Efficient classification based methods for global sensitivity analysis

Reuter, Uwe; Mehmood, Zeeshan; Gebhardt, Clemens

doi:10.1016/j.compstruc.2012.06.007

Cited by 10 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In many real problems with sparse data, polynomial models are insufficient. In addition, regression coefficients of various responses may not be normalized and thus may require more sophisticated variance decomposition methods (Reuter et al, 2011).…”

Section: Methods and Techniquesmentioning

confidence: 99%

Analytical global sensitivity analysis with Gaussian processes

Srivastava

Subramaniyan

Wang

2017

AIEDAM

View full text Add to dashboard Cite

Methods for efficient variance-based global sensitivity analysis of complex high-dimensional problems are presented and compared. Variance decomposition methods rank inputs according to Sobol indices that can be computationally expensive to evaluate. Main and interaction effect Sobol indices can be computed analytically in the Kennedy and O'Hagan framework with Gaussian processes. These methods use the high-dimensional model representation concept for variance decomposition that presents a unique model representation when inputs are uncorrelated. However, when the inputs are correlated, multiple model representations may be possible leading to ambiguous sensitivity ranking with Sobol indices. In this work, we present the effect of input correlation on sensitivity analysis and discuss the methods presented by Li and Rabitz in the context of Kennedy and O'Hagan's framework with Gaussian processes. Results are demonstrated on simulated and real problems for correlated and uncorrelated inputs and demonstrate the utility of variance decomposition methods for sensitivity analysis.

show abstract

Section: Methods and Techniquesmentioning

confidence: 99%

Analytical global sensitivity analysis with Gaussian processes

Srivastava

Subramaniyan

Wang

2017

AIEDAM

View full text Add to dashboard Cite

show abstract

“…Li and Liu [7] researched a three-dimensional hybrid governing equation for the static response analysis and sensitivity analysis and showed that it can solve the specific problem efficiently. Another research direction has divided the large-scale problem into several smaller ones, and then can be solve easily, like the domain decomposition technique [8] and filtering method [9]. These methods can solve the large-scale optimization problem parallelly, but some interactive information will be lost, even the optimal solution.…”

Section: Introductionmentioning

confidence: 99%

An analytical sensitivity theory-based optimization method for composite structures with large-scale variables

Luo¹,

Guo²,

Nie³

2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

This research focusses on large-scale variable optimization techniques for commercial composite wing structures. To achieve a more complete set of results, the layer thickness for composite layers with the same material and ply direction, are set as a design variable in this study. An analytical approach was utilized to deal with the sensitivity analysis so as to study the relationship between displacement, stress and strain constraints. In order to improve the applicability and user-friendliness, the procedure of sensitivity analysis has been programmed within Fortran and integrated into a Computer-Aided Engineering (CAE) software named HAJIF. To compare the consistency and efficiency, the analytical sensitivity calculation method, along with the finite difference method and a semi-analytical method, were employed to optimise a composite aircraft wing structure. The analytical method was found to have the best consistency and efficiency amongst the three optimization methods. With the aid of HAJIF, the wing was optimized with a weight loss ratio of 8.9%. The total optimization time was is 56 times quicker than the finite difference method and semi-analytical method.

show abstract

“…The main idea is that only the level sets of the model response are identified and approximated instead of the whole model response [12]. The full approximation problem is thus reduced to multiple partial approximation problems for the level sets, requiring a relatively smaller number of sample points.…”

Section: Classification Based Global Sensitivity Analysismentioning

confidence: 99%

“…A change in class denotes a change of the model response, i. e., the (normalized) counter represents a sensitivity measure for the corresponding parameter. This procedure is termed as classification based sensitivity analysis and is explained in detail in [12]. …”

Section: Classification Based Global Sensitivity Analysismentioning

confidence: 99%

Global sensitivity analysis for the efficient solution of optimization problem in design process

Reuter

Mehmood

Liebscher³

2012

Proc Appl Math and Mech

Self Cite

View full text Add to dashboard Cite

Application of the methods of global sensitivity analysis in design process arises due to the complexity associated with structural optimization, e. g., in the automobile industry for an optimal vehicle structure for crash load cases. Usually, thousands of variants are considered for a vehicle design project and each variant represents a combination of a number of model parameters. The computation time for a single variant ranges from several hours to days. To reduce the complexity of the optimization problem, the significant parameters should be identified but possibly with a very few sample points. Since most of the available global sensitivity methods require a good approximation of the model response, a relatively large number of sample points is usually required for the surrogate model. In this contribution, an overview of state-of-the-art methods of global sensitivity analysis is given and an alternative idea is proposed to avoid full approximation of the model response. Instead, only the approximation of the level sets of the model response is used to derive sensitivity measures.

show abstract

Efficient classification based methods for global sensitivity analysis

Cited by 10 publications

References 12 publications

Analytical global sensitivity analysis with Gaussian processes

Analytical global sensitivity analysis with Gaussian processes

An analytical sensitivity theory-based optimization method for composite structures with large-scale variables

Global sensitivity analysis for the efficient solution of optimization problem in design process

Contact Info

Product

Resources

About