, the solution of which is determined through an adjoint method. In practical applications, where only wall pressure measurements on the building are available, an accurate mean flow reconstruction is very challenging, especially in regions far away from any measurements. In the present paper, spatially distributed forces are considered as a control parameter on the momentum equation or the turbulence closure equations. These forcings are interpreted as corrections brought to the turbulence closure. Consequently, the problem corresponds to a high-dimensional optimisation problem for a spatially distributed variable with sparse pressure measurements. Some guidelines are brought here to perform accurate and physically consistent reconstructions for wind engineering applications. In order to deal as closely as possible with a practical application, this study focuses on the realizable revision of the k − ϵ RANS model applied on a high-rise building wake flow. In particular, some regularisation strategies as well as efficient techniques for control parameter selection and identification are provided. 1
Origine | Fichiers produits par l'(les) auteur(s) |
---|