Experimental design and data prediction by Bayesian statistics for adsorption of tetracycline in a GAC fixed-bed column

In the present work, the efficiency and applicability of fixed-bed columns were evaluated for the adsorption of tetracycline (TC) on granular activated carbon (GAC). An experimental planning by the Box-Benking (BOX) design was evaluated to determine the best operational conditions of the column. It is involving three independent variables: initial concentration of the pollutant (20, 50 and 80 mg.L-1), bed height (1, 2 and 3 cm) and feed flow rate (3, 6 and 10 mL.min  1); and two response factors: the saturation time (tsat) and the maximum adsorption capacity (qmax). To simulate the breakthrough curve six models were applied (Thomas, Yoon-Nelson, Yan, Clark, Gompertz and Log-Gompertz) for the best operational conditions. The parameter estimates and model selection were performed using a Bayesian technique, applying the experimental value of qmax, and five statistical metrics (classical: R2 and R2ajstd; Bayesian: AIC, AICC and BIC). The results showed that all operational adsorption conditions applied obtained 95 % of TC removal and the best ones were: 80 mg.L-1, 6 mL.min  1 and 1 cm, for TC initial concentration, feed flow rate and bed height, respectively, achieving a tsat of 23 min and qmax of 2.08 mg. g  1. The Yan and the Log-Gompertz models were the ones that best predicted the behavior of the breakthrough curve obtained. The Yan model being responsible for elucidating the mechanisms of the column and the LogGompertz model useful for the scaleup of the process, being applicable for industrial implementation.