Characteristics of Inertial Fibre Suspensions in a Turbulent Pipe Flow Stella Dearing*, Cristian Marchioli, Alfredo Soldati Dipartimento di Energetica e Macchine, Università di Udine, Italy 1 3rd SIG43 Workshop: Udine, Italy, rd SIG43 Workshop: Dynamics of non-spherical particles in fluid turbulence, Udine, Italy, April 6th - 8th, 2011

Applications: Which industries benefit Inertial fibres suspended in turbulent flows are commonly encountered industry 2 Pulp and paper processing Controlling rheological behaviour and fibre orientation distribution crucial to optimise operations Furniture Industry Pneumatic transport of fibres Fibres as an alternative to polymers as a DRA? Examples include the TAPs, medical application, firehouse Fibres provide more modest reductions but improved shear degradation and filterability Nappy Fabrication Fluff, cellulose fibres, make up 60% of nappy material. Uniform distribution is highly desirable for increased absorption Experimental data is required to validate simulation assumptions, develop and improve current models, provide industry with practical correlations for sizing of industry equipment (aim not to oversize) 3rd SIG43 Workshop: Udine, Italy, 2011

Motivation: Why Study? 3 Fibre suspensions have complicated rheological properties that are quite different than carrier fluid even at low mass fractions. The physical properties f(fibre spatial distribution & orientation) 1.Fibres rotate/align when subject to velocity gradient – hydrodynamic drag of fibres is f(rotational motion)  controls translational motion 2.Strong turbulence tends to randomise orientation (BERNSTEIN &SHAPIRO) 3.BUT coherent structures  interact with fibres causing segregation and accumulation ( MARCHIOLI et al. PHYS. FLUIDS ) y z 3rd SIG43 Workshop: Udine, Italy, 2011 DNS  one way coupling  no fibre-fibre interactions  Re τ =150; Re = 9000  Benchmarking required!

Set-up: Pipe Circuit measurements 4 Pipe length9m31m Pipe materialglassGalvanised steel Pipe diameter ϕ / 0.022m0.1m Re6,000-20,00025, ,000 3rd SIG43 Workshop: Udine, Italy, 2011 Laser Camera z x Mirror ND Yag laser 1000mJ PCO sensicam 1280 x1024 F = 8hz x z Flow Direction ϕ /7 ϕ /5

Experimental Parameters Fibre typeMass fraction, ϕ, % nL 3 Nylon AdditiveLengthDiameterSpecific Density Nylon~0.3mm24 μm rd SIG43 Workshop: Udine, Italy, 2011 Frequency, %

Software development: Phase Discrimination 6 Adjust intensity Dilate Remove noise Erode to orig. size 3rd SIG43 Workshop: Udine, Italy, 2011 zizi Mean orientations Normalised number density 1.Preprocessing 2. Object Identification 3. Discriminate object type based on length & aspect ratio 4.Ellipse fitting 5.Calculate statistics 1.Preprocessing 2. Object Identification 3. Discriminate object type based on length & aspect ratio 4.Ellipse fitting 5.Calculate statistics

Software development: Statistics calculation and validation 7 3rd SIG43 Workshop: Udine, Italy, 2011 Results  mean orientation Green triangles  DNS data (Marchioli et al., 2010) at Re = 9000, Red squares  experimental data Re = 8043 Artificial images Projected fibre onto plane  random distribution tends to 0.64 vs 0.5

Results: Mean orientations 8  Fibres strongly align to the streamwise direction close to wall  Alignment decreases with distance from wall but does not become completely random  Alignment reduction increases with Re for mass fraction of 0.01%  Alignment reductions tends to same value for all Re at mass fraction of 0.02%. 3rd SIG43 Workshop: Udine, Italy, 2011 DNS Re τ =150; Re = 9000

Results: Normalised mean number density 9  Near wall peak; reduction away from wall  Less fibres at near wall cf at higher Re cf to Re=8043  At 0.02% less fibres close to wall, increase in fibres away from wall 3rd SIG43 Workshop: Udine, Italy, 2011 DNS Re τ =150; Re = 9000 Re = 8043

Conclusions 10 Phase discrimination technique and validation has been presented. – Limitations and errors have been quantified. –Phase discrimination technique is appears suitable for the problem ( 3D set-up) First set of experimental data of orientation/distribution close to wall Mean orientation and number density data qualitatively agrees with DNS data  promising. High Re number tests: sensitive to concentrations. –Deposition at higher Re tests. Why? More information required Information on decomposed velocity flow field is required. Study for instantaneous images with high frequency camera We are in the process of doing error analysis of phase discrimination PIV algorithm using a DNS flow field 3rd SIG43 Workshop: Udine, Italy, 2011