نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی دکتری،دانشگاه تهران، تهران،ایران
2 دانشیار،دانشگاه تهران، تهران،ایران
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
The use of manifold-based models for mapping governing equations from physical space into the species composition space and constructing flamelet tables has led to a significant reduction in computational cost for simulating reactive flows. Within the framework of flamelet modeling, the manifold space is characterized by the mixture fraction in non-premixed flames, and by the reaction progress variable in premixed flames. A key parameter in the transformation from physical to manifold space is the scalar dissipation rate, which is inherently dependent on the spatial gradients of the manifold coordinates in the physical domain. Therefore, prior to solving the reactive flow and generating flamelet tables, it is essential to develop an appropriate model for the scalar dissipation rate based on the manifold components. Such a model must effectively represent the characteristics of the physical space within the manifold framework. In this study, the results obtained from solving the governing equations in manifold space are analyzed and compared with reference solutions in the physical coordinate system. This comparison aims to assess the predictive capabilities of various scalar dissipation rate models. To this end, a novel data-driven model based on deep neural networks is proposed for predicting the scalar dissipation rate of the reaction progress variable, using a dataset of freely propagating premixed flames. The results demonstrate that the proposed data-driven model yields superior accuracy compared to traditional modeling approaches, offering improved predictions of flame behavior.
کلیدواژهها [English]
)