Naslov
Prediction of groundwater hardness in slavonia using artificial neural network models

Autori
Šafranko, Silvija ; Vešligaj Turkalj, Jelena ; Romić, Željka ; Stanković, Anamarija ; Jokić, Stela ; Medvidović-Kosanović, Martina

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Book of Abstracts 8th International Scientific and Professional Conference Water for all / Habuda-Stanić, Mirna ; Lauš, Ivana ; Gašo-Sokač, Dajana ; Bušić, Valentina ; Stjepanović, Marija - Osijek : Prehrambeno tehnološki fakultet Sveučilišta Josipa Jurja Strossmayera u Osijeku, 2019, 166-166

ISBN
978-953-7005-59-7

Skup
8. međunarodni znanstveno-stručni skup: Voda za sve = 8th International Scientific and Professional Conference: Water for all

Mjesto i datum
Osijek, Hrvatska, 21.01.2019. - 22.01.2019

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
water hardness ; Slavonia ; artificial neural networks ; prediction

Sažetak
Water hardness is an important parameter for water quality determination and suitability for human consumption and agriculture purposes. Hard water is usually defined as water which contains calcium and magnesium salts principally in a form of bicarbonates, chlorides and sulfates with possible presence of ferrous ions. According to the literature, several epidemiological investigations have demonstrated the relation between risk for cardiovascular disease and other health problems with the use of hard water enriched with calcium and magnesium ions. Most water of the Slavonian region is classified as hard water which may lead to health, plumbing and structural issues. Therefore it is essential to investigate potential causes of water hardness to be able to quickly estimate water hardness parameter. This could be useful for both convenient, and also for economic reasons due to expensive analytical equipment and materials. In this study, two artificial neural network models (ANNs) have been developed in order to predict water hardness in Slavonian region covering data recorded over the last five years, between 2014. and 2018. For that purpose, a feed- forward multilayer backpropagation neural network (FFBP-ANN) and radial basis function (RBF) neural network were created varying activation functions, the number of neurons in the hidden layer and spread constant for RBFNN model. The ANNs have been trained and tested on divided and normalized dataset in the range from -1 to 1 and from 0 to 1 in order to scale-up the inputs and output parameters. The overall performance of the developed ANN predictive models was evaluated based on the obtained mean squared error (MSE) and correlation coefficient (R) parameters. Determination of the best performing model was based on the AAD (Average absolute deviation) parameter. The obtained results showed the superior performance of FFBP-ANN model compared to RBFNN. However, both models were found to be useful tools for water hardness prediction.

Izvorni jezik
Engleski

Znanstvena područja
Kemija, Javno zdravstvo i zdravstvena zaštita, Biotehnologija u biomedicini (prirodno područje, biomedicina i zdravstvo, biotehničko područje)

POVEZANOST RADA