Along with the development and application of computer codes, it has been increasingly recognized that the corresponding computational results are associated with uncertainty due to lack of knowledge on various sources. Therefore, uncertainty and sensitivity analyses are performed to get (1) a quantification of the combined influence of many of these uncertainty sources and (2) a ranking of the individual sources according to their contribution to the uncertainty of the results.
The software tool SUSA (Software for Uncertainty and Sensitivity Analyses) guides through the main steps of a probabilistic uncertainty and sensitivity analysis. These steps can be summarized as follows:
1. Identification of all phenomena, modeling assumptions, and parameters that are potentially important contributors to the uncertainty of the computational result and representation of all uncertainty sources by uncertain parameters.
2. Quantification of the state of knowledge on the uncertain parameters in terms of probability distributions and dependence measures.
3. Generation of a sample of values for the uncertain parameters according to a multivariate probability distribution which satisfies the input given in step 2.
4. Performance of computer code runs for each set of values sampled for the uncertain parameters -> random sample from the unknown probability distribution of the computational result.
5. Quantification of the uncertainty of the computational result on the basis of the sample resulting from step 4.
6. Ranking of the parameters with respect to their contribution to the overall uncertainty of the computational result (Sensitivity Analysis).
7. Comprehensive documentation of the analysis steps for scrutinizing the analysis results.
SUSA combines well established concepts and tools from probability calculus and statistics with a comfortable menu-driven user interface. All data transfers, statistical data analyses and graphical representations are performed automatically. SUSA even provides a comprehensive documentation of the analysis steps for scrutinizing the analysis output.
Various types of probability distributions and dependence structures are available for modeling uncertainty probabilistically. For sample generation, the simple random sampling or the Latin Hypercube sampling procedure may be applied. In principal, the runs of any computer code can be started from within SUSA. SUSA can even provide input decks of complex computer codes each accounting for a different set of parameter values. Diverse alternatives exist for quantifying the uncertainty of the computational results (e.g. tolerance limits) and their sensitivity with respect to the individual uncertainty sources (e.g. correlations, regression based measures or correlation ratios).
SUSA is a PC-software running under Windows (NT, XP or Windows 7). The current version 3.6 requires the installation of MS Excel 2003, 2007, or 2010. It can be used for any simple or complex application. There are, in principal, no limitations concerning the number of uncertain parameters, output quantities, and computer code runs.