SPH for the Modeling of non-newtonian Fluids with Thermal-Dependent Rheology

Similar documents
Computational Astrophysics

MECHANICAL PROPERTIES

Rheology and Constitutive Equations. Rheology = Greek verb to flow. Rheology is the study of the flow and deformation of materials.

Game Physics. Game and Media Technology Master Program - Utrecht University. Dr. Nicolas Pronost

150A Review Session 2/13/2014 Fluid Statics. Pressure acts in all directions, normal to the surrounding surfaces

Preliminary validation of lava benchmark tests on the GPUSPH particle engine

Pressure corrected SPH for fluid animation

Please remember all the unit that you use in your calculation. There are no marks for correct answer without unit.

Chapter 3 Non-Newtonian fluid

Chapter 1. Continuum mechanics review. 1.1 Definitions and nomenclature

1. Introduction, fluid properties (1.1, 2.8, 4.1, and handouts)

This chapter is a study of the shear stress as a function of the shear rate for Newtonian and non-newtonian biological materials.

Modelling of dispersed, multicomponent, multiphase flows in resource industries Section 4: Non-Newtonian fluids and rheometry (PART 1)

Soft Bodies. Good approximation for hard ones. approximation breaks when objects break, or deform. Generalization: soft (deformable) bodies

Lecture 3. Properties of Fluids 11/01/2017. There are thermodynamic properties of fluids like:

Viscous dissipation and temperature dependent shear thinning rheology

Rheological Properties

An Overview of Fluid Animation. Christopher Batty March 11, 2014

SIMULATIONS OF THE 2004 LAVA FLOW AT ETNA VOLCANO BY THE MAGFLOW CELLULAR AUTOMATA MODEL

Fundamentals of Fluid Dynamics: Elementary Viscous Flow

Duality methods for variational inequalities and Non-Newtonian fluid mechanics

Chapter 9: Differential Analysis

CHAPTER 1 Fluids and their Properties

Fluid Animation. Christopher Batty November 17, 2011

Chapter 9: Differential Analysis of Fluid Flow

Computational Fluid Dynamics 2

On a variational inequality of Bingham and Navier-Stokes type in three dimension

Modelling the Rheology of Semi-Concentrated Polymeric Composites

V (r,t) = i ˆ u( x, y,z,t) + ˆ j v( x, y,z,t) + k ˆ w( x, y, z,t)

Chapter 2: Fluid Dynamics Review

1. The Properties of Fluids

Chapter 6 Molten State

Computer Fluid Dynamics E181107

Numerical Heat and Mass Transfer

Petroleum Engineering Dept. Fluid Mechanics Second Stage Dr. Ahmed K. Alshara

Process Development for a High-Throughput Fine Line Metallization Approach Based on Dispensing Technology

Summary of the Equations of Fluid Dynamics

CSCI1950V Project 4 : Smoothed Particle Hydrodynamics

Contents. Preface XIII. 1 General Introduction 1 References 6

KELVIN-HELMHOLTZ INSTABILITY BY SPH

Fluid Dynamics. Part 2. Massimo Ricotti. University of Maryland. Fluid Dynamics p.1/17

Quick Recapitulation of Fluid Mechanics

Pharmaceutics I. Unit 6 Rheology of suspensions

(Refer Slide Time: 2:14)

Getting started: CFD notation

Nonlinear Wave Theory for Transport Phenomena

- Marine Hydrodynamics. Lecture 4. Knowns Equations # Unknowns # (conservation of mass) (conservation of momentum)

APPENDIX A USEFUL EQUATIONS (METRIC AND IMPERIAL SYSTEMS) THE DEFINITION OF VISCOSITY RHEOLOGICAL (VISCOUS BEHAVIOR) PROPERTIES OF FLUIDS

Dynamics of Glaciers

Chapter 5. The Differential Forms of the Fundamental Laws

Lecture 2: Constitutive Relations

Viscoelasticity. Basic Notions & Examples. Formalism for Linear Viscoelasticity. Simple Models & Mechanical Analogies. Non-linear behavior

Smooth Particle Hydrodynamic (SPH) Presented by: Omid Ghasemi Fare Nina Zabihi XU Zhao Miao Zhang Sheng Zhi EGEE 520

