A Review of the Modeling of Adsorption of Organic and Inorganic Pollutants from Water Using Artificial Neural Networks

Table 2

Summary of the ANN modeling of the multicomponent adsorption of water pollutants at batch operating conditions.

Adsorbent	Adsorbate	ANN model	Experiment	Data used (Training-Testing-Validation)	Input variables	Output variable	Pollutants removal	Reference
Chitosan-based hybrid hydrogels	Acid blue, allura red	Levenberg-Marquardt back propagation (5-10-10-10-2)	Binary kinetics	221-47-47	C01, C02, carbon content, porosity, t	Adsorption capacity	AB 242.5 mg/g, AR 219.35 mg/g	[7]
Ultrasound-modified chitin	Co^2+, Ni^2+, methylene blue	Levenberg-Marquardt, Bayesian regulation and back propagation with gradient derivatives	Isotherms	140-30-30	C0, T	Adsorption capacity	Co^2+ 62.34 mg/g, Ni^2+ 52.53 mg/g, methylene blue 11.30 mg/g	[27]
Bone char	Cd^2+, Ni^2+, Zn^2+, Cu^2+	Multi-layer feed forward (4-(10-20)-(5-10)-4)	Multi-component isotherms	99-28-14	C0	Adsorption capacity	Cd^2+ 1.62 mmol/g, Ni^2+ 1.35 mmol/g, Zn^2+ 0.90 mmol/g, Cu^2+ 1 mmol/g	[28]
Activated carbons	Acid orange, acid blue, caffeine, acetaminophen, benzotriazole	Hybrid ANN - homogeneous surface diffusion model	Binary and ternary kinetics	90-5-5	Type of adsorbent, pH, T, C0, ratio m/V	Removal percentage	AO 65%, AB74 85%, ACT 64%, CAF 82%, BTA 78%	[29]
Activated carbon	Nimesulide, paracetamol	Bayesian regularization back propagation algorithm	Isotherms	85-0-15	Ps, m, t, C0	Adsorption capacity	Nimesulide 0.22 mmol/g, paracetamol 0.16 mmol/g	[30]
Activated carbon from kiwi peel, cucumber peel and potato peel	Methylene blue, malachite green, rhodamine B	Three-layered feed forward (3-7-1)	Single, binary and ternary equilibrium	184-40-40	pH, t, T, m, C0	Adsorption capacity	MB 400 mg/g	[46]
Activated carbon	Herbicides 2,4-D and 4-chloro-2-methylphenoxyacetic acid	Multi-layer perceptron	Binary isotherms	21-0-9	Ce	Adsorption capacity	D 0.48 mmol/g, M 0.54 mmol/g	[48]
Graphite oxide nano particle	Methylene blue, brilliant green	Principal component analysis - three layered feed forward Levenberg–Marquardt back-propagation (4-4-1)	Binary equilibrium	100-40-0	Ce in mixture	Ce	MB 410 mg/g, BG 129.41 mg/g	[43]
Granulated activated carbon	Nitrobenzene, phenol, aniline	Three-layered feed forward Levenberg-Marquardt back propagation (3-1-1)	Ternary equilibrium	40-41-0	Ce1, Ce2, Ce3	Adsorption capacity	NI 0.39 mmol/g, AN 0.4 mmol/g, PH 0.32 mmol/g	[147]
Mixture of bentonite, zeolite, biochar, cockleshell, cement	Atenolol, ciprofloxacin, diazepan	Three-layered feed forward Levenberg-Marquardt back propagation (2-5-1)	Multicomponent equilibrium, RSM	45-15-15	t, C0	Removal percentage	ATN 90.2%, CIP 94%, DIA 95.5%	[286]
Microbioal biosorbent	Cu^2+-Cd^2+	Feed forward back propagation (3-10-2)	Binary isotherms	16-17-17	C0, pH	Adsorption capacity	Cu^2+ 9.75 mg/g, Cd^2+ 4.48 mg/g	[110]
