Characterization of the importance of mechanical properties using Sobol indices in composite structures

Authors: Luísa Natália Hoffbauer, Carlos Conceição António

ABSTRACT:

In order to study the problem of uncertainties in composite structures, an approach based on the optimal design for a specific level of reliability is proposed, corresponding to a maximum load applied to the structure. This scenario corresponds to an inverse reliability problem for the mean values of mechanical properties. To rank these properties, scattered points are generated in a domain centered on their mean values, using the uniform design method (UDM). The Tsai number for the most probable failure, the reliability index and the relative sensitivities for each point considered in the UDM are obtained. Assuming these points as input / output patterns, an Artificial Neuronal Network is developed to study the propagation of uncertainties based on Sobol indices.

Key Words:
artificial neuronal network, composite structures, reliability, Sobol indexes, uncertainties, uniform design
Affiliation:

Luísa Natália Hoffbauer 1 | Carlos Conceição António2
1 INEGI/LAETA, Instituto Superior de Engenharia, P. Porto, Porto, Portugal
2 INEGI/LAETA, Faculdade de Engenharia, Universidade do Porto, Porto, Portugal