Method of Selecting Active Parameters Using Sensitivity Analysis and Linear Programming

Abstract

ABSTRACT It is necessary to select appropriate active parameters to ensure both the accuracy and computation efficiency before the global analysis of chemical kinetic model. This paper proposes a new method for selecting the active parameters on the base of the combination of sensitivity analysis and linear programming. Compared with the usual methods for selecting active parameters, such as the local sensitive analysis, the characteristics of the proposed method is preliminary visualization of the possible influence of the selected active parameters on the model outputs in the process of parameter selection, ensuring the reliability of the selected active parameters. Considering the computation efficiency, the number of selected active parameters can be controlled in a suitable size through combining with dichotomy or other screening techniques. In the study, the pre-exponential factors of the Arrhenius equations in the USC-Mech II model were considered as the candidate parameters and the uncertainties of the pre-exponential factors were set. Taking the ignition of ethylene for example, the 10 reactions that can increase the ignition time of ethylene under a wide range of conditions with equivalence ratio of 1, 0.1 ~ 1 MPa and 1000 K ~ 1500 K were successfully selected using the proposed method. Then, the 10 active parameters were tested in each condition. The results showed that the selected active parameters can make the ignition delay time close to the target for each condition, which reflects the reliability of active parameter selection.