MnO2-loaded activated	Brilliant green, crystal violet, methylene blue	Three-layered feed forward Levenberg Marquardt back-propagation (6-12-3)	Ternary equilibrium, RSM (CCD)	62-14-14	m, t, C0, Ce	Removal percentage of each dye	BG 206 mg/g, CV 234 mg/g, MB 263 mg/g	[126]
Magnetic ɤ-Fe2O3-loaded activated carbon	Methylene blue, malachite green	Feed forward Levenberg-Marquardt back propagation (5-11-2)	Binary equilibrium, RSM (CCD)	36-7-7	pH, m, t, C01, C02	Removal percentage of each dye	MB 195.55 mg/g, MG 207.04 mg/g	[130]
Mn@CuS/ZnS nanocomposite-loaded activated carbon	Methylene blue, malachite green	Radial basis function neural network with Kernel stone algorithm	Binary, RSM (CCD)	22-10-0	pH, m, t, C01, C02	Removal percentage of each dye	MB 126.42 mg/g, MG 115.08 mg/g	[136]
SnO2 nanoparticles loaded on activated carbon	Sunset yellow, disulfine blue	Three-layered feed forward back propagation (5-11-2)	Binary equilibrium, RSM (CCD)	26 data	pH, m, t, C01, C02	Removal percentage	SY 83.34 mg/g, DB 94.94 mg/g	[143]
CuO nanoparticles supported on activated carbon	Rose bengal, safranin O, malachite green	Three-layered feed forward Levenberg-Marquardt back propagation (6-10-3)	Ternary equilibrium, RSM (CCD)	65-16-0	pH, m, t, C01, C02, C03	Removal percentage	MG 94.26%, SO 71%, MG 76%	[146]
Zeolite	Ag^1+, Co^2+, Cu^2+	ANN Bayesian regularization backpropagation (4-10-1) and Adaptive Neuro-fuzzy Inference Systems (4-8-X-X-1)	Isotherms	275-49-0	T, Si/Al ratio, C0, molecular weight of adsorbate	Adsorption capacity	Ag^1+ 5 mmol/g, Co^2+ 2.51 mmol/g, Cu^2+ 3.88 mmol/g	[157]
Sargassum filipendula biomass	Cd^2+-Zn^2+	Feed forward back propagation (7-5-2)	Binary isotherms	16	Ce	Adsorption capacity	Cd^2+ 291.2, Zn^2+ 169 mg/g	[287]
Activated carbon, wood charcoal and rice husk ash	Phenol, Resorcinol	Three-layer feed forward Levenberg-Marquardt back propagation	Binary kinetics	15-7-7	pH, t, C0, m	Removal percentage	Phenol 75%, Resorcinol 90%	[288]
Chitosan foams	Cu^2+, Zn^2+, Cr^6+	Multi-layer perceptron Levenberg Marquardt (3-1-1)	Multicomponent Isotherms	--	Ce	Adsorption capacity	Cu^2+ 61mg/g, Zn^2+ 68 mg/g, Cr^6+ 58 mg/g	[289]
Aegel marmelos fruit shell	Ni^2+ Cr^6+	Elite-ANN with ACM model (4-10-10-6)	Equilibrium	17-0-4	C0, m, type of adsorbent	Removal percentage, adsorption capacity and Ce	Ni^2+ 59.26 mg/g, Cr^6+ 12.67 mg/g	[290]
SnO2 nanoparticle-loaded activated carbon	Acid yellow 41, sunset yellow	Principal component analysis-artificial neural network	Binary equilibrium, RSM (CCD)	10-5-5	pH, m, t	Removal percentage	AY41 95.6%, SY 97.9%	[291]

Nomenclature: adsorbent dosage (m), central composite design (CCD), final adsorbate concentration (Ce), initial adsorbate concentration (), particle size (Ps), response surface methodology (RSM), temperature (), time (), and volume ().