Artificial neural network modeling of diuron and irgarol-based HPLC data and their levels from the seawaters in Izmir, Turkey

Abstract

ABSTRACTThe LC determination of two well-known antifouling booster biocides, diuron and irgarol, was investigated from the seawaters in İzmir, Turkey. The biocide levels were pre-concentrated through C18 solid-phase extraction cartridges and they were analyzed by the LC-UV method. An artificial neural network (ANN) was used to model the data obtained from LC optimization. Column temperature, percentage of acetonitrile, flow rate, wavelength, pH, and concentration of biocides were used as input parameters. The retention time was selected as output parameter. The best back-propagation algorithm in ANN modeling for diuron and irgarol was found to be the Levenberg–Marquardt algorithm. The limits of detection for diuron and irgarol were calculated as 25.38 and 39.49 ng L−1, respectively. The inter-day and intra-day precisions were obtained less than 13.5% for each biocide. The recovery rate for diuron was 96.9% and for irgarol it was 84.6%. The maximum diuron and irgarol levels were measured as 1779 ng L−1 and 908 ng L−1, respectively. In conclusion, ANN is a robust modeling method to predict the retention time in LC studies. Since diuron and irgarol have been detected in Turkish waters, it is therefore suggested that booster biocides with less impact on the environment should be used in antifouling paint formulas.