APPLICATION OF UNCERTAINTY ANALYSIS OF ARTIFICIAL NEURAL NETWORKSFOR PREDICTING COAGULANT AND ALKALIZER DOSAGES IN A WATER TREATMENT PROCESS

Menezes, F. C. de; Fontes, R. M.; Oliveira-Esquerre, K. P.; Kalid, R.

doi:10.1590/0104-6632.20180354s20170039

Acessibilidade / Reportar erro

Brasil

Brazilian Journal of Chemical Engineering

Español English

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

Process Systems Engineering • Braz. J. Chem. Eng. 35 (4) • Dec 2018 • https://doi.org/10.1590/0104-6632.20180354s20170039 copy

APPLICATION OF UNCERTAINTY ANALYSIS OF ARTIFICIAL NEURAL NETWORKSFOR PREDICTING COAGULANT AND ALKALIZER DOSAGES IN A WATER TREATMENT PROCESS

Authorship SCIMAGO INSTITUTIONS RANKINGS

ABSTRACT

Artificial neural networks (ANNs) were built to predict coagulant (Model I) and alkalizer (Model II) dosages given raw and treated water parameters from a water clarifying process. Different ANN architectures were tested and optimal results were obtained with [10-10-10-01] and [08-12-12-01] nodes of input, hidden and output layers for Models I and II, respectively. Two algorithms based on GUM-S1weredevelopedto evaluate the artificial neural network parameter uncertainty and the coverage interval of model outputs. The results show that these algorithms can provide a better set of parameters for the ANN compared with the traditional training method. The present research provides a unique unifying view that considers neural networks and uncertainty analysis in a well-documented industrial case study.

Key words:
Artificial intelligence; Parameter uncertainty; Coverage interval; Aluminum sulfate; Sodium hydroxide

Authors (year)	Modeling techniques	Data/information required
Heddam and Dechem (2015)Heddam, S., and Dechem, N., A new approach based on the dynamic evolving neural-fuzzy inference system (DENFIS) for modeling coagulat dosage (Dos): Case study of water treatment plant of Algeria. Desalination and Water Treatment, 53(4), 1045-1053 (2005).	Neural-fuzzy inference system (DENFIS)	Seven online measurements of raw water and alum dosage
Heddan et al (2011)Heddam, S., Abdelmalek, B., and Dechemi, N., Applications of radial-basis function and generalized regression neural networks for modeling of coagulant dosage in drinking water-treatment plant: Comparative study. Journal of Environmental Engineering, 137(12), 1209-1214 (2001).	Generalized regression neural network (GRNN) and radial basis neural network (RBFNN)	Six input variables 725 samples in 2 year-period
Heddam et al. (2012)Heddam, S., Bernard, A., and Dechemi, N., ANFIS-based modelling for coagulant dosage in drinking water treatment plant: a case study. Environmental Monitoring and Assessment, 184(4), 1953-1971 (2012).	Adaptative neuro-fuzzy inference system (ANFIS)	Seven online measurements of raw water and alum dosage
Hernandez and Le Lann (2006)Hernandez, H., and Le Lann, M.-V., Development of a neural sensor for on-line prediction of coagulant dosage in a potable water treatment plant in the way of its diagnosis. J. S. Sichman et al (Eds.) IBERAMIA-SBIA, LNAI 4140, 249-257. Springer-Verlag Berlin Keideberg(2006).	ANN and linear regression model	9 water quality variables
Lamrini et al. (2005)Lamrini, B., Benhammou, A., Le Lann, M.-V. and Karama, A., A neural software sensor for online prediction of coagulant dosage in a drinking water treatment plant. Transactions of the Institute of Measurement and Control, 27(3), 195-213 (2005).	Neural-fuzzy inference system	Seven online measurements of raw water and alum dosage.
Robenson (2009)Robenson, A., Shukor, S.A., and Aziz, N., Development of process inverse neural network model to determine the required alum dosage at Segama water treatment plant Sabah, Malaysia. Computer Aided Chemical Engineering, 27, 525-530 (2009).	Inverse neural network
Wu and Lo (2008)Wu, G.-D., and Lo, S.-L., Predicting real-time coagulant dosage in water treatment by artificial neural networks and adaptive network-based fuzzy inference system. Engineering Applications of Artificial Intelligence, 21(8), 1189-1195 (2008).	Levenberg-Marquardt neural networks	Raw water turbidity and coagulant dosage on day t − 1
Zhang et al. (2013)Zhang, K., Achari, G., Li, H., Zargar, A., and Sadiq, R., Machine learning approaches to predict coagulant dosage in water treatment plants. Int. J. Assur. Eng. Manag., 4(2), 205-214 (2013).	K-nearest neighbors (KNN)	3 water quality variables 966 samples collected between 2008 and 2010

