Application of artificial neural networks for conformity analysis of fuel performed with an optical fiber sensor

摘要: The liquid fuel quality control is an important issue that brings benefits for the State, consumers and environment. conformity analysis, in special gasoline, demands a rigorous sampling technique among gas stations other economic agencies, followed by series of standard physicochemical tests. Such procedures are commonly expensive time demanding and, moreover, specialist often required to carry out tasks. drawbacks make development alternative analysis tools research field. refractive index additional parameter help besides prospective optical fiber sensors, which operate like transducers with singular properties. When this correlated sample density, it becomes possible determine zones cannot be analytically defined. This work presents application artificial neural networks based on Radial Basis Function these zones. A set 45 gasoline samples, collected several previously analyzed according rules Agencia Nacional do Petroleo, Gas Natural e Biocombustiveis, Brazilian regulatory agency, constituted database build two networks. input variables first network samples indices, measured Abbe refractometer, density digital densimeter. For second included, densities, wavelength response long‐period grating indices. used was written using point‐to‐point submitting consecutive electrical arcs from splice machine. output both Networks represented status each sample, report tests carried following American Society Testing Materials and/or Association Technical Rules standards. subset 35 randomly chosen database, design calibrate (train) topologies (numbers neurons hidden layer function radius) were built order minimize root mean square error. composed 10 validate final architectures. obtained results have demonstrated reach good predictive capability.