IF MULTISCALE --> THEN MULTIDISCIPLINARY?

It's almost two years that I decide how to model the phenomenon of ASR in concrete structures. My choice is a multiscale material model based on poromechanics theory developed by Coussy, Dormieux, Ulm and further applied by Lemarchand and Charpin.

I like a lot this approach because you can bridge scales, the equations can be solved analytically and it proved to describe properly the mechanical behavior of concrete.

But today while reading the book titled "Mechanics and Physics of Porous Solid", I was asking to myself if it is indeed necessary to have a multidisciplinary approach to model this problem.

Can ASR in concrete be explained considering only mechanics? or do we need physics and chemistry too?

I don't have any answer at the moment, if you find one just let me know :)

Apparently, by crossing multiple scales you need to learn multiple disciplines!

Modeling ASR in concrete structures, it looks like finishing a puzzle. (picture from biotuesdays.com)

Modelling efforts and control over realistic data

I have been recently busy with trying to improve damage model by implementing better ways to represent variety in material definitions. After finishing the analysis of NanoIndent results now we have the opportunity to assign this information (basically E-modulus and Strength values driven from hardness) for different phases in the system.

I recently introduced

• Random distribution based on Nano Indent test results - using mean and variance of data
• Triangular random generation of mesh with around 112000 elements to represent 10cm x 10cm sections
• Probabilistic distribution of reaction (in this case expansion) based on petrographic identification of reactive phases in aggregates (different for each mix design)
• and implementation of saturated cracks by updating the loading system while simulation is running (by introducing plastic behavior stage non elements stress strain curves)

Here are some representative images;

Histogram showing distribution of E values grouped by phases (Agg-Mortar-Bond-Void)

Heat map of E Mod values based on Digital image, Mesh geometry representing heterogeneity in aggregates and mortar

Initial Load elements with varied Intensities

Fracture evolution after 20000 simulation steps

Timing of Damage - Cutoff level in simulation results

Lattice simulations are capable of simulating crack pattern due to various loading modes. Either Aggregate, ITZ or paste (open cracks). As simulations run till the system failure (or predeifned failure threshold strain) there is a challenge on deciding on the cut-off level of damage progress in 2D system. Thanks to help from my colleague Branko Savija, now we introduced a itterative way to calculate total crack opening on every step. This approach allows us to compare damaged images (with known duration, environmental conditions (accelerated test) mix properties and aggregate geometry distribution) with threshold cracks (based on pixel counts) with relative damage in the system of lattice beams. Total crack width calculations shall be smooted for outliers but comparison graphs look like tese:

Sample 1: Core from Nautesund Bridge S31-3
Simulations are based on loading modes; AGG, ITZ, AGG+ITZ, Paste (20% randomly distributed). Based on image analyses cut-off line is 6%

Sample 2: Cube from accelerated tests (60°C) duration 140 days - original mix design from Nautesund Bridge (w/c=0.5) - No Alkali boosting (Expansion value: 0.24%)
Simulations are based on loading modes; AGG, ITZ, AGG+ITZ, Paste (20% randomly distributed). Based on image analyses cut-off line is 6%

Sample 3: Cube from standard accelerated tests (60°C) duration 140 days - original aggregates from Nautesund Bridge (w/c=0.5) - Based on standard grading curve - No alkali boosting. (Expansion value: 0.14%)
Simulations are based on loading modes; AGG, ITZ, AGG+ITZ, Paste (20% randomly distributed). Based on image analyses cut-off line is 10%

 This approach allows us indetify which loading modes progress faster. But there is still room for progress but initial results seems promising. Next step will be intentifying crack width growth by phase types and recalibration of material properties with Nanoindentation techniques.

SIMULATION: Crack formation under Aggregate Expansion

We have been trying to run simulations on various loading conditions and material properties to investigate crack formation in 2D sections. While we are busy with sample analyses and running simulations, here is a taster (if you may call it a result) video of how cracks grow under expanding aggregates. This simulation is based on a real test specimen (Upside down though).

