STRUCTURAL GLOBAL SENSITIVITY METHOD BASED ON PARTIAL DERIVATIVE WHOLE DOMAIN INTEGRAL

In view of the problems that the traditional variance-based Sobol’method had a low solving efficiency,lacked of enough robustness,and it cannot further effectively decompose and reasonably distribute the influences of the high-order cross subterms,a practical and effective structural global sensitiv...

Full description

Saved in:
Bibliographic Details
Main Authors: TU LongWei, LIU Jie, LIU GuangZhao, ZHANG Zheng
Format: Article
Language:zho
Published: Editorial Office of Journal of Mechanical Strength 2019-01-01
Series:Jixie qiangdu
Subjects:
Online Access:http://www.jxqd.net.cn/thesisDetails#10.16579/j.issn.1001.9669.2019.06.015
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In view of the problems that the traditional variance-based Sobol’method had a low solving efficiency,lacked of enough robustness,and it cannot further effectively decompose and reasonably distribute the influences of the high-order cross subterms,a practical and effective structural global sensitivity method was proposed in this paper based on partial derivative whole domain integral and optimal polynomial surrogate model. Firstly,optimal surrogate model was constructed through polynomial structure-selection,which had good fitting and predictive ability,and it was convenient for direct integral operations. Then,local sensitivity method based on partial derivative was extended to a global sensitivity method by integrating partial derivatives of model variables in variable sapces. In addition,the paper redefined a more conveniently calculated sensitivity indice that can achieve effective decomposition for the high-order sensitivity indices,and the sensitivity results directly corresponded to model variables without the high-order indices,which had more practical engineering significance. Numerical example 1 shows the deficiency of Sobol’total sensitivity indices in application. Numerical example 2 illustrates the validity of the proposed method for complex high-dimensional model. Engineering example demonstrates the applicability and effectiveness of the present method for complex engineering structure problems.
ISSN:1001-9669