It is quite frequent in chemical kinetics, in combustion studies or in meteorology, to use large computational models that depend - generally in a non-linear fashion - on a certain number of "parameters", that is quantities, fixed during the model calculation, that allow to modify the model behaviour in order to fit it to the specific problem studied. Often these parameters are not well known (typically, cross sections or rate constants in chemical kinetics). When the number of parameters to be fitted is low (less than three or four, for example), generally it is relatively easy to search for the better values by simply varying the parameters in some ordered way and comparing the obtained results with experimental knowledge of the investigated phenomenon. But that is impossible to do if the parameters to be varied are tenths, because of the huge computational effort and memory storage capacity requested, which grow as a power of the number of varied parameters. In this case, a workable strategy is to reduce as much as possible the number of parameters to be adjusted, leaving fixed the "less important" ones, that is those parameters not largely affecting the output of the model, when restricted to specified intervals. Sensitivity analysis is a tool to know the "importance" of each parameter, once its variation interval has been assigned (this last could be the parameter incertitude around the experimental value, or the allowed range of variation of a controllable quantity).
APPLICATION OF MONTE CARLO SENSITIVITY ANALYSIS TO ELECTRON·MOLECULE CROSS SECTIONS IN NITROGEN DISCHARGES
F Esposito;
1996
Abstract
It is quite frequent in chemical kinetics, in combustion studies or in meteorology, to use large computational models that depend - generally in a non-linear fashion - on a certain number of "parameters", that is quantities, fixed during the model calculation, that allow to modify the model behaviour in order to fit it to the specific problem studied. Often these parameters are not well known (typically, cross sections or rate constants in chemical kinetics). When the number of parameters to be fitted is low (less than three or four, for example), generally it is relatively easy to search for the better values by simply varying the parameters in some ordered way and comparing the obtained results with experimental knowledge of the investigated phenomenon. But that is impossible to do if the parameters to be varied are tenths, because of the huge computational effort and memory storage capacity requested, which grow as a power of the number of varied parameters. In this case, a workable strategy is to reduce as much as possible the number of parameters to be adjusted, leaving fixed the "less important" ones, that is those parameters not largely affecting the output of the model, when restricted to specified intervals. Sensitivity analysis is a tool to know the "importance" of each parameter, once its variation interval has been assigned (this last could be the parameter incertitude around the experimental value, or the allowed range of variation of a controllable quantity).I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