1 Exercise: Linear, incompressible Stokes flow with FE

Differential relations for fluid flow

Liquid fuels viscosity (non-newtonian fluids)

Introduction to Fluid Mechanics

Particle-based Fluids

Continuum mechanism: Stress and strain

2. FLUID-FLOW EQUATIONS SPRING 2019

Lattice Boltzmann approach to liquid - vapour separation

The Physics of Non-Newtonian Fluids

Chapter 1 Fluid Characteristics

Entropy generation and transport

Pharmaceutical compounding I Colloidal and Surface-Chemical Aspects of Dosage Forms Dr. rer. nat. Rebaz H. Ali

7 The Navier-Stokes Equations

Introduction to Marine Hydrodynamics

NUMERICAL SIMULATION OF THE FLOW OF A POWER LAW FLUID IN AN ELBOW BEND. A Thesis KARTHIK KANAKAMEDALA

ME 262 BASIC FLUID MECHANICS Assistant Professor Neslihan Semerci Lecture 4. (Buoyancy and Viscosity of water)

Conservation of Mass. Computational Fluid Dynamics. The Equations Governing Fluid Motion

Introduction to Viscometry and Rheology, Basics, Rotational Testing. Basic Seminar Applied Rheology

Liquids and solids are essentially incompressible substances and the variation of their density with pressure is usually negligible.

CESSATION OF VISCOPLASTIC POISEUILLE FLOW IN A RECTANGULAR DUCT WITH WALL SLIP

UNIVERSITY of LIMERICK

Petroleum Engineering Department Fluid Mechanics Second Stage Assist Prof. Dr. Ahmed K. Alshara

Polymerization Technology Laboratory Course

Part II Fundamentals of Fluid Mechanics By Munson, Young, and Okiishi

Modeling and simulation of bedload transport with viscous effects

Fluid Mechanics Abdusselam Altunkaynak

Rheometer: Procedure: Part A: Viscosity v Time

Dynamics of the Mantle and Lithosphere ETH Zürich Continuum Mechanics in Geodynamics: Equation cheat sheet

CH5716 Processing of Materials

Introduction to Geology Spring 2008

CHAPTER (2) FLUID PROPERTIES SUMMARY DR. MUNZER EBAID MECH.ENG.DEPT.

Physics-Based Animation

Notes 4: Differential Form of the Conservation Equations

Math background. Physics. Simulation. Related phenomena. Frontiers in graphics. Rigid fluids

Smoothed Particle Hydrodynamics

Chapter 2. General concepts. 2.1 The Navier-Stokes equations

Numerical Simulation of Newtonian and Non-Newtonian Flows in Bypass

CH.9. CONSTITUTIVE EQUATIONS IN FLUIDS. Multimedia Course on Continuum Mechanics

FUNDAMENTAL STUDY OF BINGHAM FLUID BY MEANS OF DAM-BREAK FLOW MODEL

ESCI 485 Air/Sea Interaction Lesson 1 Stresses and Fluxes Dr. DeCaria

Smoothed Particle Hydrodynamics (SPH) 4. May 2012

Fluid Mechanics II Viscosity and shear stresses

Navier-Stokes Equation: Principle of Conservation of Momentum

Fluid Dynamics Exercises and questions for the course

Chapter 1. Governing Equations of GFD. 1.1 Mass continuity

Smoothed Particle Hydrodynamics (SPH) Huamin Wang

We may have a general idea that a solid is hard and a fluid is soft. This is not satisfactory from

Transcription:

SPH for the Modeling of non-newtonian Fluids with Thermal-Dependent Rheology G. Bilotta 1,2 1 Dipartimento di Matematica e Informatica, Università di Catania, Italy 2 Istituto Nazionale Geofisica e Vulcanologia, Sezione di Catania, Italy 20140701 BAW/RUB workshop on Applied SPH

Outline 1 Motivations 2 The problem 3 Non-Newtonian Rheologies 4 SPH discretization 5 Lava flow simulations 6 Conclusions

Motivations 1 Motivations 2 The problem 3 Non-Newtonian Rheologies 4 SPH discretization 5 Lava flow simulations 6 Conclusions

