燃烧反应动力学中的灵敏性和不确定性分析

发布时间：2018-04-25 20:20

本文选题：燃烧反应动力学 + 不确定性分析　；参考：《中国科学技术大学》2017年博士论文

【摘要】：可靠的燃烧反应动力学模型有助于我们深入地理解燃烧反应过程,进而帮助我们设计更高性能的发动机,提高燃烧效率,降低污染物排放。燃烧反应动力学模型的研究主要围绕着以下三个方面:(1)基元反应的热力学及动力学理论计算;(2)基础燃烧实验,包括宏观燃烧的属性和微观组分的测量;(3)基于前面两方面的最新研究成果,发展和改进燃烧反应动力学模型。这三部分既各自发展又相互影响。如单分子解离反应的理论计算方法从Lindemann机理发展到RRKM理论,越来越接近物理化学过程的本质,这是其独立发展的部分;利用某些条件下实验测量的速率系数修正RRKM理论中的部分参数,从而得到更准确的温度压力依赖的速率系数,这是其相互影响的部分。燃烧反应动力学模型发展的最终目标之一是降低模型预测的不确定性,得到能够进行准确预测的燃烧反应动力学模型。燃烧反应动力学模型的发展过程就是不断降低其预测不确定性的过程。本工作围绕着如何更加高效地降低模型的不确定性,以灵敏性和不确定性分析为基本工具,探究不确定性在理论计算、模型和实验中的传递机制,进而发展相关计算方法,最终构造一条降低模型不确定性的快速通道。首先对传统的全局灵敏性分析方法进行了改进,提出了利用人工神经网络加速传统的全局灵敏性分析的方法。此方法结合人工神经网络算法拟合能力强、收敛速度快的优点以及高维模型表征法易于计算灵敏性系数的优点,构造了双层替代模型,大大加快了全局灵敏性分析的收敛速度,使之适用于更加复杂的模型。然后从理论计算、模型和实验三个方面深入探究如何利用灵敏性和不确定性分析高效地降低模型的不确定性。(1)在理论计算方面,利用不确定性分析评估RRKM/ME理论计算得到的速率系数的不确定性,发现了在有多条通道的反应体系中,次要通道的速率系数理论计算结果的不确定性要远远高于主要通道的结果。同时提出了一种简单有效评估具有温度压力依赖效应的速率系数不确定性的方法,从而大大降低了评估具有温度压力依赖效应的速率系数不确定性的难度,提高了其实用性。(2)在燃烧实验方面,我们提出了利用模型辅助提高实验探测极限的方法,通过分析不同条件下模型预测不确定性以及寻找在不同条件下系统误差之间的联系,使得我们可以测量某些之前无法定量的实验目标。(3)在模型发展方面,为了得到对模型优化更加有效的实验数据,我们提出了灵敏性熵的概念来帮助我们设计实验,并在甲醇燃烧体系中验证了这种方法的可行性。本论文的一系列工作表明,利用合适的方法,通过合理的分析不确定性在理论计算、动力学模型和燃烧实验中的传递机制,可以有效地提高部分研究目标的测量精度,也可以设计出对模型优化最有效的实验,进而为降低动力学模型的不确定性提供了一条快速通道。
[Abstract]:A reliable kinetic model of combustion reaction helps us to understand the combustion process in depth, and then help us to design a higher performance engine, improve combustion efficiency and reduce the emission of pollutants. The research on the kinetic model of the combustion reaction mainly focuses on the following three aspects: (1) the thermodynamic and kinetic theoretical calculation of the elemental reaction; (2 ) basic combustion experiments, including the properties of macroscopic combustion and the measurement of microscopic components; (3) developing and improving the kinetic model of combustion reaction based on the latest research results in the first two aspects. The three parts both develop and interact with each other. For example, the theoretical calculation method of single molecule dissociation reaction is developed from the Lindemann mechanism to the RRKM theory, and more and more connected. The essence of the physicochemical process is the part of its independent development; the rate coefficients in the RRKM theory are corrected by the rate coefficients measured experimentally under certain conditions, thus obtaining a more accurate rate coefficient of temperature and pressure dependence. This is part of its interaction. One of the ultimate goals for the development of the kinetic model of the combustion reaction is to reduce it. The development process of the combustion reaction dynamics model is to reduce the uncertainty of the prediction. This work focuses on how to reduce the uncertainty of the model more efficiently, and explore the sensitivity and uncertainty analysis as the basic tool. The transfer mechanism in the theoretical calculation, model and experiment is not determined, and then the relative calculation method is developed, and a fast path to reduce the uncertainty of the model is finally constructed. First, the traditional global sensitivity analysis method is improved and the method of using the artificial neural network to accelerate the traditional global sensitivity analysis is proposed. Method combining the advantages of artificial neural network algorithm with strong fitting ability, fast convergence speed and the advantages of high dimensional model characterization method which is easy to calculate the sensitivity coefficient, a double layer substitution model is constructed, which greatly accelerates the convergence rate of global sensitivity analysis and applies to more complex models. Then, the theoretical calculation, model and experiment three are used. How to use sensitivity and uncertainty analysis to effectively reduce the uncertainty of the model. (1) in theoretical calculation, using uncertainty analysis to evaluate the uncertainty of the rate coefficient calculated by the RRKM/ME theory, we find the theoretical calculation of the rate coefficient of the secondary channel in the reaction system with multiple channels. The uncertainty of the fruit is much higher than the result of the main channel. At the same time, a simple and effective method to evaluate the rate uncertainty of the temperature pressure dependence effect is proposed, which greatly reduces the difficulty of evaluating the rate uncertainty with the temperature pressure dependence effect and improves its practicality. (2) in the combustion experiment party On the other hand, we put forward the method of using model aided to improve the limit of experimental detection. By analyzing the uncertainty of model prediction under different conditions and finding the connection between system errors under different conditions, we can measure some experimental targets that can not be quantified before. (3) in the development of the model, we can get the model optimization more. With the effective experimental data, we propose the concept of sensitivity entropy to help us design the experiment and verify the feasibility of this method in the methanol combustion system. A series of work in this paper shows that the appropriate method is used to analyze the transfer of the kinetic model and the combustion experiment by reasonable analysis of the uncertainty in the theoretical calculation, the dynamic model and the combustion experiment. The mechanism can effectively improve the measurement accuracy of some research targets, and can also design the most effective experiment for the model optimization, and then provide a fast channel for reducing the uncertainty of the dynamic model.

【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TK16

【相似文献】