Characteristic wavelength selecting method for near infrared spectrum in ant colony optimization algorithm

Abstract

The invention provides a characteristic wavelength selecting method for a near infrared spectrum in an ant colony optimization algorithm. The method comprises the following steps: creating a partial least squares analysis model by utilizing all wavelength points of the near infrared spectrum as initial selection equivalent variables of the ant colony optimization algorithm and utilizing the quality or characteristics of a test object as reference standard, re-weighing and calculating to update a pheromone vector according to the predicated mean square error of the model, searching and obtaining the optimal near infrared spectrum wavelength combination through using iterative computations, performing circular computation for a plurality of times, and automatically judging so as to obtain the optimal characteristic wavelength of the near infrared spectrum. The characteristic wavelength selecting method provided by the invention adopts the global search and positive feedback mechanisms of the ant colony optimization algorithm so as to effectively avoid the defects of subjective wavelength selection in the modeling process, so that the model has strong robustness and applicability.

Description

technical field [0001] The invention relates to the technical field of near-infrared spectrum analysis, in particular to a method for selecting near-infrared spectrum characteristic wavelengths using an ant colony optimization algorithm. Background technique [0002] According to the definition of ASTM, the near-infrared spectrum region refers to electromagnetic waves with a wavelength in the range of 780-2526nm. It is a molecular vibration spectrum frequency multiplication and combined frequency absorption spectrum. It has rich structure and composition information and can be used for the measurement of the composition and properties of hydrocarbon organic substances. . Compared with traditional analysis techniques, it has the advantages of non-destructive testing, high analysis efficiency, low cost, good reproducibility, sample measurement generally does not require pretreatment, and is suitable for on-site testing and online analysis. With the rapid development of near-i...

AI Technical Summary

Problems solved by technology

However, the multiplied and combined frequency absorption signals of substances in this spectral region are weak, and the spectral bands overlap, making the analysis complicated; and because the data collected by the instrument contains other irrelevant information and noise besides the information of the sample itself, such as electrical noise, sample It is difficult to eliminate all these information in preprocessing; secondly, the information of samples in some spectral regions is very weak, and the degree of correlation with the composition or properties of samples is not high; in addition, there is a collinear relationship in the spectral data of the same sample, which is easy to generate data redundancy

If all these data are involved in modeling, not only the amount of calculation is large, the model is complex, but the accuracy is also affected

Images