Motivations

Motivations

Motivations Work started as part of LAVA-V3 project (INGV / DPC). Aim: develop a new numerical model for lava flows capable of modeling all aspects of the flow, including: solidification crust/front/levee formation and rupture tunnel formation ephemeral vent opening

People Alexis Hérault modeling and GPU implementation Giuseppe Bilotta modeling and GPU implementation Eugenio Rustico multi-gpu implementation Ciro Del Negro project coordiator

The problem 1 Motivations 2 The problem 3 Non-Newtonian Rheologies 4 SPH discretization 5 Lava flow simulations 6 Conclusions

The problem Navier-Stokes equations: Dρ Dt = ρ u, eqn. of state or u = 0, ρdu Dt = P + τ + F, where ρ, u, P, F: density, velocity, pressure, external forces (gravity) τ = µ u: viscous stress

The problem Navier-Stokes and heat equations: Dρ Dt = ρ u, eqn. of state or u = 0, where ρdu Dt = P + τ + F, DT ρc p = (κ T) + τ u Dt ρ, u, P, F: density, velocity, pressure, external forces (gravity) τ = µ u: viscous stress c p, κ: specific heat capacity and thermal conductivity

Non-Newtonian Fluids with Thermal-Dependent Rheology Thermal depenency: heat equation has a viscous heating term: i,j τ ij u i / x j ; (Costa & Macedonio, 2003) µ = µ(t). Non-Newtonian fluids: τ µ u. (or rather, µ is not independent of u).

1 Motivations 2 The problem 3 Non-Newtonian Rheologies 4 SPH discretization 5 Lava flow simulations 6 Conclusions

Rheology Rheology: study of flow of deformable matter when subject to stress. Also: the law with which deformable matter flows, i.e. the relation between shear stress and rate of strain. Newtonian rheology: linear relation τ = µ u, coefficient µ (viscosity) does not depend on u. (Or conversely: u = φτ, with φ = 1/µ fluidity, independent from τ.) Most fluids are actually non-newtonian: stress/strain relation depends on the stress/strain rate, but we can always talk about the apparent viscosity (ratio of stress to strain rate).

Non-Newtonian rheologies Categories: shear-thinning or pseudo-plastic: (apparent) viscosity decreases with stress/shear rate (e.g. paint); shear-thickening or dilatant: (apparent) viscosity increases with stress/shear rate (e.g. cornstarch in water); Can be time-dependent (shear thinning: thixotropic; shear thickening: rheopectic) or time-independent: Bingham plastic, power-law rheologies, Herschel-Bulkley fluids, etc.

Beyond Newton One-dimensional formulations: Bingham plastic : τ = τ 0 + Kγ, no flow if τ < τ 0, Newton-like behavior afterwards with τ stress, γ shear rate, τ 0 limit stress, K consistency index, n power index.

Beyond Newton One-dimensional formulations: Bingham plastic : τ = τ 0 + Kγ, no flow if τ < τ 0, Newton-like behavior afterwards Power law : τ = Kγ n ; Newton for n = 1, pseudo-plastic for n > 1, dilatant for n < 1; with τ stress, γ shear rate, τ 0 limit stress, K consistency index, n power index.

Beyond Newton One-dimensional formulations: Bingham plastic : τ = τ 0 + Kγ, no flow if τ < τ 0, Newton-like behavior afterwards Power law : τ = Kγ n ; Newton for n = 1, pseudo-plastic for n > 1, dilatant for n < 1; Herschel-Bulkley : τ = τ 0 + Kγ n, general case that includes Newtonian, Bingham and power-law behavior; with τ stress, γ shear rate, τ 0 limit stress, K consistency index, n power index.

