Sensitivity analysis of Engineering Structures Utilizing Artificial Neural Networks and Polynomial Chaos Expansion

Sensitivity analysis of Engineering Structures Utilizing Artificial Neural Networks and Polynomial Chaos Expansion

Stať ve sborníku v databázi WoS či Scopus

This paper is focused on sensitivity analysis of engineering structures using surrogate models. Two different techniques for surrogate modeling are utilized in order to obtain various sensitivity measures of quantity of interest. The artificial neural networks and polynomial chaos expansion are used for efficient sensitivity analysis. Each of the techniques is superior in different areas of uncertainty quantification and thus each of them is used for estimating of different sensitivity measures in two engineering examples – simplified analytical function and complex non-linear finite element model of an existing concrete bridge. On the one hand, artificial neural network is utilized for estimation of sensitivity measures based on Monte Carlo simulation and on the other hand, polynomial chaos expansion is exploited for derivation of global sensitivity measures without additional simulations. It is shown that utilization of both methods leads to efficient and complex sensitivity analysis of engineering structures, and it could be recommended to use combination of both techniques in industrial applications.

This paper is focused on sensitivity analysis of engineering structures using surrogate models. Two different techniques for surrogate modeling are utilized in order to obtain various sensitivity measures of quantity of interest. The artificial neural networks and polynomial chaos expansion are used for efficient sensitivity analysis. Each of the techniques is superior in different areas of uncertainty quantification and thus each of them is used for estimating of different sensitivity measures in two engineering examples – simplified analytical function and complex non-linear finite element model of an existing concrete bridge. On the one hand, artificial neural network is utilized for estimation of sensitivity measures based on Monte Carlo simulation and on the other hand, polynomial chaos expansion is exploited for derivation of global sensitivity measures without additional simulations. It is shown that utilization of both methods leads to efficient and complex sensitivity analysis of engineering structures, and it could be recommended to use combination of both techniques in industrial applications.

Artificial neural network; Polynomial chaos expansion; Sensitivity analysis; Uncertainty quantification

Artificial neural network; Polynomial chaos expansion; Sensitivity analysis; Uncertainty quantification

NOVÁK, L.; LEHKÝ, D.; NOVÁK, D.

2024

09.03.2023

Springer Science and Business Media Deutschland GmbH

Germany

9783031255984

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

1

15

15