Proceedings of the International rilem conference Materials, Systems and Structures in Civil Engineering 2016

115

International RILEM Conference on Materials, Systems and Structures in Civil Engineering 

Conference segment on Service Life of Cement-Based Materials and Structures 

22-24 August 2016, Technical University of Denmark, Lyngby, Denmark 

 

THE IMPORTANCE OF MULTIPHYSICS AND MULTISCALE 

MODELLING OF CONCRETE TO UNDERSTAND ITS COMPLEX  

MACROSCOPIC PROPERTIES 

 

Jörg F. Unger 

(1)

, Vitaliy Kindrachuk 

(1)

, Volker Hirthammer 

(1)

, Thomas Titscher 

(1)

, 

Christoph Pohl 

(1)

 

 

(1) Federal Institute for Materials and Structures, Berlin, Germany 

 

 

 

 

 

Abstract 

Concrete is a complex material. Its properties evolve over time, especially at early age, and 

are dependent on environmental conditions, i.e. temperature and moisture conditions, as well 

as the composition of the material. This leads to a variety of macroscopic phenomena such as 

hydration/solidification/hardening, creep and shrinkage, thermal strains, damage and inelastic 

deformations. Most of these phenomena are characterized by specific set of model 

assumptions and often an additive decomposition of strains into elastic, plastic, shrinkage and 

creep components is performed. Each of these phenomena are investigated separately and a 

number of respective independent models have been designed. The interactions are then 

accounted for by adding appropriate correction factors or additional models for the particular 

interaction. This paper discusses the importance of reconsider even in the experimental phase 

the model assumptions required to generalize the experimental data into models used in 

design codes. It is especially underlined that the complex macroscopic behaviour of concrete 

is strongly influenced by its multiscale and multiphyscis nature and two examples (shrinkage 

and fatigue) of interacting phenomena are discussed. 

 

 

1. Introduction 

 

The understanding and the prediction of the macroscopic behaviour of concrete is very 

important to ensure a safe design of structural components. In the design phase, it is often 

very difficult and certainly not appropriate to use complex models with a multiscale and 

multiphysics approach. The focus is in many cases a safe design for the ultimate limite state.  

However, a profound understanding of the material behaviour with a realistic modelling of 

hardening and the interacting phenomena is required to understand the behaviour under 

service conditions or the evaluation of a structure after extraordinary loading situations. This 

ranges from the detection of microcracking, crack spacing and crack opening, creep and 

shrinkage deformations up to the influence of high temperatures or fatigue loading. In this 

---

116

International RILEM Conference on Materials, Systems and Structures in Civil Engineering 

Conference segment on Service Life of Cement-Based Materials and Structures 

22-24 August 2016, Technical University of Denmark, Lyngby, Denmark 

 

case, the experimental investigations of interacting phenomena are usually performed with a 

specific modelling assumption, e.g. a linear decomposition of the strain into elastic, plastic, 

shrinkage and creep components. Interactions are then taken into account by additional 

interaction terms. 

An example is concrete under combined thermal and mechanical loads, where often an 

additive decomposition into mechanical and thermal strains is done. The mechanical model is 

calibrated for constant room temperature, whereas the thermal model is adjusted for zero 

mechanical load. The (nonlinear) interaction is then characterized by so called load induced 

thermal strains [1].  

Even though this approach seems to be straightforward in general, it has several drawbacks 

 

these interaction terms have to be calibrated separately and only model the real 

physics in a homogenized, macroscopic way; 

 

it is often difficult to design experiments, where only a single parameter is changed 

and calibrated, and the distribution of this parameter is often not fully homogeneous 

within the specimen; 

 

the predictive capacities of models with interaction factors are often limited to the set 

of training/validation data, since the interaction model is purely phenomenological and 

does not capture the real physics; 

 

the consideration of additional inputs requires a recalibration of the model, i.e. the 

heating of concrete does not only change the temperature, but also induces drying 

which itself leads to shrinkage phenomena. So depending on whether shrinkage is 

considered as additional (additive) strain, the interaction term varies; 

 

the number of parameters for those models is often large and particularly the 

interaction terms often do not have a clear physical meaning; 

 

the thermodynamic consistency of the model is sometimes difficult to ensure, e.g. 

when modelling a series of loading scenarios with a parallel temporal evolution of 

macroscopic properties such as strength, Young’s modulus or fracture energy, which 

is especially important at early age. 

As a consequence, the choice of the right model assumptions is very important, especially 

when dealing with multi-physics problems. Two models might give very similar results, e.g. a 

damage or softening plasticity model for monotonic loading. However, both models will give 

significantly different results for general loading sequences including unloading. A similar 

example is the modelling of shrinkage in concrete based either on a strain or stress based 

approach. Pure shrinkage deformations can be modelled appropriately with both approaches, 

but it will be demonstrated that combined loading (mechanical, thermal, moisture content) 

might lead to significantly different results. 

A second reason for the macroscopic complex behaviour of concrete is the importance of its 

heterogeneous structure. This comprises very different scales. Some (non exhaustive) 

examples are 

 

atomistic models  

 

the influence of moisture on macroscopic creep behaviour is often explained by 

the disjoining pressure due to very thin layers of water; 

 

microscale models 

 

micro-cracks lead to a macroscopic softening with a quasi-brittle post-peak 

---