Authors:
Riccardo Tosi | Centro Internacional de Métodos Numéricos en Ingeniería (CIMNE) | Spain
Marc Nunez | Centro Internacional de Métodos Numéricos en Ingeniería (CIMNE) | Spain
Brendan Keith | Technical University of Munich | Germany
Barbara Wohlmuth | Technical University of Munich | Germany
Jordi Pons-Prats | Centro Internacional de Métodos Numéricos en Ingeniería (CIMNE) | Spain
Riccardo Rossi | Centro Internacional de Métodos Numéricos en Ingeniería (CIMNE) | Spain
Nowadays, we are facing the increasing necessity of dealing with uncertainties in many different engineering fields, together with the constant development of computational capabilities. Current solvers must satisfy both statistical convergence and computational efficiency; this talk aims to explore their integration. When running on supercomputers, the main issue one should consider is how to avoid wasting resources by leaving the machine empty. The proposed framework avoids the synchronization point bottleneck by adding a new level of parallelism, between batches. Additionally, the reliability is guaranteed through CVaR estimation of the Quantity of Interest, and the dynamic scheduling of all independent tasks is provided. We will then present and analyze the application of the framework to wind engineering problems, specifically focusing on the variability estimation of relevant parameters, such as the building base moment and pressure coefficient.
This work has been supported by the European Commission through the H2020 Research and Innovation program under contract 800898.
[1] Dadvand, P., Rossi, R., & Oñate, E. (2010). An object-oriented environment for developing finite element codes for multi-disciplinary applications. Arch Comput Methods Eng, 17(3), 253–297.
[2] Pisaroni, M., Nobile, F., & Leyland, P. (2017). A Continuation Multi Level Monte Carlo (C-MLMC) method for uncertainty quantification in compressible inviscid aerodynamics. Comput Method Appl M, 326, 20–50.