Abstract View
A Machine Learning Based Aerosol Dynamics Model for Log-Normal Aerosols
ONOCHIE OKONKWO, Rahul Patel, Ravindra Gudi, Pratim Biswas, University of Miami
Abstract Number: 382
Working Group: Aerosol Physics
Abstract
Aerosol reactors play a crucial role in industrial-scale synthesis of advanced ceramics powders and commodity particles including carbon blacks and titania. The product properties – which include light scattering, and photo-catalytic properties for titania – depend on the aerosol product characteristics such as particle size, size distribution, morphology, and crystal phase. These characteristics are determined by various aerosol dynamics phenomena, including, reaction, nucleation, condensation, sintering, coagulation, and charging occurring in the reactor systems. For optimal design and control of aerosol reactors, an understanding of the evolution of particulate systems which accounts for the dynamics phenomena is required. The aerosol dynamics phenomena are described by the general dynamics equation (GDE). The GDE is a non-linear partial integro-differential equation which, therefore, requires the development of efficient numerical solution methods. The complexity and computational burden associated with solving the GDE lies in the coagulation term [1].
Artificial neural networks (ANN) have been shown to accurately describe highly complex phenomena in a computationally efficient manner [2]. This presentation shows the first step in the development of an accurate and highly computationally efficient solution method for the GDE using a simplified coagulation term based on ANN. The ANN model, which is trained is used to evaluate two proxy coagulation kernels. Using the proposed ANN solution approach, simulations are done for pure coagulation scenarios. The results from the proposed model are compared with the moment model and sectional model to demonstrate its accuracy and superior computational performance.
References
[1] H. Zhang et al., Aerosol Science and Technology, 54 (2020).
[2] J. Behler, International Journal of Quantum Chemistry, 115 (2015).