Conventional anddata‐drivenmodeling of filtered drag, heat transfer, and reaction rate ingas–particleflows

Abstract

This study presents conventional and artificial neural network‐based data‐driven modeling (DDM) methods to model simultaneously the filtered mesoscale drag, heat transfer and reaction rate in gas–particle flows. The dataset used for developing the DDM is filtered from highly resolved simulations closed by our recently formulated microscopic drag and heat transfer coefficients (HTCs). Results reveal that the filtered drag correction is nearly independent of filter size when including the filtered gas phase pressure gradient. We further find that the filtered HTC correction critically depends on the added filtered temperature difference marker while the filtered reaction rate correction shows weak dependence on the additional markers. Moreover, compared with conventional correlations, DDM predictions agree better with filtered resolved data. Comparative analysis is also conducted between existing HTC corrections and our work. Finally, the applicability of conventional and data‐driven models coupled with coarse‐grid computational fluid dynamics simulations for pilot‐scale (reactive) gas–particle flows is validated comprehensively.

Review of Machine Learning for Hydrodynamics, Transport, and Reactions in Multiphase Flows and Reactors

“…However, the relative importance and relevance between the newly introduced variables and the old variables may need to be identified. This is because the filtered drag correction is nearly independent of filter size when introducing the closure marker of the gas phase pressure gradient in traditional physics-based modeling . Moreover, whether the trained model can be well extrapolated to different gas–solid flow patterns is also an important issue in NN modeling .…”

“…However, the data source of this study is generated by simulations of a relatively small domain in which there are possible numerical artifacts since the clusters or bubbles can grow to the domain scale and hence lead to unrealistic settlement velocity. Their later study further found that the correction model can still perform well by using the closure marker of the fluid phase pressure gradient in the absence of filter size . Except for the use of simulations to generate data sets, theoretical methods such as the energy minimization multiscale (EMMS) can also provide data for training.…”

“…Besides, some researchers have also applied ML to study heat and mass transfer inside porous media. − However, ML has not yet witnessed extensive adoption to problems of transport phenomena accompanied by chemical reactions in multiphase reactors. To tackle this issue, Zhu et al implemented ANN-based data-driven modeling of heat transfer and mass transfer accompanied by chemical reactions in gas–particle fluidized bed reactors, with a prediction accuracy of MAPE = 11.14%, r = 0.98 vs the traditional correlation MAPE = 21.11%, r = 0.96. The key advance of this work is the discovery of a new closure marker (e.g., the mesoscale interphase temperature difference) as the additional feature input which thus boosts the model performance.…”

“…The hybrid model enabled better representation of the petrol yield data than those estimated by a pure mechanistic model or a pure data-driven ANN model. Recently, we performed data-driven modeling of the filtered subgrid reaction rate source term to correct the reaction rate observed in a coarse-mesh TFM simulation based on the data from fine-grid simulations . Due to a relatively simple relationship between the filtered reaction rate output and two feature inputs (filter size and solid hold-up), conventional fitting and data-driven methods showed comparable prediction performance of a solid-catalyzed first-order reaction.…”

Modeling filtered heat transfer model by considering dimensionless temperature difference between gas and solid phases

“…By filtering CFD–DEM data, Lei et al 29 proposed a heat transfer model as a function of the filtered solid volume fraction, the filter size and the dimensionless temperature difference between the two phases. Zhu et al 30 proposed a similar model with dimensional temperature difference. Lei et al 31 then extended the model by Zhu et al into the bidisperse systems.…”