论文部分内容阅读
When multivariate calibrations are regarded as a powerful tool of extracting chemical information from measurement data, the calibration transfer is indispensable for maintaining the predictive ability of a multivariate calibration model when the instrument or measurement conditions are altered.Therefore, a wide array of different methods found in literatures were dedicated to calibration transfer over the past years.However, almost all the existing methods achieve the goal to transferring the spectral responses from a secondary instrument to a primary one in the case of ignoring the variance from sensibilities of components to instrumental response.So the samples used in standardization set must be representative enough to describe the differences between the instruments, and should ideally span the entire experimental domain, and otherwise, it will lead to a certain deviation in multivariate calibrations.In this paper, a novel strategy is presented to eliminate the variance from sensibilities of components to instrumental response when standardizing two-way responses measured on two fluorescence spectrometers by combining piecewise directive standardization (PDS) with parallel factor analysis (PARAFAC).After PDS, the spectra of the prediction samples on the primary instrument are considered to be missing values.The PARAFAC algorithm is carried out on the four-way array consisting of the spectra of standardization samples and the missing spectra of the prediction samples on the primary instrument, and the PDS transferred spectra of standardization samples and prediction samples on the secondary instrument.And then, the property of concentration can be obtained.The performance of proposed method has been evaluated and compared with PDS, using different number of standardization samples.The finding is that the PDS/PARAFAC achieves better results compared with the PDS, and eliminates the effect of standardization sample number.The strategy is expected to be applicable for lower way data measured on spectrometers, such as Raman spectrometers, near infrared spectrometry, etc.