Mechanical properties measurement and modelling

Mechanical properties measurement andmodelling

Ruzena Chamrova

Laboratory of Construction MaterialsEcole Polytechnique Federale de Lausanne

Lausanne, Switzerland 2008

Chamrova, Scrivener, Guidoum Project 10 - Mechanical properties measurement and modelling

Scientific view differs from the industrial

I scientific view - fitting a curveI industrial view - same porosity, considerable difference in

strength

porosity

com

pres

sive

str

engt

h [M

Pa]


The aim is to study mechanical performance ofhydrating cement paste

Aim of the project

I measurement and prediction at early ageI development of in-house FEM code for the predictionI study relationship between microstructural development

and mechanical performance


Mechanical properties of microstructure are computedby FEM

Mechanicalproperties

Mesh + FEM

Experiment

Model


Factors influencing the FEM prediction

Mechanical propertiesprediction

Microstructural input

Porosity approach

Connectedness approach

Mesh

Boundaryconditions


Unstructured meshes capture the geometry better

UnstructuredSemi-structuredStructured

Smoothing

Octree

DelaunayVoxel

Regular tetrahedral


Semi-structured meshes offer a trade-off

UnstructuredSemi-structuredStructured

Smoothing

Octree

DelaunayVoxel

Regular tetrahedral


Delaunay mesh does not seem to be practical atpresent

I mesh improvement - spike reduction - is time consumingI memory limitations and complexity of overlaps allow to go

up to 50000 particlesI this approach might be necessary e.g. for transport

properties modelling


Structured tetrahedral mesh offers trade-off betweensimplicity of generation and capturing of geometry

Tetrahedral meshVoxel mesh

Subdivision into 6 tetrahedra

I mesh is stillunnecessarilyjagged


Problem of the subdivision into 6 lies in the orientationof elements

I orientation ofelements notsuitable forspheres

I rotating elementswould meanincompatibility forFEM


2D example shows that subdivision into 5 tetrahedra ismore suitable

Voxel mesh Subdivision - 5

Subdivision - 6


Tetrahedral meshes produce higher E (10 - 15 %)

degree of hydration [%]

Ela

stic

Mod

ulus

[GP

a]tetrahedra – 6tetrahedra – 5voxels

E for various types of meshes


Tetrahedral meshes produce higher E (10 - 15 %)


voxelstetrahedra – 5tetrahedra - 6

Poi

sson

's r

atio

[-]


On the contrary volume ratios differ by 1 %

Possible explanations

I solely volume ratios cannot precisely explain thisphenomenon

I the connections in the structure got denser with thetetrahedral mesh

I linear tetrahedral element (CST) might not be comparablewith the same size of a hexahedral element


And we are still not happy with the geometry...

Under implementation

Smoothing Octree





Porosity approach


Mesh

Boundaryconditions


More realistic way is not to mesh porosity

Porosity meshing

I sensitive to the elasticproperties assigned toporosity

Real porosity

I computational timeseveral times higher


Real porosity approach predicts lower E...


Ela

stic

Mod

ulus

[M

Pa]

meshed porosityreal porosity


...and lower Poisson’s ratio


K =2.18 GPareal porosity

Poi

sson

's r

atio

[-]





Porosity approach


Mesh

Boundaryconditions


Mechanical properties do not develop from the verybeginning of hydration

All phases intocomputation

Only connectedphases intocomputation

I unconnected phases are thrown away by the burningalgorithm which is provided by the microstructural model


Connectedness seems to be the most influential factor


All phasesOnly connected

Ela

stic

Mod

ulus

[M

Pa]





Porosity approach


Mesh

Boundaryconditions


There is little difference between BCs - RVEhypothesis correct


Ela

stic

Mod

ulus

[M

Pa]

displacement BCstatic BCperiodic BC


2D plain strain of slices is not a substitution for 3Dcomputation

Elastic Modulus of the microstructural slices – displacement BC

Slice of the microstructure

Ela

stic

Mod

ulus

[M

Pa]

direction Xdirection Ydirection Z3D structure


High E of some slices is caused by low porosity andhigh alite ratio

3D phase ratios 2D phase ratios

Comparison of the phase ratios

phas

e ra

tios

I for correct 2Dpredictionnecessary togenerate new 2Dmicrostructurebased on 3Ddistribution


Next plans

code development and prediction of

mechanical propertiesEPFL

porosity prediction, strengthtests - Aalborg

Prague – April 2008

code development EPFL

validation and thesis writingEPFL


Mechanical properties measurement and modelling

Documents