Machine Learning Approach to Predict the Surface Charge Density of Monodispersed Particles in Gas-Solid Fluidized...

Area of Science:

• Chemical Engineering
• Particle Technology
• Computational Modeling

Background:

• Gas-solid fluidized beds exhibit complex particle electrostatic behavior influenced by multiple factors.
• Existing research often examines single factors, lacking a comprehensive understanding of interactions.
• Developing mathematical models for multi-factor particle charging is challenging due to system complexity.

Purpose of the Study:

• To develop a predictive model for particle surface charge density in monodispersed gas-solid fluidized beds.
• To investigate the application of machine learning for analyzing particle charging behavior.
• To identify the most reliable machine learning model for this complex phenomenon.

Main Methods:

• Utilized machine learning techniques including kernel ridge regression (KRR), support vector machine regression (SVR), and multilayer perceptron (MLP).
• Trained models using literature and experimental data for predicting surface charge density.
• Performed sensitivity analysis to evaluate model reliability.

Main Results:

• SVR and MLP models achieved high prediction accuracy, with R-squared values of 0.980 and 0.979, respectively.
• Sensitivity analysis indicated that the MLP model demonstrated greater reliability compared to the SVR model.
• The developed models accurately predict particle surface charge density within a defined range.

Conclusions:

• Machine learning is a feasible approach for analyzing particle charging in fluidized beds.
• The proposed MLP model serves as an effective correlative tool for rapid estimation of particle surface charge density.
• This study provides a novel method for understanding complex electrostatic interactions in gas-solid systems.