Adam Sciegaj (foto Marcus Folino)
Populär beskrivning (engelska)
Have you ever noticed how reinforced concrete cracks in buildings and engineering structures around us? This is normal and usually not dangerous, as the reinforcement prevents the cracks from growing too much. However, these cracks open up the inside of the structure for potentially harmful substances, which can cause corrosion of the reinforcement. This negatively affects the durability of the structure and is highly undesired from the sustainability point of view. Unfortunately, we cannot totally prevent cracking. We would therefore like to be able to model the cracking process, to be able to predict and control crack widths.
Even though the actual physical phenomena involved in the cracking process are quite complicated, models exist which can give us accurate predictions. These models simulate what's happening to the material when forces, like e.g., gravity or traffic loads, act on it. In practice, we create computer models of the engineering structures we want to analyse. To facilitate the computations, the computer model is divided into small pieces called finite elements. Cracks can have lengths in the order of decimetres, and are thus much smaller than the structure, which usually ranges from tens to hundreds of metres. In terms of crack modelling, this means that the finite elements must also be very small, which results in very large computer models requiring a long time to produce results. Fortunately, there exist multiscale modelling techniques, which are able to provide detailed small-scale results even if the structure is modelled with fairly large finite elements.
In this thesis, steps are taken to extend the existing multiscale modelling techniques to reinforced concrete structures. This way, detailed results such as crack widths and patterns can be obtained even for very large structures such as bridges or nuclear reactor containment buildings. More specifically, this is achieved by analysing the material response at different length scales, and connecting these scales to each other in an appropriate way. Additionally, thanks to parallel computing, the methods proposed in this thesis can potentially shorten the time it takes to analyse reinforced concrete structures with computer models.