Generalized decay and artificial neural network models for fixed-Bed phenolic compounds adsorption onto activated date palm biochar

Abstract

Batch adsorption of phenolic compounds on different types of biochars have been intensively, reported in literature. However, such studies under fixed-bed column for ortho-cresol and phenol are sparsely investigated with none that employed artificial neural network (ANN) modeling despite its reported unique merits in adsorption field in recent years. In this study, fixed-bed breakthrough column tests for aqueous-phase ortho-cresol and phenol adsorption onto activated date palm biochar (DPBC), under operational conditions of flowrate (5-50 ml/min), bed depth (10-40 cm) and initial concentration (10-100 mg/l).were undertaken. The breakthrough curve (BTC) data obatined were employed to develop a feed-forward artificial neural network (ANN) and nonlinear regression generalized decay-function (GEDF). The mathematical form of the GEDF models fit the adsorption process well with reasonable accuracy, while the ANN model exhibited better predictions. The breathrough kinetic data best fitted Thomas model (R2 = 0.9507−0.997) with the maximum bed capacity obtained at initial concentration 100 mg/L, 30 ml/min flow-rate and bed-depth of 13.3 cm for both phenol and ortho-cresol. The ANN model’s sensitivity analysis coupled with Garson algorithm, profile method and performance decomposition were used to rank the operational variables. The investigated parameters influence of the phenols uptake follows the order initial concentration > flow rate > time > bed depth or DPBC mass. Increasing flow rate leads to linear and exponential increase in adsorbates effluent concentration, respectively. Higher DPBC bed depth only causes slight linear decline in the adsorbates effluent concentration. The GEDF models provide single generalized approach capable of predicting the adsorption process BTC under different operational conditions