The uncertainty of the rate coefficients of elementary reactions in detailed chemical kinetic mechanisms can affect the predicted species concentrations and global properties. The size of reduced mechanisms generated by reduction methods, such as directed relation graph (DRG), can also be greatly affected by the uncertainties in the rate coefficients, because these methods eliminate species by comparing contributions of the production rates of important species from other species. In order to obtain the optimum reduced mechanism in the uncertainty domain of reaction rate parameters, the species-targeted global sensitivity analysis (STGSA) method is proposed to measure the importance of species in kinetic mechanisms and to simplify them by considering the input uncertainties. The demonstrations are performed for the ethylene oxidation mechanism using the USC-Mech II with 111 species and the n-heptane oxidation mechanism using the JetSurf (version 1.0) with 194 species. It is found that STGSA can accurately identify the redundant global species and generate more compact reduced mechanisms compared with the DRG-based methods, resulting in a 33-species reduced mechanism for ethylene oxidation and a 37-species reduced mechanism for n-heptane oxidation. These reduced mechanisms show accurate predictions for the ignition delay times and laminar flame speeds over a wide range of operating conditions.

在详细的化学动力学机理中基本反应速率系数的不确定性会影响预测的物种浓度和整体性质。速率系数的不确定性还可能极大地影响诸如归因关系图（DRG）之类的归约方法所产生的归约机制的大小，因为这些方法通过比较其他物种的重要物种生产速率的贡献来消除物种。为了在反应速率参数的不确定性范围内获得最佳的还原机理，提出了一种以物种为目标的全局灵敏度分析（STGSA）方法来测量物种在动力学机制中的重要性，并通过考虑输入不确定性来简化它们。使用USC-Mech II的111种物种和使用JetSurf（1.0版）的194种物种进行正庚烷氧化机理的演示。结果发现，与基于DRG的方法相比，STGSA可以准确地识别多余的全局物种，并产生更紧凑的还原机制，从而导致33种物种的乙烯氧化还原机理和37种物种的正庚烷氧化还原机理。这些简化的机构显示了在广泛的运行条件下对点火延迟时间和层流火焰速度的准确预测。结果发现，与基于DRG的方法相比，STGSA可以准确地识别多余的全局物种，并产生更紧凑的还原机制，从而导致33种物种的乙烯氧化还原机理和37种物种的正庚烷氧化还原机理。这些简化的机构显示了在广泛的运行条件下对点火延迟时间和层流火焰速度的准确预测。结果发现，与基于DRG的方法相比，STGSA可以准确地识别多余的全局物种，并产生更紧凑的还原机制，从而导致33种物种的乙烯氧化还原机理和37种物种的正庚烷氧化还原机理。这些简化的机构显示了在广泛的运行条件下对点火延迟时间和层流火焰速度的准确预测。