Intelligent real-time quantification of cheese whey in rivers and reservoirs in Madrid (Spain)

A computerized approach to quickly, easily, and efficiently detect discharged sewage of the cheese sector into rivers or reservoirs has been presented in this work. It is based on developing and/or training linear regression, linear multiple regression, and multilayer perceptron-based models using the data obtained from measuring different contaminated waters with a portable digital refractometer. This equipment measures the refractive index, Brix degrees, and salinity of aqueous samples. To carry it out, 451 samples were made by mixing five different waters with cheese whey from the Spanish "FaThrica de quesos de Villalon de Campos S.L.". The water samples used were distilled water and four non-drinkable waters taken from four different Spanish locations (the Valmayor Reservoir, the Mari Pascuala Lagoon, the Manzanares River and the Guadarrama River). The lowest mean absolute percentage error (0.79%) was achieved using three independent variables and a multilayer perceptron-based model. The validation of this approach has been done in two ways: k-fold cross-validation and internal validation. The outcome of this research is a new method for a fast detection of waste spillages into rivers and reservoirs.