HYBRID EVENT: You can participate in person at Boston, Massachusetts, USA or Virtually from your home or work.

Guo Fengshi

Catalysis 2023
Guo Fengshi, Speaker at Catalysis 2023
Korea University, Korea, Republic of
Title : Prediction of permeable reactive barrier width for arsenic treatment in groundwater using machine learning techniques


Permeable reactive barrier (PRB) is an effective in-situ technology for aquifer and groundwater remediation [1]. The important factors in PRB design are the width and selection of the reactive material [2]. In this study, the beaded coal mine drainage sludge (BCMDS) was employed as the filling material for the PRB to adsorb arsenic pollutants in groundwater. The determination of PRB width using the conventional adsorption mechanism is determined by determining the delay coefficient through batch experiments and substituting it into Eq. 1, while the mass transfer zone is determined through the column experiment, and the final PRB width is determined by combining the equilibrium and mass transfer zone widths (WMTZ), i.e., Eq. 2 [3].  WTot =  tlife * μGround  / R --- [1] , WTot= (WEqu + WMTZ ) ---[2]. Both experimental data and machine learning predictions were used for the design of the PRB width. The calculation of WMTZ was conducted through column experiments under various conditions: different pollutant initial concentrations (1.5, 5, and 8.1 mg/L), pH values (3, 5, 8, and 9.5), and flow rates (0.56, 0.85, and 1.13 ml/min). Breakthrough curves were generated based on the experimental results. In this study, we try to predict the WMTZ of the PRB through machine learning and finally determine the PRB's width by considering the PRB's age. It is combined with XGBoost and SHapley Additive exPlanations (SHAP) [4], which simultaneously consider the type of adsorbent material, pollutant, and environmental conditions, with 157 data collected from articles to predict WMTZ. Data preprocessing and model training improved the prediction accuracy. The experimentally derived WMTZ values were also used to validate the machine learning predictions, demonstrating the accuracy of the predictions and the feasibility of using machine learning


Student GUO FENGSHI studied environmental engineering at Hebei University of Technology in China and graduated as a bachelor in 2021. Then started to study civil, environmental, and architectural engineering at Korea University, and joined the research group of Prof. Jeehyeong Khim at the AOP lab.