Simulations continue till the specimen fails (or reaches a predefined stain level). You may realise some simplification on the section identification. Here are how the original sections look like. These specimens have been kept in 60°C reactor (RH 100%) for 20 weeks before sample preparation process.

 1- Normal Light Image

2- UV Light Image

Specimens are based on original mix design of Nautesund Bridge. you can find more information about this testing campaign here.

I have also done damage rating index analyses on this (and various other) specimens based on method defined by Grattan-bellew (1992). Some results regarding These analysis are as follows:

 Damage rating index - calculated based on Grattan Bellew (1992)

Distribution of Crack types observed on our samples

DRI analyses and simulations are continuing in full speed. Soon I will provide more data (and definitely more results) on our blog.

Busy months ahead..

Gel Expansion and Loading differences in model

Crack formation changes according to localization of the gel. this correlation brings a challenge in modelling. Which loading point will successfully represent Gel expansion?
Recently explored various loading options on a hypothetical section with a low aggregate-paste ratio. Different images represent various loading points, On; aggregate, ITZ, aggregate+ITZ, respectively.

Also comparative stress distribution after certain amount of crack growth

First 2D simulations

Using digital imaging from reacted structure samples, we managed to run a 2D simulation for ASR expansion (size of 9.6 cm x 9.6 cm) .

Using thresholding and edge detection images are convereted to binary images. Three Phases (aggregate - mortar - ITZ) identified and superimposed with a lattice mesh including (79600 beam elements). Based on earlier studies (Schlangen and Copuroglu, 2010) material properties are assigned and expansion is defined on all interfacial transition zone (ITZ) elements.

Cracked beam elements after 3000-8000-20000 steps.

First attempts assumed local lodings on the ITZ elements only and swelling to start simultaneously. Currently we are working on identification reactive components in a 2D medium and implementing non-simultaneous expansion patterns in the section. After execution of small scale testing device we will have more insights about pressure progress and gel propeties (those details for another post). Alternate reactive regions and shifted timing regimes to be defined later on.

Deformed version of sample after 20000 steps

Still a lot to do, in modeling but first results seems promising for improvement.



Schlangen, E. and C. Copuroglu (2010). Modeling of expansion and cracking due to ASR with 3D lattice model. Framcos7, Jeju, Korea.

Study outline

Concrete is a heterogeneous building material consisting of natural and man made products. The inherent heterogeneity of these material makes it hard to set a standard procedure for reactivity assessment. Common practice is to avoid reactive aggregates, where reactivity tests are either time consuming (impractical for engineers) or not reliable enough. Thus, there exists no internationally accepted ASR performance testing procedure; however studies on RILEM AAR-4 performance testing procedure is still in progress, yet not published. It is still a challenge to estimate the ASR effect on the existing structures together with potential impacts. There exist several recommendations, around the world, on ASR detection and prevention around the world. None, so far, has been internationally accepted.

In Netherlands, CUR regulations committee, CUR Technical Committee VC 62: Alkali Silica Reaction in concrete was set up in 2000. After severe criticism on CUR recommendation 38 (1994), the new committee published CUR recommendation 89 – ‘Measures to prevent damage to concrete by alkali-silica reaction (ASR)’ in 2002, which was followed by CUR recommendation 102 ‘Inspection and assessment of concrete structures in which the presence of ASR is suspected or has been established’. These recommendations are mostly provide decision trees suggest a ‘GO-NO GO’ decision.

A great number of studies, carried out worldwide, investigate the material performance considering various aspects of ASR effect. Most of these studies aim to determine the sole effect of ASR on micro-to-macro structural elements. Yet, there are missing links between chemical reaction parameters and gel product’s deleterious behaviour. Due to discontinuity of material properties in a mix, there may occur variations between laboratory test results and real structure measurements. A statistical approach to sample space data, harvesting available literature, can be used to increase the reliability of such a prediction model.

The aim of the project is to develop a material performance model for simulation of ASR effect on concrete durability. The final outcome of this project is an integrated support tool for engineers for ASR conscious design – a guideline for engineers.