	Parameter	Minimum	Maximum	Average	S.D.	Skewness	Kurtosis	Unit
Raw water	pH_r	5.7	8.5	6.5	0.4	1	3.3	-
	col_r	11	696	132	86	2.2	8.3	(HU)
	ss_r	2.0	84,0	9,3	7,9	4.9	36.4	(mg/L)
	turb_r	0.5	38,0	11,5	7,3	1.3	1.9	(NTU)
	alka_r	8	24	13	3	0.9	0.6	(mg/L)
Dosage of chemical agents	NaOH_in	0.0	35,0	9,3	6,3	0.9	0.5	(mg/L)
Dosage of chemical agents	Sulf_in	5.0	75,0	33,9	10,9	0.6	-0.7	(mg/L)
Treatedwater	pH_t	4.5	7,2	6,1	0,3	-0.7	3.8	-
	col_t	1	102	8	6	6.3	73.7	(HU)
	ss_t	1.0	12.0	1.5	1.1	3.6	17.7	(mg/L)
	turb_t	0.1	9,1	0,8	0,6	5.5	56.6	(NTU)
	alum_t	0.000	1.000	0.091	0.002	6.1	50.8	(mg/L)

Model	Inputs		Output
(I)	Process inputs	pH_r, col_r, turb_r, ss_r, alka_r, NaOH_in	Sulf_in
(I)	Process outputs	col_t, turb_t, ss_t, alum_t	Sulf_in
(II)	Process inputs	pH_r, col_r, turb_r, ss_r, alka_r, Sulf_in	NaOH_in
(II)	Process outputs	col_t, alum_t	NaOH_in

Principal component	Variance (%)	Cumulative variance (%)
1	34.7	34.7
2	19.4	54.1

	pH_r	col_r	ss_r	turb_r	alka_r	NaOH_in	Sulf_in	pH_t	cor_t	ss_t	turb_t
col_r	-0.129
ss_r	-0.018	0.772
turb_r	0.032	0.880	0.732
alka_r	0.051	-0.155	-0.066	-0.153
NaOH_in	-0.269	0.341	0.225	0.259	-0.12
Sulf_in	-0.231	0.255	0.174	0.156	-0.056	0.837
pH_t	0.129	0.047	-0.038	0.066	0.046	0.114	-0.038
cor_t	0.108	0.451	0.356	0.435	-0.11	0.014	-0.054	0.065
ss_t	-0.16	0.164	0.213	0.053	-0.036	0.113	0.156	-0.006	0.119
turb_t	0.164	0.511	0.435	0.502	-0.13	-0.005	-0.059	0.072	0.815	0.129
alum_t	0.075	-0.027	0.101	-0.019	-0.04	-0.156	-0.045	-0.134	0.287	0.170	0.284

Model	R²	R²_adj	MAE	RMSE
I	0.77	0.77	3.79	5.17
II	0.81	0.81	1.99	2.68

Uncertainty	R²	R²_adj	MAE / (mg/L)	RMSE / (mg/L)
0.5u_c	0.78	0.77	3.73	5.10
u_c	0.78	0.77	3.69	5.10
2u_c	0.76	0.75	3.82	5.33

Uncertainty	R²	R²_adj	MAE / (mg/L)	RMSE / (mg/L)
0.5u_c	0.82	0.82	1.98	2.66
u_c	0.85	0.85	1.82	2.42
2u_c	0.87	0.87	1.82	2.42

Brazilian Society of Chemical Engineering Rua Líbero Badaró, 152 , 11. and., 01008-903 São Paulo SP Brazil, Tel.: +55 11 3107-8747, Fax.: +55 11 3104-4649, Fax: +55 11 3104-4649 - São Paulo - SP - Brazil
E-mail: rgiudici@usp.br

Acompanhe os números deste periódico no seu leitor de RSS

[1] Corresponding author: Tel.: +55 71 3283-9800, Fax.: +55 71 3283-9802 - Email address: karlaesquerre@ufba.br