Numerical and experimental investigation of one-way fluid structure interaction of a vertical cantilever beam in an air stream

Joubert, Eugene Christiaan (2015-12)

Thesis (PhD)--Stellenbosch University, 2015.

Thesis

ENGLISH ABSTRACT: This thesis uses open source software for simulating one-way fluid structure interaction (FSI) of a bluff-body geometry and validating the results with wind tunnel measurements. A vertical cantilever beam with a rectangular crosssection is used as main subject as it provides complex flow-induced structural loading conditions, stiffness for ensuring one-way FSI and facilitates the placement of sensors internally during wind tunnel measurements to prevent flow obstruction. The beam is orientated parallel to flow resulting in a streamwise cross-section ratio of L/D = 2:63 and perpendicular to the flow to result in a cross-section ratio of D/L = 0:38 where the cross-section length (long side) is L and the width (short side) is D. Two wind speeds are considered for each orientation causing the Reynolds number to vary between 7.6 ×10[to value of 4] < Re < 4×10[to value of 5]. Laboratory measurements were done to provide boundary conditions, material properties, instrumentation calibration and validation data. Wind tunnel measurements were done for two wind speeds, two beam orientations and at various locations around and on the beam surface. This included velocity field measurements using particle image velocimetry (PIV), pressure, vibration and strain. OpenFOAM is used for the computational fluid dynamics (CFD) work. The CFD simulations can be described as turbulent, incompressible, New-tonian, three-dimensional and unsteady. In the near-wall region Spalding's "all-y+" wall function is used to provide some mesh flexibility and reduce computational requirements without compromising accuracy. The Realizable k-ϵ, k-ω SST, Non-linear k-ϵ Shih and Spalart-Allmaras IDDES turbulence models are compared against measured time-averaged velocity and pressure results. The time-averaged CFD results show that all turbulence models accurately reproduce measured flow fields in the upstream and side-wall regions. In the wake the IDDES turbulence model is the most accurate being able to reproduce a recirculation length of 3D compared to the measured length of 3.2D. The time-dependent results show that all models produce loading frequencies with Strouhal numbers between 0:05 < St < 0:06 which is comparable to literature with St ≈ 0:06 but the amplitudes vary. The IDDES model produces the most accurate results and is also second fastest after the k-ω SST model. A segregated FSI method is employed and coupling is achieved using a combination of newly developed OpenFOAM utilities and Python programs which link the fluid and structural meshes. Code-Aster is used for the finite element methods (FEM) for structural analysis. A shell mesh is used with linear approach and a time-dependent load is applied to the structure as provided by the CFD results. Only the parallel beam at U = 30 m/s shows significant oscillations. The results show that the FSI simulations accurately reproduce the strain of 5.8 µm/m in the beam due to the flow drag loading of 12.9 N.

AFRIKAANSE OPSOMMING: Hierdie tesis gebruik vrylik beskikbare sagteware om een-rigting vloeistofstruktuur- interaksie (VSI) van 'n stomp geometrie te simuleer waarna die resultate met windtonnelmetings vergelyk word. 'n Vertikale kantelbalk met 'n reghoekige deursnit is die geometrie wat ondersoek word. Hierdie geometrie verskaf die styfheid benodig vir een-rigting VSI en maak dit moontlik om instrumentasie te huisves tydens windtonneltoetse om sodoende nie die vloei te beïnvloed nie. Die balk word parallel tot die vloei geplaas wat 'n deursnit stroomrigting-verhouding veroorsaak van L/D = 2.63, asook loodreg tot die vloei om 'n stroomrigtings-verhouding van D/L = 0.38 te gee. Twee windsnelhede word ondersoek vir elke oriëntasie wat gevolglik Reynolds getalle van 7.6 × 10[tot mag van 4] < Re < 4 × 10[tot mag van 5] gee. Labratorium toetse is gedoen om die simulasies se randwaardes, materiaal eienskappe, instrumentasie kalibrasie en validasie data te voorsien. Windtonnel toetse is gedoen vir twee windsnelhede en twee balk oriëntasies. Dit sluit in vloeiveld-, druk-, vibrasie- en vervormingsmeetings. OpenFOAM word gebruik vir die numeriese vloeidinamika (NVD) simulasies. Hierdie simulasies kan beskryf word as turbulent, onsamedrukbaar, Newtonies, drie-dimensioneel en tyd-afhanklik. Naby die wand word Spalding se "alle-y+" funksie gebruik wat die simulasie minder sensitief maak vir roosterverfyning in die gebied sonder om die akkuraatheid te verminder. Die Realizable k-ϵ, k-ω SST, Nie-lineêre k-ϵ Shih en Spalart-Allmaras IDDES turbulensiemodelle word vergelyk met gemete snelheid en druk data. Die resultate wys dat die gemete snelhede goed vergelyk met die numeries berekende snelhede vir al die turbulensie-modelle wat oorweeg is, maar slegs die IDDES model gee 'n vergelykbare voorspelling van die stroom-af hersikulsie lengte voorspel as 3D in vergelyking met die 3.2D gemete waarde. Die tyd-afhanklike simulasies van alle modelle bepaal 'n sleur frekwensies met die Strouhal getal tussen 0.05 < St < 0.06 wat vergelykbaar is met die literatuur wat 'n Strouhal getal van St ≈ 0:06 verskaf. Die berekende hefkrag-amplitudes verskil egter tussen die turbulensie modelle. Die IDDES-model verskaf die mees akkurate resultate en is ook die tweede vinnigste na die k-! SST model. 'n Verdeelde VSI metode word gebruik en die koppelling word gedoen deur 'n kombinasie van nuwe kode in OpenFOAM en Python. Code-Aster word gebruik vir die eindige-element metodes om die struktuur te analiseer. Plaatelemente en lineêre modellering word gebruik waarvoor die NVD resultate die tyd-afhanklike kragte verskaf. Slegs die parallele balk by U = 30 m/s toon tydafhanklike gedrag. 'n Vervorming van 5.8 µm/m op die balk word akkuraat bereken deur die VSI-simulasies, wanneer die balk onderwerp word aan 'n ekwivalente sleurkrag van 12.9 N.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/98039
This item appears in the following collections: