SOLUTION OF SOIL-STRUCTURE INTERACTION PROBLEMS IN GPU ENVIRONMENTS

Abstract eng:
Soil-structure interaction (SSI) phenomena are of crucial importance when considering a structure's response to an earthquake excitation. On top of modeling the actual structure as accurately as possible, it is imperative to take into consideration the dynamic characteristics of the underlying soil. In order to achieve that, a multiphase view of the soil is upheld, resulting to a coupled problem formulation. In such a formulation, the monolithic approach where all field equations are solved simultaneously, is considered as the most reliable one [3,4] and is widely adopted in geotechnical problems, seismic or wave loading of soil-structure interaction problems and draining systems in landfills and reservoirs. In order to tackle such large-scale problems, graphics processing units (GPUs) are considered. GPUs and their programming have attracted the attention of the scientific community due to their inherently remarkable performance features on matrix and vector operations. Moreover, recent advances in GPU technology have provided lower cost, higher performance, enhanced computational precision and improved programming tools, widening their adoption on solving computationally expensive, large algebraic systems occurring from scientific problems in various fields. In this work, a large-scale SSI problem featuring a detailed concrete structure founded on a 3-layered nonlinear soil is considered, solved by a cluster of GPUs using domain decomposition techniques and showcasing the performance and scalability of the aforementioned techniques.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket