Experimental In-Situ Tests and Image Processing For A Realistic Image-Based FE Meso-Model For Cementitious Materials

Experimental in-situ tests and image processing for a realistic
image-based FE meso-model for cementitious materials

∗1,2 †1 ‡2
Author A. One , Presenting Author , and Author C. Three
1
Affiliation 1, Country1
2
Affiliation 2, Country2
Keywords: x-ray tomography, image processing, cementitious materials, meso-scale FE modelling
Summary: This work makes the link between advanced experimental techniques of x-ray and neutron CT and
the reliable characterization of the heterogeneities of a cementitious material. The final goal is to provide a
reliable data set which can be used as an input for a realistic imaged-based finite element (FE) meso-model.
1. INTRODUCTION
From a physical point of view, concrete is not a homogeneous material, as it appears to be on a structural level.
It can be rather considered as a multiphase and multiscale geomaterial exhibiting different characteristics at
different observation scales. Due to this heterogeneous multiscale nature, the failure processes in concrete are
complicated and still constitute a very active area of research.
In meso-scale it can be considered as a three-phase heterogeneous geomaterial consisted of aggregates, cement
paste and macro-pores. Modelling concrete at this scale requires a reliable characterization of its geometrical and
topological features. X-ray CT can be a really suitable tool for this particular case, providing non-destructively
a 3D image of the internal material’s structure.
Although being a highly innovative technique, experimental data coming from x-ray scans require primarily
a series of image analysis operations for quantitative and qualitative information to be extracted. Specifically,
due to the physics of x-rays, this can be quite rigorous when the individual materials (phases) consisting the
scanned object have similar density values. This exactly happens to be the case of concrete’s mesostructure:
aggregates embedded into cement paste, adding a considerable technical challenge to the whole procedure.
This technical challenge will be the principal issue addressed on this work: how to get a reliable morphological
representation of the mesostrucure of a heterogeneous cementitious material which can be later used as an input
for a realistic imaged-based FE meso-model.
2. IMAGE ANALYSIS
On a representative horizontal slice of a reconstructed 3D image of concrete, the three phases (aggregates,
cement paste, macro-pores) can be easily distinguished by a human eye. However, pixels (or voxels in 3D)
belonging to each phase don’t share the same value, rather they have a range of grey values corresponding to
X-ray attenuation coefficients of each material. This means that the separation of each phase is essential to
simplify the representation of the image into something more meaningful and easier to analyse: a trinary 3D
image where each phase is strictly composed of a set of voxels sharing a pre-decided value representing that
phase.
The separation of the void from the solid phase can be straightforward due to their obvious density contrast
directly depicted on the reconstructed image. The separation of the solid phase into cement paste and aggregates
is more complicated due to their similar densities, resulting in close grey levels. However, since the cement paste
is (by nature) a more heterogeneous material compared to the aggregates, the variation of grey values should be
more irregular. Thus, instead of separating the solid phase based on the absolute grey values, a separation based
∗ e-mail: [email protected]
† e-mail: [email protected]
‡ e-mail: [email protected]
on the variation of the grey values is the approach followed in this work. A sequence of morphological operations
and filters using structuring elements is applied so as to end up with a fairly well solid phase segmentation.
Generally, the whole procedure is based on specifically developed during this work Python functions, avoiding
any existing image analysis software. The generated functions are combined to compose a general procedure for
getting a concrete’s trinary 3D image, developed in an accessible and open source way to encourage auditability
and further developments.
For the reliability of the presented segmentation procedure, a representative set of aggregates, used for
concrete mix, is embedded into a gel (representing the cement paste) and scanned. The obvious density contrast
between gel and aggregates leads to a straightforward segmentation. A casting of a real concrete using the same
aggregates follows and a new x-ray scan is performed. Each step of the presented segmentation procedure is
applied on the real concrete scan and the resulted aggregate phases are compared.
As a further validation step, a neutron scan is perfomed to the real concrete, where due the the physics of
neutrons, the segmentation of the solid phase is much easier compared to that of the x-rays scan.
3. RESULTS AND APPLICATIONS
Once each phase of concrete’s mesostructure is isolated, a quantitative analysis can be performed. Individual
aggregates and pores are reliably identified and characterised through a set of measurements. The pore and
aggregate size distribution are calculated. A comparison between the aggregate’s size distribution curve (used
for the concrete mix) and the ones calculated from the x-ray and neutron scans leads to a validation of the
presented segmentation procedure.
Having obtained a reliable characterization of concrete’s mesostructure heterogeneities, the 3D trinary image
is used to build a morphologically realistic imaged-based FE model. Mechanical applications (uniaxial tension
and compression tests) are performed so as the main characteristics of the quasi-brittle material macroscopic
behaviour to be reproduced.
As a further step for validating and enriching the FE model, uniaxial tensile and compressive tests are
perfomed inside the x-ray scanner. An experimental set-up is developed compatible with the x-ray apparatus
and the in-situ evolution of the microstructure is followed during the tests and compared with the numerical
simulations.
References
[1]
(a) (b) (c)
Figure 1: (a) Obtained trinary 3D image of concrete (blue: macro-pores, grey: aggregates, cement paste
occupies the remaining volume) (b) Heterogeneous morphology of concrete mesostructure used in FE simulations
(c) Macrospopic crack patterns in tensile numerical tests

Experimental In-Situ Tests and Image Processing For A Realistic Image-Based FE Meso-Model For Cementitious Materials

Uploaded by

Copyright:

Available Formats

Experimental In-Situ Tests and Image Processing For A Realistic Image-Based FE Meso-Model For Cementitious Materials

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Experimental In-Situ Tests and Image Processing For A Realistic Image-Based FE Meso-Model For Cementitious Materials

Uploaded by

Copyright:

Available Formats

Experimental in-situ tests and image processing for a realistic

image-based FE meso-model for cementitious materials

Keywords: x-ray tomography, image processing, cementitious materials, meso-scale FE modelling

3. RESULTS AND APPLICATIONS

(a) (b) (c)

You might also like