Computational modeling and experiments of an elastoviscoplastic fluid in a thin mold-filling geometry

Abstract

Materials which exhibit elastic and yielding behavior are present in many industrial processes including thin-film coating, oil extraction, manufacturing of consumer products, and food processing. Numerical simulation is a powerful tool for gaining insights into the flow behavior of complex fluids and can facilitate the design of commercially relevant processes, such as mold filling and coating flows. In this study, we perform numerical simulations of an elastoviscoplastic fluid expanding into a thin, rectangular mold. We use a Saramito model to describe the rheology of the fluid (Saramito, 2007) as well as the Bingham–Carreau–Yasuda generalized Newtonian model. The Saramito model used describes the material as an Oldroyd-B fluid above yield and an elastic solid below yield; the yield criterion is based on the von Mises stress. Conservation equations for momentum and mass and the Saramito constitutive equations for stress are solved using the finite element method coupled to a free surface moving mesh algorithm. We assess our Saramito model implementation by comparing computations to benchmarks in the literature, including flow past a cylindrical obstacle and channel flow. We compare results from two and three-dimensional mold filling simulations to flow visualization experiments where fluid fills a thin gap between transparent plates. For both two and three-dimensional, the Saramito model is generally more predictive of the shape of the growing fluid droplet than the Bingham–Carreau–Yasuda model. Saramito model results for 2D computations match experimental droplet shapes well with the addition of a model for the drag terms to capture the effects of the unresolved third dimension. For fully 3D computations, the Saramito model is able to reproduce experimentally-observed droplet shapes for the smallest (5 mL/min) and largest (20 mL/min) flow rates, but struggles to accurately reproduce experimental observations at an intermediate flow rate (10 mL/min). This discrepancy is probably due to the over prediction of slip from the wetting model. • Growing droplet simulated using elastoviscoplastic and generalized-Newtonian models. • EVP model reproduced droplet shapes more accurately than the GN model. • Generalized-Newtonian model typically less accurate than elastoviscoplastic model. • Accuracy of 2D predictions improved by a drag model accounting for unresolved walls.