Numerical study of flow in diffuser- Investigation with three turbulence models and comparison with experimental data

Sammanfattning: Fluid dynamics is a branch of physics that has many applications when it comesto solving real-world problems. Areas that utilise the knowledge in fluid dynamicsin order to develop technology that is more efficient and less resource cravingare diverse and range from the energy sector to the transport sector. Furthermore,as computational power increases, it becomes both more cost- and time effective tomove from conducting practical experiments to instead perform numerical computersimulations.This project has studied the possibility of conducting numerical simulationsof the separation phenomenon and also if it is possible to capture the effects offlow control meant to minimise separation. More specifically this thesis has focusedon the case with flow through a conical diffuser with an annular inlet and a centerbody present, where the center body causes the flow to separate. This center bodybefore the diffuser part could in practical applications be in the form of a bearinghub. Separation is usually an undesirable feature for flows in confined space and theability to counteract its development through both passive (geometrical alterations)and active (for example injection of jets) flow control mechanisms are importanttools to an optimised diffuser design. The availability of the well performing opensource program such as OpenFOAM is a further reason to why the development ofaccurate numerical methods is of particular interest.The simulations undertaken in this project has produced results in the form of threecomponent mean velocity distributions of the flow in a conical diffuser with an annularinlet and center body present. Results were then compared with experimentalreference data. The investigation covered three turbulence models (k-!, k-!-SSTand k-!-SST-SAS) in two different geometries, corresponding to the implementationof a passive separation control feature by implementing a straight section afterthe center body (in order to minimise separation and in accordance with referenceexperiment for comparison). Additionally, the effect of a flow control method calledCoanda blowing in order to minimise separation was thoroughly investigated. Aswirl component of the inlet velocity in order to increase pressure recovery wasanother implementation made and compared to reference data.The results showed that the k-! and k-!-SST models managed to capture the generalmotion of the flow with and without passive flow control. However, the Coanda effectfrom the jets proved difficult to capture, even when a refined mesh was created. Thek-!-SST-SAS model proved ineffective here, despite its said superiority in previousexperiments. Probably, lack of previous experience in CFD simulations as well ashigher mesh demands for the SAS model explains its poor results. The results,however, validated the proposed coupling between separation in the wake and in thewall boundaries as well as how swirl can generate an increased pressure recovery.