Rheology plots/i shear stress more viscous less viscous strain rate

Rheology plots/ii strain rate less viscous more viscous shear stress

Rheology plots/iii dilatant strain rate newtonian pseudoplastic Bingham plastic Herschel-Bulkley pseudoplastic shear stress

Herschel-Bulkley General formulation, with temperature dependency: { µ µ(t) = 0 (T) if ɛ ɛ 0 (T), K(T)ɛ n(t) + τ 0(T) ɛ if ɛ > ɛ 0 (T) (ɛ: second invariant of rate of strain tensor) infinite viscosity for null strain

Herschel-Bulkley General formulation, with temperature dependency, regularized (Zhu, following Papanastasiou): { µ µ(t) = 0 (T) if ɛ ɛ 0 (T), K(T)ɛ n(t) + τ 0 (T) 1 exp( αɛ) ɛ if ɛ > ɛ 0 (T) (ɛ: second invariant of rate of strain tensor) (α some large regularizing constant)

SPH discretization 1 Motivations 2 The problem 3 Non-Newtonian Rheologies 4 SPH discretization 5 Lava flow simulations 6 Conclusions

Smoothed Particle Hydrodynamics Smoothed Particle Hydrodynamics. Lagrangian meshless method with smoothed field (position, velocity, temperature) interpolation: f (x) = f (y)δ(x y)dy f (y)w(x y, h)dy Ω Ω j f (x j )W(x x j, h)v j where W is a smoothing kernel that approximates the Dirac s δ, V j is the volume of particle j, kernel with compact support (radius h) = summation extended to neighbors only.

SPH kernel properties E.g. density: ρ(x) j ρ(x j )W(x x j, h)v j = j m j W(x x j, h) where m j = V j ρ(x j ) is the mass of particle j. Kernel normalization (Shepard): W(x x i, h) = W(x x i ) j W(x x j)v j allows exact reconstruction of constant fields: normalized kernels ensure zero-order accuracy, symmetric kernels ensure first-order accuracy, positive kernels ensure isolated particles have nonnegative density.

SPH gradients/i For gradients, we use Green s formula to get f ( x) = f ( y)δ( x y)d y f ( y)w( x y, h)d y Ω Ω f ( y) y W( x y, h)d y + Ω + f ( y)w( x y, h) nds j Ω m j ρ j f j j W( x x j, h) +??? = = j m j ρ j f j i W( x x j, h) +???. If supp W i Ω, then??? = 0. Otherwise, correction terms have to be added.

SPH gradients/ii Other possible formulas: f (x i ) m j (f ρ j ± f i ) i W(x i x j, h), j j f (x i ) 1 m ρ j (f i f j ) i W(x i x j, h), i f (x i ) ρ i j j m j ( fj ρ 2 j + f i ρ 2 i ) i W(x i x j, h). Chosen for different fields for different stability properties. Common property: explicit computation, no need to invert linear systems.

SPH discretization/i Discretized continuity equations (W ij = W(x i x j, h), f ij = f i f j ): ( ) dρ = m dt j u ij i W ij or i or (summation density): ( ) dρ m j = ρ dt i u i ρ ij i W ij, j ρ i = j m j W ij. (but summation density can only be used reliably in absence of a free surface).

SPH discretization/ii Discretized momentum and thermal equations (W ij = W(x i x j, h), f ij = f i f j ): ( ) du = ( Pj m dt j i ρ 2 j + m j µ i + µ j ρ i ρ j ( ) dt c pi = 4m j dt i ρ i ρ j 2m j ρ i ρ j + P i ρ 2 i ) i W ij + r ij i W ij u ij + g, r 2 ij κ i κ j r ij i W ij T κ i + κ ij j rij 2 + µ i µ j µ i + µ j (r ij u ij ) 2 r ij i W ij r 4 ij

Weakly compressible SPH To preserve direct computation, slightly compressible flows with equation of state P = ρ (( ) 0c s ρ γ 1) γ (ρ 0 density at rest, c s (artificial) speed of sound, taken an order of magnitude larger than the maximum expected speed of the phenomenon). Speed of sound is chosen ficticious because it limits time-stepping in explicit methods. (For lava we don t really worry about that, because the viscosity is actually what limits the timestep.) ρ 0

Viscosity Non-Newtonian fuid: µ i is the apparent viscosity of the particle, computed separately before each force computation. Strain rate tensor: u i = j u ij i W ij V j Second invariant: ɛ i = 1 ( (tr u) 2 2 tr( u 2 ) ) Then compute µ i from (regularized) Herschel-Bulkley

