Interpretable machine learning analysis and automated modeling to simulate fluid-particle flows (Open Access)--颗粒学报PARTICUOLOGY

Ouyang, B., Zhu, L., & Luo, Z. (2023). Interpretable machine learning analysis and automated modeling to simulate fluid-particle flows. Particuology, 80, 42-52. https://doi.org/10.1016/j.partic.2022.12.004

Interpretable machine learning analysis and automated modeling to simulate fluid-particle flows (Open Access)

Bo Ouyang, Litao Zhu, Zhenghong Luo^*

Department of Chemical Engineering, School of Chemistry and Chemical Engineering, State Key Laboratory of Metal Matrix Composites, Shanghai Jiao Tong University, Shanghai, 200240, China

10.1016/j.partic.2022.12.004

Volume 80, September 2023, Pages 42-52

Received 14 September 2022, Revised 6 November 2022, Accepted 9 December 2022, Available online 20 December 2022, Version of Record 1 March 2023.

E-mail: luozh@sjtu.edu.cn

Highlights

• Interpretable machine learning provides valuable insight into filtered drag model.

• Slip velocity and volume fraction contribute the most to the filtered drag correction.

• The addition of gas pressure gradient as a third marker improves the prediction.

• Automated machine learning significantly simplifies the process of optimizing model structures and hyperparameters.

Abstract

The present study extracts human-understandable insights from machine learning (ML)-based mesoscale closure in fluid-particle flows via several novel data-driven analysis approaches, i.e., maximal information coefficient (MIC), interpretable ML, and automated ML. It is previously shown that the solid volume fraction has the greatest effect on the drag force. The present study aims to quantitatively investigate the influence of flow properties on mesoscale drag correction (H_d). The MIC results show strong correlations between the features (i.e., slip velocity () and particle volume fraction ()) and the label H_d. The interpretable ML analysis confirms this conclusion, and quantifies the contribution of , and gas pressure gradient to the model as 71.9%, 27.2% and 0.9%, respectively. Automated ML without the need to select the model structure and hyperparameters is used for modeling, improving the prediction accuracy over our previous model (Zhu et al., 2020; Ouyang, Zhu, Su, & Luo, 2021).

Graphical abstract

Keywords

Filtered two-fluid model; Fluid-particle flow; Mesoscale closure; Interpretable machine learning; Automated machine learning; Maximal information coefficient

文章导读