Parallel simulation of reinforced concrete sructures using peridynamics

The failure of concrete structures involves many complex mechanisms. Traditional theoretical models are limited to specific problems and are not applicable to many real-life problems. Consequently, design specifications mostly rely on empirical equations derived from laboratory tests at the component level. It is desirable to develop new analysis methods, capable of harnessing material-level test parameters. To overcome limitations and shortcomings of models based on continuum mechanics and fracture mechanics, Stewart Silling introduced the concept of peridynamics in 1998. Similar to molecular dynamics, peridynamic modeling of a physical structure involves simulating interacting particles subjected to an empirical force field. The evolution of interacting particles determines the deformation of the structure at a given time due to the applied boundary condition. As a particle-based model, peridynamics requires the repeated evaluation of many particle interactions which is computationally demanding. However, with todays inexpensive computing hardware, parallel algorithms can be utilized to run such problems on multi-node supercomputers with fast interconnects. However, existing codes tend to be domain-specific with too many built-in physical assumptions. In this work, a novel method for parallelization of any particle-based simulation is presented which is quite general and suitable for simulating diverse physical structures. A scalable parallel code for molecular dynamics and peridynamics simulation, PDQ, is described which implements a novel wall method parallelization algorithm, developed as part of this thesis. PDQ partitions the geometric domain of a problem across multi-nodes while the physics is left open to the user to decide whether to simulate a solvated protein or alloy grain boundary at the atomic scale or to simulate cracking phenomena in concrete via peridynamics. A further extension of PDQ brings more flexibility by allowing the user to define any desired number of degrees of freedom for each particle in a peridynamics simulation. At the end of this thesis, plain, reinforced and prestressed concrete benchmark problems are simulated using PDQ and the results are compared to available design code equations or analytical solutions. This research is a step toward next level of computational modeling of reinforced concrete structures and the revolutionizing of how concrete is analyzed and also how concrete structures are designed.\u2