Discrete Element Simulations and Machine Learning for Improving the Performance of Dry Catalyst Continuous Impregnation Processes

Joseph Shovlin,Kuang Liu,Yangyang Shen,Bill Borghard,Hernan A. Makse,Maria S. Tomassone

DOI: https://doi.org/10.48550/arXiv.2212.00634

2022-12-02

Abstract:In this work, discrete element method (DEM) simulations coupled with machine learning are used to study the process of dry impregnation. Our results show that the particle bed contains two regimes. Regime 1 shows smaller inclination angles and a larger mass hold-up which implies more forces restricting the particle movement. Regime 2 reveals larger inclination angles and rotational speeds and a smaller mass holdup, which indicates a smaller bed height. Using Machine learning, we found a general function for the Relative Standard Deviation (RSD) as a function of time, angle of inclination and speed of rotation for both even and uneven flow rates for a full range of the parameters fed in the LASSO algorithm. Machine learning gives insight on both regimes and reveals that for low RPM and low angles, uneven spraying gives a lower RSD which is consistent with our observations of the DEM studies.

Applied Physics,Materials Science

What problem does this paper attempt to address?

This paper attempts to study and improve the performance of the continuous impregnation process of dry catalysts through the Discrete Element Method (DEM) and machine - learning methods. Specifically, the paper focuses on the following issues:

1. **Influence of mixing and flow**: How does it affect mixing and flow when the particles have a certain humidity or are saturated with fluid?
2. **Improvement of fluid distribution**: How to improve the distribution of fluid within the particle bed to ensure a more uniform impregnation effect?
3. **Distribution of dead zones**: The range of dead zones under a specific impregnator configuration and their influence on the process.
4. **Comparison between metal transport and fluid transport**: The relationship between the transport of metal and the transport of fluid in the particle bed.
5. **Geometric optimization**: Improve mixing and content uniformity by using different geometric structures (such as baffles, spray patterns or systems).
6. **Influence of rotation mode, nozzles and spray patterns**: How do these factors affect the mixing profile within the particle bed?
7. **Process scale - up rules**: Provide empirical rules for process scale - up.

The paper uses DEM simulation combined with machine - learning techniques, especially the LASSO algorithm, to analyze the influence of different parameters (such as rotation speed, tilt angle, spray rate, etc.) on the relative standard deviation (RSD) within the particle bed, thereby optimizing the continuous impregnation process. The research shows that for low rotation speeds and small tilt angles, non - uniform spraying can reduce the RSD, which is consistent with the DEM simulation results. In addition, the change curves of RSD over time under different parameter combinations are predicted by machine - learning, providing theoretical basis and technical support for practical applications.