Mathematical Problems in Engineering

Research Article

Employing Artificial Neural Networks to Predict the Performance of Domestic Sewage Treatment Terminals in the Rural Region

Table 8

ANN application cases in the previous works.


Application scenario	Model	Input parameters	Predicted parameters	R ²	Error	Authors

WWTP	Adaptive neuro fuzzy inference system	Influent pH, influent temperature, influent SS, and influent COD.	Effluent SS	0.88 for training; 0.89 for validation	1.87 for training; 0.71 for validation	[34]
			Effluent COD	0.79 for training; 0.83 for validation	3.83 for training; 2.47 for validation
			Effluent pH	0.93 for training; 0.93 for validation	0.07 for training; 0.05 for validation

Fenton reactor	Artificial neural network	Reaction time, pH, antibiotic concentration, H₂O₂, etc	COD removal efficiency	0.997	0.000376	[47]

WWTP	Artificial neural network-feed forward back propagation	COD, BOD, TSS	COD	0.93 for training;	—	[48]
				0.86 for validation

Anoxic/oxic reactor	Artificial neural network	Influent COD, reflux ratio, carbon-nitrogen ratio	Effluent COD	0.98596	2.82	[49]
	Artificial neural network-genetic algorithm			0.99476	1.12

The upflow anaerobic filter reactor	Multilayer perceptron neural network	Influent chemical oxygen demand, hydraulic retention time, and influent cyanide concentration	Effluent COD	0.983 for training; 0.876 for validation	104.75 for training; 98.35 for validation	[31]
	Radial basis neural network			0.995 for training; 0.708 for validation	56.81 for training, 157.95 for validation
	Regression neural network			0.951 for training; 0.751 for validation	175.19 for training, 140.51 for validation

Lakes	Artificial neural network	NH4-N, secchi depth	TN	0.72 for training, 0.69 for validation	0.15 for training, 0.14 for validation	[50]

Partial-nitritation/anammox reactor	Back propagation artificial neural network	Influent NO3-, NO2-, pH and enfluent NO3-, NO2-,	TN removal efficiency	0.889	0.091	[51]

WWTP	Fuzzy rough-back propagation	Return ratio, sludge volume, MLSS, DO, Q (flow), efﬂuent suspended solid, influent TP, influent TN, influent COD, influent ammonia concentraion.	Effluent COD	0.921	2.86	[23]
			Effluent TN	0.882	1.35
	Principal component analysis-back propagation		Effluent COD	0.886	3.15
			Effluent TN	0.862	1.52
	Back propagation-neural networks		Effluent COD	0.813	3.46
			Effluent TN	0.822	1.97

In the error column, numbers with one asterisks is MSE, and numbers with two asterisks is RMSE.