Abstract (EN):
This work focuses on the prediction of research and motor octane numbers of gasolines throughout multivariate statistical analysis, which may significantly increase the celerity of the quality control process. The aim of this work is to compare the performance of two different multivariate models, based on partial least squares regression and artificial networks. The results show that both models predict octane numbers with accuracy, presenting coefficients of determination above 0.95 for the calibration data set. For the test data set, partial least squares model is more accurate, which might indicate the existence of linear correlations between spectral data and octane numbers. The statistical parameters also show that the research octane number prediction is more accurate than the motor octane number.
Idioma:
Inglês
Tipo (Avaliação Docente):
Científica
Nº de páginas:
6