Identification and detection of organophosphorus pesticides using artificial neural networks based on spectra obtained in gas chromatography

Background: The analysis associated with detection of compounds from spectra is a process in which the results tend to be a tentative identification of compounds. Different factors as including technical, methodology, environmental and climatic effects processes affect the detection method in its precision, separation, quantification and characterization of compounds. This process can be tedious and complex due to the number of existing compounds on existing databases. Objective: Training an artificial neural network to identify three main organophosphate pesticides, based on different chromatographic graphs obtained in a preliminary study. Conclusions: In training, the algorithm corrects the worst error converging on the optimal solution from the fourth iteration with final error of 0.0218 in 10th iteration. The total error of the test was 0.0212, even lower than the value obtained in the training. The algorithm developed in this paper decreased by up to 65% the initial error, obtaining as total mean squared error of 0.0218 value in its execution for 10 iterations. The development of this study identified three types of pesticides organofosfarados with a total error test 0.0212, from parameters such as retention time, intensity and concentration.