IUTAM Symposium on Data assimilation in fluid flow

Data assimilation is a powerful technique which has been widely applied by the meteorology, oceanography and more recently the fluid mechanics communities. Its original idea is to optimally combine the equations describing a dynamic system (usually in the form of a numerical model) with observations, which implies a hybrid modelling approach using both observed data and equations. In other words, it integrates numerical and experimental results instead of validating/calibrating the former by the latter. Mathematically, this process can be interpreted as a nonlinear optimization that can be achieved by exploiting e.g. physics-informed neural network, Kalman filter, variational methods, etc. and it faces similar challenges as optimal flow control and perturbation analyses. It has been used for calibration and better prediction/reconstruction of flow fields, as well as the modification of parameters in turbulence models.

In the era of big data, in response to the massive data produced, data assimilation has been extended to the fusion of flow data from various sources, which can be with different fidelities (e.g. simulations with various resolutions) and regarding different quantities (e.g. measurements targeting at different physics). Such a fusion can be achieved by constructing a data-based assimilation without utilizing the governing equations.

Therefore the term data assimilation incorporates several active subjects in fluid dynamics, as well as some latest advances in the application of machine learning in fluid mechanics. Further considering that local financial supports have been granted by the Chinese research council NSFC, this symposium will bring together researchers in several areas including data-based turbulence modelling, optimal perturbation analyses, flow control, reduced-order modelling, etc.