Inference of the biodiesel cetane number by multivariate techniques

In this work, we have implemented a method which uses structural information from 1H NMR spectra of fatty esters and biodiesels to infer the corresponding Cetane Number (CN). The method consists of the successive application of Principal Component Analysis (PCA), Fuzzy Clustering and a feed-forward Artificial Neural Network (ANN) to the data set. PCA recognized redundant information, and determined the number of clusters for subsequent Fuzzy Clustering classification. At the final stage ANN used membership values from the Fuzzy Clustering process as inputs to predict the cetane number of different types of biodiesel (complex mixtures) from data of pure substances (esters). Root-mean-square deviations were in the range of 0.2–2.4.