A machine learning based approach to solve the aerosol dynamics coagulation model

AbstractSolving aerosol dynamic models accurately to obtain the size distribution function is often computationally expensive. Conventional artificial neural network (ANN) models offer an alternative procedure to solve the aerosol dynamic equations. However, conventional ANN models can result in violation of aerosol mass conservation. To further enhance accuracy and reduce computational time, a hybrid ANN approach to solve the aerosol coagulation equation is developed, validated, and demonstrated. The methodology and assumptions for the development of the hybrid ANN model which provides an analytical closed form solution for aerosol coagulation is described. The ANN model is trained and validated using a dataset from an accurate sectional model. Following this, the hybrid ANN aerosol model is used to describe the evolution of aerosol in a furnace aerosol reactor. The hybrid ANN model results are compared to the accurate sectional and moment coagulation models. The hybrid ANN coagulation model prediction was found to accurately describe the evolution of the size distribution at a computational cost which is slightly more than the moment model but orders of magnitude less than the sectional model.Copyright © 2023 American Association for Aerosol ResearchEDITOR: Nicole Riemer Data availability statementThe data that support the findings of this study are openly available in Mendeley Data at https://data.mendeley.com/datasets/pt4wjkhmyk/1 (Okonkwo et al. Citation2023).Disclosure statementNo potential conflict of interest was reported by the author(s).Additional informationFundingPartial support for this work was provided by a grant from the US Department of Energy: Development of Critical Components for the Modular Staged Pressurized Oxy-Combustion Power Plant; DE-FE0031925.