Reconstruction of large‐scale flow structures in a stirred tank from limited sensor data

Abstract

We combine reduced order modelling and system identification to reconstruct the temporal evolution of large scale vortical structures behind the blades of a Rushton impeller. We performed Direct Numerical Simulations at Reynolds number 600 and employed proper orthogonal decomposition (POD) to extract the dominant modes and their temporal coefficients. We then applied the identification algorithm, N4SID, to construct an estimator that captures the relation between the velocity signals at sensor points (input) and the POD coefficients (output). We show that the first pair of modes can be very well reconstructed using the velocity time signal from even a single sensor point. A larger number of points improves accuracy and robustness, and also leads to better reconstruction for the second pair of POD modes. Application of the estimator derived at Re=600 to the flows at Re=500 and 700, shows that it is robust with respect to changes in operating conditions. 1This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process which may lead to differences between this version and the Version of Record. Please cite this article as

Analysis of wall mass transfer in turbulent pipe flow combining extended proper orthogonal decomposition and Fukagata-Iwamoto-Kasagi identity

“…The governing equations are solved using an in-house unstructured finite volume solver, Pantarhei [39][40][41][42]. The convection and diffusion terms are discretised using a second-order central approximation.…”

Investigation of the temporal and spatial flow features within the high‐shear mixer by modal decomposition techniques

“…For instance, de Lamotte et al 13,19 applied POD and DMD to identify the dominant flow structures in the stirred tank equipped with 2 four‐blade Rushton turbines through the dataset determined by PIV and CFD independently. Analogously, Papadakis et al 20 and Mayorga et al 21 also utilized the CFD‐calculated velocity field of the Rushton turbines stirred tanks to implement the POD analysis; Tan et al utilized DMD to decompose the CFD‐obtained flow field of the mixed flow pump 22 and three‐stage multiphase pump 23 ; Lacassagne et al 24 and Fernandes del Pozo et al 25 analyzed the flow behavior of non‐Newtonian shear‐thinning fluid in oscillating grid stirred tanks and standard A310 hydrofoil stirred tanks, respectively, by means of POD technique. Due to the limitation of sampling frequency and optically‐accessible regions of PIV, CFD is more favored for generating the dataset of modal decomposition analysis because the instantaneous flow fields at a high spatial and temporal resolution are much more available.…”

On the interaction of turbulence with nucleation and growth in reaction crystallisation

“…The computational implementation is carried out by coupling the in-house DNS code Pantarhei with the in-house population balance modelling code CPMOD. Pantarhei has been used extensively to simulate transitional and turbulent flows in boundary layers, around airfoils, behind fractal grids and inside stirred vessels (Thomareis & Papadakis 2017, 2018; Xiao & Papadakis 2017, 2019; Başbuğ, Papadakis & Vassilicos 2018; Paul, Papadakis & Vassilicos 2018; Mikhaylov, Rigopoulos & Papadakis 2021). The code CPMOD has been employed for various population balance problems, including aerosol synthesis and soot formation (Sewerin & Rigopoulos 2017 a , 2018; Liu & Rigopoulos 2019; Sun, Rigopoulos & Liu 2021).…”