Insights into the estimation of heavy metals ions sorption from aqueous environment onto natural zeolite

In this work, we present how soft computing approaches can be used to study the sorption performance of natural zeolite to eliminate heavy metals ions including Zn2+, Ni2+, Cd2+, and Pb2+, from aqueous environment. The models include least-squares support-vector machine (LSSVM), genetic programming (GP), particle swarm optimization-adaptive neuro-fuzzy inference system (PSO-ANFIS), and artificial neural network (ANN), as well as multivariate nonlinear regression (PNR). The inputs of the models are electronegativity, hydrated ionic radii (A), first ionization energy (E1, kJ/mol) and molecular weight (Mw, g/mol) of heavy metal ion, initial pH (pHi, mmol/l) and equilibrium pH (pHe, mmol/l) of solution, and Si concentrations in the aqueous phase (Sie, mg/l). The output of the models is ionic species sorbed per gram of zeolite, f, (mmol/g). The importance of initial concentration, pollutant molecular weight, and acidic functional groups in the sorption of heavy metals is emphasized by sensitivity analysis. Ion exchange variables are determined by the adsorption isotherm expressions: Redlich–Petersen (RP), Langmuir–Freundlich (LF), Dubinin–Radushkevich (DR), Toth (T), Lineweaver–Burk (LB), and modified Dubinin–Radushkevich (MDR). Here, we focus on the applications of smart techniques in modeling complicated sorption systems used in wastewater treatment and environmental pollution control. The % AARD values for LSSVM, GP, ANN, PNR (with 5 order of polynomial), PSO-ANFIS, LF, RP, T, DR, MDR, and LB are 0.77, 5.37, 2.47, 1.23, 1.59, 177.1, 15.1 4215.8, 159.3, 128.9, and 11.2, respectively. The results of the computational techniques are found better than the adsorption isotherms.