Evaluation of enhanced chemical coagulation method for a case study on colloidal liquid particle in wastewater treatment: Statistical optimization analysis and implementation of machine learning
Behnam Fayyaz Shahandashty,Narges Fallah,Mohammad Shamsi,Bahram Nasernejad,Morteza Afkhamipour
DOI: https://doi.org/10.1016/j.jenvman.2024.122345
2024-09-10
Abstract:Coal mines are one of the largest sources of energy supply and generate significant volumes of wastewater. Chemical coagulation is one of the most effective methods for wastewater treatment. In this research, ferric and aluminum-based coagulants, along with polyacrylamide flocculants with positive, negative, and neutral charges, were utilized in chemical coagulation. After applying the Plackett-Burman screening method, it was found that ferric chloride coagulant, neutral flocculant, and slow mixing duration had the greatest impact. The chemical coagulation process was modeled and optimized by examining these factors using the Box-Behnken statistical design as input parameters and sedimentation velocity as the output. Under optimal conditions, the values for ferric chloride coagulant, neutral flocculant, mixing time in slow mode, and sedimentation velocity were determined to be 106.3 mg/L, 3.98 mg/L, 29.6 min, and 1.10 cm/min, respectively. Under optimal conditions, the removal percentages of pollutants, including TSS, turbidity, TDS, COD, and BOD, were obtained at 100%, 100%, 87%, 93%, and 81%, respectively. The experimental data were fitted using the BBD and ANN methods. Both models demonstrated very high agreement, but the ANN method performed better with an AAD% of 0.66, an MSE of 0.0001, and an R2 value of 0.99. All results were calculated with a confidence level above 98%, indicating that both models had very high reliability in modeling and prediction.