Naslov
Application of Artificial Neural Networks for Method Development in Ion Chromatography

Autori
Bolanča, Tomislav ; Cerjan-Stefanović, Štefica ; Novič, Milko

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

Izvornik
Sixteenth Annual International Ion Chromatography Symposium / - San Diego (CA), 2003, 77-77

Skup
16th Annual International Ion Chromatography Symposium

Mjesto i datum
San Diego (CA), Sjedinjene Američke Države, 21.09.2003. - 24.09.2003

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
ion chromatography; artificial neural networks; retention model; optimization; wastewater

Sažetak
In this work the possibilities of application of artificial neural networks for method development in ion chromatography has been investigated. By using experimental design methods the experimental part of the work was planed, and obtained experimental results were used for development of the retention model in ion chromatography. Following chromatography set-up was applied: Dionex DX500 chromatography system (Sunnyvale, CA, USA) equipped with quartenary gradient pump (GP50), eluent generator module (EG40), chromatography module (LC25) and detector module (ED40). Separation and suppressor columns used were Dionex IonPac AG15 (4x50 mm) guard column, IonPac AS15 (4x250 mm) separation column and ASRS-ULTRA-4mm suppressor column, working in recycle mode, were used, respectively. The sample-loop volume was 50  l. The whole system was computer controlled through PeakNet 5.1 software. Retention model was developed for analysis of inorganic anions: fluoride, chloride, nitrite, sulphate, bromide, nitrate, phosphate, and it includes following ion chromatographic parameters: eluent flow rate and eluent composition. Multi layered feed forward artificial neural networks trained with a Levenberg - Marquardt batch error back propagation algorithm was used for development of retention model. Following neural network parameters were optimized: number of hidden layer neurones, number of iteration steps and minimal number of experimental data in training set. Developed retention model was validated with external experimental data set (experimental results which are not used for models development) and used for optimization of ion chromatographic system. The program for neural network was made by the authors in the MATLAB environment (MATLAB 6.0, MathWorks, Sherborn, MA, USA). All the calculations were performed on the IBM compatible personal computer equipped with 2400 MHz Pentium IV processor, and 512 Mb RAM. In order to obtain good optimization results, several criteria separation functions were used, different in terms of mathematical formulations as well as in terms of chemical interpretations. Finally it is shown that optimized ion chromatographic method can be applied for the analysis of wastewater generated in fertilizer industry Peretrokemija d.d., Kutina, Croatia.

Izvorni jezik
Engleski

Znanstvena područja
Kemijsko inženjerstvo

POVEZANOST RADA