CFD modeling and optimization by metamodels of a squirrel cage fan using OpenFoam and Dakota: Ventilation applications. (November 2021)

• Record Type: Journal Article
• Authors: Benchikh Le Hocine, Alla Eddine; Poncet, Sébastien; Fellouah, Hachimi
• Is Part Of: Building and environment. Volume 205(2021)
• Publication Date: 2021-11
• Subjects: Squirrel cage fan -- OpenFoam -- Dakota -- Design of experiments -- Surrogate model -- Latin Hypercube Sampling -- Turbulence; Buildings -- Environmental engineering -- Periodicals; Building -- Research -- Periodicals; Constructions -- Technique de l'environnement -- Périodiques; Electronic journals; 696
• DOI: 10.1016/j.buildenv.2021.108145
• Languages: English
• ISSNs: 0360-1323

Abstract

Abstract: The constraints related to air quality are progressively increasing making the optimization of the air system crucial. Multiple fan configurations are currently used in vacuum systems, the most widely employed one being the squirrel cage fan (SCF), also known as forward-curved multi-blade centrifugal fan. Most of researches so far used commercial software to optimize the SCF focusing mostly on the impeller region. In the present work, a complete automatic optimization process loop is developed based only on open source libraries: Dakota to manage all the processes, Salome to generate the geometry and the mesh grid and OpenFoam for the numerical calculations. Up to seven design parameters are selected for the impeller, blades, and volute. The Latin Hypercube Sampling (LHS) method is preferred to determine the design points and then different metamodels are built. A 3D incompressible simpleFoam solver is coupled to the Multiple reference frame (MRF) approach to model the flow in the SCF. The total efficiency is improved by 8.46% by changing the hub shape and by adding further design parameters. A strong interaction is observed between the cut-off vortex and flow separation on the blade upper side. The optimal design is finally validated against the produced prototype, with an error of 3.4% on the pressure gradient at the design point. Highlights: The modeling and optimization loop are developed using OpenFoam and Dakota. Kriging Metamodel predicts accurately the efficiency (2.11%) and forces (2.38%). The Efficient Global Optimization (EGO) allows to improve the efficiency by 8.46%. The cut-off vortex causes flow separation around the blades and efficiency decrease. The numerical approach is validated against experiments with an error of 3.4%.