A reduction approach based on the directed relation graph method with error propagation (DRGEP) by incorporating curve-matching-based error measure

The chemical kinetic models are often simplified to a skeletal level by using automatic reduction methods. For this, the impact of the removal of species and reactions on model prediction is normally evaluated to identify the redundant species and reactions in the mechanism during the reduction. Even though the influences of error propagation are sometimes taken into account in the reduction, the induced errors by removing species or reactions are typically quantified at individual conditions, which do not consider the dependence of quantities of interest on physical conditions. Nevertheless, the quantities of interest are the functions of various physical conditions, such as temperature and pressure. Thus, in this work, a reduction approach based on the widely used directed relation graph method with error propagation by incorporating curve-matching (CM)-based error measure approach (DRGEPCM) is proposed. During the reduction with DRGEPCM, curves are generated by fitting the responses of reduced and detailed models based on one physical condition parameter. The CMbased error measure evaluates the similarity between the curves of reduced and detailed models in terms of values and shapes. ... mehrBy using this error measure, the development of dependence of prediction of reduced models on physical conditions is revealed. As example, a chemical model of 𝑛-decane is reduced by using the DRGEP and DRGEPCM methods in this study. The results are compared and discussed. It is shown that the prediction dependence of model predictions on temperature can be well-captured by the models reduced by using the CM-based error measure in model reduction. The values of error tolerance for the DRGEPCM method have significant impacts on the predictive performance of reduced models, at specific narrow ranges of error tolerance. The reduced model provided by using the DRGEPCM is more accurate and compact for certain error tolerance, compared to that by DRGEP at identical error tolerance specified for the reduction. The results highlight the importance of the definition of error measure in the model reduction. Novelty and significance statement This work proposes a novel reduction approach based on the DRGEP by incorporating a curve-matching-based error measure (DRGEPCM). The dependence of model predictions on physical condition parameters is thus taken in model reduction into account for the first time. It is found that the prediction dependence of model predictions on temperature can be captured well by the models reduced by using the DRGEPCM method. The models reduced by using the proposed method demonstrate better performance for certain cases, compared to those generated with the DRGEP method at identical error tolerance specified for the reduction. The results of this work highlight the importance of the definition of error measure in the model reduction