A compound rapid assay method for crude oil properties based on similarity discrimination and partial least squares

The measurement of crude oil properties is of paramount importance to oil processing in refinery plants. Although some methods for the rapid assay of crude oil properties have been proposed, the measurement accuracy is still not high enough in some cases. A compound rapid assay method for crude oil properties based on similarity discrimination and partial least squares (PLS) is put forward in this paper. First, we apply similarity discrimination method in which we calculate the spectral distance between the calibration crude oil samples in database and the oil sample under test. Then we count the number of calibration samples of which spectral distance is less than a threshold. Finally, we determine the credibility of similarity discrimination according to the number of these calibration samples. If the similarity discrimination prediction is considered credible, the modeling method will be employed to measure the properties of the unknown sample, otherwise properties will be measured by PLS. The test results demonstrate that the proposed method is not only convenient and practical, but also can provide enhanced prediction accuracy of crude oil properties.