Framework for adaptive multi-scale simulation of textile reinforced concrete
Peiffer, Frank; Meskouris, Konstantin (Thesis advisor)
Aachen : Publikationsserver der RWTH Aachen University (2009)
Dissertation / PhD Thesis
Composite materials have been receiving growing attention in recent years. In a composite, material components are combined in a way that their respective strengths are optimally employed. The work at hand has originated from the research on textile reinforced concrete (TRC) that uses high strength textiles embedded in a fine-grained cementitious matrix to compensate for its low tensile strength. Moreover, this combination of materials leads to quasi-ductile tensile behavior as required for load carrying structural elements. The thorough understanding and simulation of the mechanical behavior of composites is still a demanding task. In TRC both the reinforcement and the concrete matrix exhibit heterogeneity on similar scales of resolution. The failure of matrix, reinforcement or bond causes a localization of interacting damage mechanisms. The available micro- and meso-scale models regard these effects but the implied numerical effort restricts their applicability to the simulation of only small parts of a structure. Macro-scale models disregard the complex damage mechanisms and are limited to problems with homogeneous or periodic stress and strain fields. The work at hand proposes the adaptive multi-scale simulation of TRC for an efficient solution of more general problems (e.g. boundary effects or shear zones). The method demands for new concepts both in the field of modeling and in the design of suitable simulation software. A multi-scale modeling approach is presented that allows for adaptive enrichment of an initially coarse model. It accounts both for the discontinuity in the matrix field and for the local effect of debonding. For efficiency reasons the material structure is resolved only where needed, namely in the vicinity of the crack bridges and in the boundary zones. For the sake of explanation simplicity, the modeling approach is exemplified for the one-dimensional case. No significant enhancements for two and three dimensional cases should be necessary. The triggering event for the adaptive model refinement is the development of a crack. The time stepping algorithm must monitor the material state, scale the time step appropriately and then modify the discretization. It is shown that a thoughtless implementation of the adaptive rules within the time stepping algorithm breaks the principles of information hiding and encapsulation. Instead, an adaptive time-stepping framework is presented that allows for the structured extension of object-oriented finite element code with additional adaptive features. The design takes into account not only the requirements of the TRC multi-scale simulation but also the demands of several other adaptive applications. The concept of adaptive strategies is introduced that aims at encapsulating the application specific adaptive features and thus keeps the time stepping algorithm clear of special purpose code. The realization of the target applications within the presented framework demonstrates the applicability of the chosen design in various contexts.