Application of GA-DS to Calibration Transfer of Aviation Fuel Density in Near Infrared Spectroscopy

摘要

Direct standardization (DS) algorithm, a familiar calibration transfer algorithm in near infrared (NIR) spectroscopy, needs excessive transfer samples and has the disadvantage of the supercorrection during transfer. In order to simplify the process of DS, a calibration transfer algorithm based on genetic algorithm (GA) was presented. The characteristic NIR wavelength was selected by GA first. Then the differences among instruments were corrected by DS. A partial least squares model of aviation fuel density was transferred successfully from one instrument to another three ones with only five (or six) transfer samples. Precision of predication results with standardization model is nearly as high as that of calibration model and is higher than that of DS model. Compared with traditional methods of DS algorithm, the GA-DS algorithm can not only increase the model precision, but also compress the NIR data to make the NIR model development simplified.