Abstract:

More than one thousand diesel samples were collected to establish near infrared spectroscopic database for determination of Cetane Number. On the research, disposable headspace bottle was used to satisfy the spectral fast acquisition. By adding a small amount of samples to the database could get suitability model in Maoming petrochemical. Robust and uniform calibration models were developed based on different types of diesel by partial least squares method，SVM and LSSVM algorithm. The standard error of prediction on diesel oil samples using PLS, SVM and LSSVM calibration model were 1.6, 1.4 and 1.3, respectively. The work is provided with advantages such as high-speed, saving modeling cost and low workload.

Key words: NIR, diesel, cetane number, database, chemometrics