Statistical analysis, machine learning modeling, and text analytics of aggregation attachment efficiency: Mono and binary particle systems

Authors: GOMEZ-FLORES, ALLAN - HANYANG UNIVERSITY; Bradford, Scott; HONG, GILSANG - HANYANG UNIVERSITY; KIM, HYUNJUNG - HANYANG UNIVERSITY

Submitted to: Journal of Hazardous Materials
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/22/2023
Publication Date: 4/24/2023
Citation: Gomez-Flores, A., Bradford, S.A., Hong, G., Kim, H. 2023. Statistical analysis, machine learning modeling, and text analytics of aggregation attachment efficiency: Mono and binary particle systems. Journal of Hazardous Materials. 454. Article 131482. https://doi.org/10.1016/j.jhazmat.2023.131482.
DOI: https://doi.org/10.1016/j.jhazmat.2023.131482

Interpretive Summary: Soil solutions, groundwater, and surface water contain small particles (e.g., colloids) that can aggregate over time. A knowledge and ability to predict this aggregation process is needed to predict the fate colloids in the environmental, to assess risks from colloidal contaminants, and to manage clogging during managed aquifer recharge. The paper provides a statistical analysis of published papers on colloid aggregation, and identifies and compares measured trends with theoretical expectations. Existing gaps in knowledge are identified and future research directions are proposed to overcome these limitations. This information will be of interest to scientists and engineers concerned with predicting and managing colloids in the environment.

Technical Abstract: The aggregation attachment efficiency (a) is the fraction of particle–particle collisions resulting in aggregation. Despite significant research, a predictions have not accounted for the full complexity of systems due to constraints imposed by particle types, dispersed matter, water chemistry, quantification methods, and modeling. Experimental a values are often case–specific, and simplified systems are used to rule out complexity. To address these challenges, statistical analysis was performed on a databases to identify gaps in current knowledge, and machine learning (ML) was used to predict a under various particle types and conditions. Moreover, text analytics was employed to support knowledge from statistics and ML, as well as gain insight into the ideas communicated by current literature. Most studies investigated a in mono–particle systems, but binary or higher systems require more investigation. Furthermore, our work highlights that numerous variables, interactions, and mechanisms influence a behavior, making its investigation complex and difficult for both experiments and modeling. Consequently, future research should incorporate more particle types, shapes, coatings, and surface heterogeneities, and aim to address overlooked variables and conditions. Therefore, building a comprehensive a database can enable the development of more accurate empirical models for prediction.