Boundary conditions: ground/i Lava ows on complex topographies, we assume them de ned through a Digital Elevation Model (DEM): 2D grid of elevations, cell side p. Interpolate to nd z0 height at projection (x0, y0) of particle on DEM, again to nd height at (x0 + p, y0), (x0, y0 + p), nd plane by three points, interact with plane.!y P(x,y,z)!!x " z0!y P(x,y,z)!!x " z0 z2 z1

Boundary conditions: ground/ii! n h! u r! u T Friction:!! u! f = µs 1 u T f = µ T! n! " ( h2 # r 2 ) with S 1 ground particle surface. Ground particle surface can be estimated by S 1 = S g /N g or set as problem parameter.

Boundary conditions: ground/iii Conduction to ground: particle close to the DEM lose T = κs 1 c p m T T ground t r due to heat exchange with ground. Ground temperature T ground is assumed constant.

Boundary conditions: free surface/i Need to detect free surface for radiative losses. Standard approach: check for neighbors in cone pointing towards gradient of characteristic function. (a) (b) Trick: lava flows are essentially two-dimensional, check for presence of particles in upwards cone. (a is on the surface, b is not).

Boundary conditions: free surface/i Radiative losses: T = K BεS 2 c p m (T 4 T 4 ext ) T. with K B Stefan-Boltzmann constant, ε emissivity. External temperature T ext is assumed constant. Free-surface particle surface S 2 is either a problem parameter, or can be computed as S 2 = S f /N f.

Phase change q: fraction of latent heat for phase change; q = 0: liquid, q = 1: solid; T : new temperature from integration of heat equation; if T < T s and q < 1, then: T = T s increment q by c p (T s T )/L if q 1, then: T = T s L(q 1)/c p q = 1 otherwise T = T.

Implementation/I Weakly compressible SPH is embarrasingly parallel (each particle can be computed concurrently and independently from the others, by just reading their data).

Implementation/II SPH: excellent candidate for GPU implementation, > 150 faster than (serial) CPU, 50 faster than parallel CPU implementation. GPUs also give us free interpolation (in hardware), good for topography.

Implementation/III Based on GPUSPH, open source implementation of SPH on GPU using CUDA, available from http://gpusph.org implements SPH for Newtonian fluid-dynamics multiple SPH formulations variety of smoothing kernel choices variety of viscosity models variety of boundary conditions single GPU, multi-gpu, multi-node

Lava flow simulations 1 Motivations 2 The problem 3 Non-Newtonian Rheologies 4 SPH discretization 5 Lava flow simulations 6 Conclusions

Parameters parameter symbol values unit density ρ 2400 kg m 3 heat capacity C p 1150 J kg 1 K 1 latent heat L 3.3 10 5 J kg 1 conductivity κ 2.2 W m 1 K 1 emissivity ε 1 vent temperature T e 1350 K solidus temperature T s 1163 K ground temperature T ground 293 K air temperature T ext 293 K

Temperature-dependent parameters Consistency factor from Giordano & Dingwell (2003): log K(T) = 4.643 + 58122.44 427.04 [H 2O] T 499.31 + 28.74 ln([h 2 O]) with T in Kelvin and [H 2 O] is the water content in weight percentage. In our tests [H 2 O] = 0.005%. Yield strength from Miyamoto (1997): log τ 0 (T) = 0.0089(T T e ) + 1.9. Regularizing coefficient α = 1000, and for Herschel-Bulkley: n(t) = 0.1 (shear-thinning).

Apparent viscosities 200 100 50 1363 K 1263 K 100 1363 K 1263 K 20 10 10 5 1 2 0.1 0.01 100 Bingham 0.1 100 Herschel-Bulkley

Newtonian Constant viscosity Temperature-dependent viscosity

Bingham Constant viscosity Temperature-dependent viscosity

Herschel-Bulkley Constant viscosity Temperature-dependent viscosity

Conclusions we can simulate both Newtonian and non-newtonian fluids with SPH; coupled thermal/dynamic problems are possible, including phase transition and thermal dependent rheology; GPU implementation gives us good performance at low cost, and some things for free (e.g. DEM interpolation); our model can be used to explore non-newtonian fluids, and help finding what is the most correct rheology for lava.