Investigations Into The Model-based Combustion Kinetic Experimental Design

Computational fluid dynamic simulation s in combustion devices impose an increased demand on the predictive ability of combustion kinetic models.Model optimization is an important method to improve the predictive capability of combustion kinetic models,and experimental design methods can find the most valuable experimental conditions for model optimization.However,the current experimental design and model optimization methods involve m any mathematical methods.The high computational cost limits the application of these methods in complex systems such as large molecular models.This study aims to improve the model-based experimental design and model optimization methods to enhance the computational efficiency.Integrating all aspects of experimental design and constructing a systematic computational framework can promote the computational efficiency of experimental design and model optimization.The current work develops a logical framework for model-based experimental design and model optimization.An integrated software named Opt Ex,which is designed for systematic experimental design and model optimization,is developed based on this logical framework.This logical framework enhances the efficiency of experimental design and the effectiveness of model optimization by synthesizing the results of multiple model analysis methods.In this study,the experimental design method of similarity clustering is proposed based on the previously developed sensitivity entropy and surrogate model similarity.This method promotes the complementary advantages of both methods and enables effective constraints on model predictions under a wide range of conditions by using a small fraction of experimental data.The applicability and efficiency of the logical framework and computational software in optimization problems for small and large molecule models,are illustrated using methanol and n-decane as examples,respectively.In order to further improve the efficiency of experimental design and solve the problem that existing experimental design methods are difficult to be applied to experimental group design,an efficient experimental design method based on heuristic algorithms is developed.The performances of different heuristic algorithms are compared.In the application case s of experimental design for dimethyl ether,the experimental design method based on the heuristic algorithm can achieve the experimental design target with less than one-tenth of the computational costs compared with the traditional enumeration method.Moreover,the heuristic algorithm can also achieve a high-efficient process of the experimental group design.Based on the above-mentioned computational methods for efficient experimental design,this study investigates the effects of parameter correlation on model analysis as well as experimental design results.This work performs quantitative calculations of correlations between molecular properties based on quantitative chemistry methods.The correlations between molecular properties in the C₄H₇ potential energy surface system are quantified with the Pearson correlation coefficient.The effects of correlations between molecular properties on the theoretical calculations of reaction rate coefficients are studied.This study also develops a method for calculating the correlation coefficients between branching reactions.Two branching reactions in a 2-butene model are investigated to show the effects of correlations on the model uncertainty quantification and sensitivity analysis.