Novel Correlation for Critical Speed for Solid Suspension in Stirred Tanks Developed Using Machine Learning Models Trained on Literature Data

Critical speed for solid suspension(N (js)) in stirred tanks is an importantdesign parameterin several chemicalprocesses. There is a need to develop a generalized correlation thatapplies to a broad range of literature data. Also, the literaturelacks a comparative study of different machine learning (ML) modelsfor the prediction of N (js). In this paper,3240 data points have been extracted from 35 papers on solid suspensionand N (js) has been modeled as the dependentvariable, initially using ML models. Three ML models (random forestregression (RF), CatBoost regression (CatB), and artificial neuralnetwork regression (ANN)) are compared. The CatB model was the mosteffective, resulting in R (2) = 0.99 forthe testing dataset, better than any empirical correlation published.Further, we have successfully used the CatB model to obtain the functionalform of the behavior of various parameters. We then developed novelcorrelations (in closed form and with parametric uncertainty) forHE-3, PTD, and DT impellers with R (2) =0.89, 0.84, and 0.86, respectively, where the correlation constantswere tuned using experimental data published in the literature. Ourclosed-form correlations significantly outperform earlier correlationspublished in the literature. Such a methodology has not been reportedin the literature we have surveyed, and we believe that it is noveland original.

Automated summary

Critical speed for solid suspension in stirred tanks is significant in chemical processes. A generalized correlation is needed for various literature data, while a comparative study of different ML models for the prediction of the critical speed is lacking in the literature. In this study, ML models including RF, CatB, and ANN were compared, and the CatB model achieved the highest effectiveness. Novel correlations for HE-3, PTD, and DT impellers were developed, outperforming previous correlations in the literature.