Short Column (Sensitivity Analysis)

From COSSAN Wiki
Jump to: navigation, search

This page shows how to perform local and global sensitivity analysis of the Short Column.

Local Sensitivity

Analysis

This section demonstrates how the local sensitivity analysis can be performed. Begin by creating a physical model.


Sc sensitivity1.png

The analysis wizard starts by clicking on the green arrow on the top-right corner of the physical model editor. Select analysis type to be 'Sensitivity Analysis', as shown in the following figure:


Sc sensitivity2.png

Click next and the following wizard page should appear. Select the quantity of interest and the output of interest as shown below.


Sc sensitivity3.png

To finish, select the reference point and the method for estimate the local sensitivity analysis (add random variables by clicking the '+' button):


Sc sensitivity4.png

Results

The results of the local sensitivity analysis can be viewed in COSSAN-X by opening the folder 'Sensitivity Analysis' of the current project. Results of the performed sensitivity analysis (i.e. the LocalSensitivityMeasures) are contained in the EngineOutputs object.


Sc sensitivity5.png

Details of the object can be seen by selecting the object of interest. E.g. selecting LocalSensitivityMeasures reveals the description of the analysis and a summary of the object. Different means of presenting the data may be selected, such as a table, bar histogram etc.


Sc sensitivity6.png

Global Sensitivity

Analysis

Again create a physical model, and select the type of analysis to be 'Sensitivity Analysis'


Sc sensitivity7.png

Select quantity of interest to be 'Global Sensitivity Measures'


Sc sensitivity8.png

The next wizard page allows you to: 1) Select the input variable to be considered in the global sensitivity analysis, 2) Set the number of bootstrap resamples used to estimate the confidence intervals of the estimated quantities, 3) Set the number of samples used, 4) Select the algorithm to be used


Sc sensitivity9.png

  • The Random Balance Design is a very efficient method but allows to compute only the first order indices.
  • The Upper Bounds of the upper bounds are computed using Markov chain and the gradient of the function along the point of the chains.
  • The Sobol' indices are computed adopting the Saltelli method that allows to estimate the first order indices and the total indices.

Results

The results of the global sensitivity analysis can be viewed in COSSAN-X by opening the folder 'Sensitivity Analysis' of the current project and selecting the EngineOutputs object. Details of the object can be seen by selecting the object of interest, and different means of presenting the data, e.g. table, bar histogram etc., may be selected


Sc sensitivity10.png

See Also