Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
208515 | Fuel | 2006 | 9 Pages |
Several physical and chemical parameters (such as saturates, aromatics, resins, and asphaltenes, element contents, density, viscosity, and carbon residue) are necessary to characterize residual oils. The combined use of mid-infrared (MIR) attenuated total reflection (ATR) spectroscopy and multivariate calibration allows those parameters to be estimated accurately. In order to improve the prediction results, samples from different processing units require different calibration models relative to the spectral similarities. This paper builds a strategy to classify and discriminate different types of residual oils by use of partial least square regression. The calibration models for the physical and chemical parameters of three types of residual oils were developed, respectively. The consistencies between the MIR predicted and reference values testify to the creditability of the proposed method.