The present work is dedicated to implement an uncertainty quantification tool in MATLAB using the non-intrusive polynomial chaos expansion method and to examine the tool performance with an application test case. Given uniformly or normally distributed uncertain parameters, the tool computes the output quantities of interest. The test case is based on data of a recent scientific paper in the field of uncertainty quantification of thermoacoustic instabilities. The tool results for analytic moments of the uncertain quantities were compared to the results from the paper, where authors were using adjoints and Monte Carlo simulation. It was shown that the outputs of both methods are very comparable, the higher the order - the lower the difference, while the computation time was reduced. The tool results for local sensitivities were verified using the finite difference method. The results were still comparable, however, it was found that increase in polynomial chaos expansion order could lead to higher local fluctuations of the response function (approximation). The reduction of computation time and reliable results show, that the tool can be used instead of Monte Carlos simulation for small numbers of input parameters. Furthermore, this tool is the only choice to conduct an uncertainty quantification study, if a single function evaluation takes a lot of time as in CFD simulation.     «

The present work is dedicated to implement an uncertainty quantification tool in MATLAB using the non-intrusive polynomial chaos expansion method and to examine the tool performance with an application test case. Given uniformly or normally distributed uncertain parameters, the tool computes the output quantities of interest. The test case is based on data of a recent scientific paper in the field of uncertainty quantification of thermoacoustic instabilities. The tool results for analytic moment...     »

The present work is dedicated to implement an uncertainty quantification tool in MATLAB using the non-intrusive polynomial chaos expansion method and to examine the tool performance with an application test case. Given uniformly or normally distributed uncertain parameters, the tool computes the output quantities of interest. The test case is based on data of a recent scientific paper in the field of uncertainty quantification of thermoacoustic instabilities. The tool results for analytic moments of the uncertain quantities were compared to the results from the paper, where authors were using adjoints and Monte Carlo simulation. It was shown that the outputs of both methods are very comparable, the higher the order - the lower the difference, while the computation time was reduced. The tool results for local sensitivities were verified using the finite difference method. The results were still comparable, however, it was found that increase in polynomial chaos expansion order could lead to higher local fluctuations of the response function (approximation). The reduction of computation time and reliable results show, that the tool can be used instead of Monte Carlos simulation for small numbers of input parameters. Furthermore, this tool is the only choice to conduct an uncertainty quantification study, if a single function evaluation takes a lot of time as in CFD simulation.     «

The present work is dedicated to implement an uncertainty quantification tool in MATLAB using the non-intrusive polynomial chaos expansion method and to examine the tool performance with an application test case. Given uniformly or normally distributed uncertain parameters, the tool computes the output quantities of interest. The test case is based on data of a recent scientific paper in the field of uncertainty quantification of thermoacoustic instabilities. The tool results for analytic moment...     »