Evaluation and machine learning-based prediction of Zn2+treatment by the cyanobacterium biomaterial at packed columns

Abstract

Excessive accumulation of zinc in the environment poses ecological risks. Hence, an approach with cost-effective and eco-friendly treatment media, such as microbial biomass, should be explored to remediate zinc from effluents. This study demonstrated the removal of Zn 2 + by packed columns, comprising Spirulina platensis (cyanobacterium) biomass fixed on polyurethane. Machine learning (ML) was also used to optimize and predict the concentration of Zn 2 + in the effluent. The results show optimum condition for Zn 2 + removal at a biomaterial height of 25 cm, a flow rate of 5 mL/min, and an inlet Zn 2 + concentration of 100 mg/L. Further, the Cubist algorithm predicted the Zn 2 + concentrations (new data) with R 2 of 0.988 and 5.34 mg/L in RMSE, and artificial neural networks (ANN) achieved 0.979 in R 2 and 6.94 mg/L in RMSE. Based on both the estimated accuracy and computing time, the Cubist model was found more advantageous than the ANN and linear regression models. This study provides optimum conditions for scaling up Zn 2 + treatment systems and an ML-based tool to estimate Zn 2 + effluents to support the design of adsorption column systems. • Spirulina platensis -polyurethane packed column built for Zn 2+ removal. • Machine learning used to explore adsorption column for Zn 2+ treatment. • Optimal conditions were height 25 cm, flow 5 mL/min, and inlet 100 mg/L. • Cubist and ANN achieved R 2 0.98 and RMSE 5.34-6.94 mg/L. • Findings could optimize and scale up Zn 2+ treatment column systems.