Agricultural product element quantitative detection model building method based on X-ray fluorescence analysis

Abstract

The invention discloses an agricultural product element quantitative detection model building method based on X-ray fluorescence analysis. The method comprises the following steps: first, conducting pre-processing on an obtained sample spectrogram to obtain a standard spectrogram, wherein the pre-processing particularly comprises differential processing, normalization, multiplicative scatter correction (MSC), centralization, standard normal variable exchange (SNV), light spectrum smoothing and the like; then, carrying out sample set partition and abnormal sample rejecting through a principal component analysis (PCA) and partial least squares (PLS ) data processing method, so as to obtain a calibration set and a forecast set; finally, conducting calibration of indexes to be detected on agricultural products through principal component analysis and artificial neural network (PCA+ANN), as well as partial least squares and artificial neural network (PLS+ANN). The method adopts a peak value method in model building, so as to solve the problem of overlapping in application of the conventional peak area model building method, and improve the accuracy of a sample calibration model.

Description

technical field [0001] The invention relates to an X-ray fluorescence spectrum analysis technology, in particular to a peak method modeling of the X-ray fluorescence spectrum. Background technique [0002] Most traditional energy dispersive X-ray fluorescence spectrometers use the peak area method when establishing statistical models. The peak area method has the following advantages: it is not necessary to accurately know the injection volume of the sample, and the influence of a slight change in the operating conditions on the test results is relatively small, the calculation is convenient, and it is suitable for simultaneous analysis of multiple elements. [0003] Based on the principle of simulated annealing algorithm, Zeng Guoqiang et al. established a new peak-finding model algorithm. This algorithm uses simulated annealing to find the convergence characteristics of the global optimum, uses the Metropolis criterion as the basis for peak-valley judgment, and introduces ...

Images