Estimación de la Demanda Bioquímica de Oxígeno (DBO5) en aguas residuales de las empresas de jeans de la ciudad de Pelileo utilizando redes neuronales artificiales

Abstract

Within artificial intelligence, the use of artificial neural networks is becoming increasingly important. This study seeks to estimate the error of the Biochemical Oxygen Demand in wastewater from laundries of jeans of the city of Pelileo using neural networks. First, a database consisting of 6 physical parameters, 9 chemical parameters and an-output variable was developed. These variables were selected using the TUSLA Standard and were compiled from the cadastre of the GAD Pelileo of the years 2017-2018 and a certified water analysis laboratory. Next, the Matlab software was used with the design of the Artificial Neural Network FeedForward Backpropagation with the input layer of 15 variables. The first hidden layer had ten neurons, the second summatory layer with one neuron and the output layer with the response variable corresponding to the estimation of the Biochemical Oxygen Demand with the Levenberg-Marquardt learning algorithm. Values of the correlation coefficient, the estimated network error, and the comparison of the Tukey Test were obtained, susch as: 0.98081; 0.8890; 0.9833. These figures reveal a concordance between the values estimated by the network and the actual values. Finally, the results showed that Biochemical Oxygen Demand was estimated numerically in wastewater through neural models. This type of model of neurons represents only part of the mathematical function that the network builds from the set of observations.