Instrument-to-Instrument Calibration Transfer And Standardization Of Spectra

Spectral analysis is a classical detection technology with the characteristics of fast response, non-destruction, low cost and safety. It has been used widely in the oil industry, food industry, chemical engineering and biological industry. Meanwhile, the robustness of calibration models and the standardization of spectra are critically complex and important. Based on the deep understanding of Raman and near-infrared (NIR) spectral features and the existing methods of calibration transfer, several improved calibration transfer and standardization algorithms are proposed. Besides, a fast analyzer for methanol gasoline is developed. The main contents are as follows:1. To implement calibration transfer between Raman spectrometers, an improved piecewise direct standardization (PDS) is proposed. Standard normal variate (SNV) is firstly introduced to reduce the influence of spectral background and intensity corresponding to the master spectrometer and the slave spectrometer; then PDS algorithm is used to eliminate the differences between Raman spectra for a specific sample. Moreover, a new quantitative criterion, called transfer error rate, is proposed to evaluate the performance of calibration model transfer. This improved PDS is applied to Raman spectral analysis of gasoline. The results show that the proposed algorithm can obtain high transfer accuracy and strong model robustness, and only a small quantity of transfer samples is needed.2. A new calibration transfer algorithm is proposed to eliminate the differences between various NIR instruments. A linear baseline is firstly removed with detrend algorithm, then PDS algorithm is used to eliminate the differences between NIR spectra for a specific sample, which is also called DT-PDS. It needs no pre-processing which preserves the original spectral information. The experimental results prove its validness for calibration transfer of NIR spectra with high transfer performances. 3. For various spectra of methanol gasoline samples from different base gasoline, a new standardization algorithm is proposed. The standardized spectra are applied to build the calibration model to predict the methanol concentration in unknown methanol gasoline samples with the range from 0% to 100%. The standardization algorithm can predict the methanol concentration accurately and remove the disturbances from the component variation of base gasoline.4. An NIR analyzer for methanol gasoline has been developed to meet the sharp need of fast detection for methanol gasoline. And this new version has been used by several enterprises to detect methanol gasoline. Application results show that this new analyzer can predict the interested properties in methanol gasoline accurately as well as be easy to be operated and maintained.