A New Wavelengths Selection Method For Molecular Spectra Based On Ant Colony Optimization And Fundamental Applications

Intellignent Plant and Industry 4.0 have become the strategic direction of future development.The intelligent solution is widely used in quality control and optimizing intermediate parameters of a industrial process.The molecular spectral analysis method has recently been a research hotspot in the area of online detection.It processes data and analyze model through spectrogram formatted by laser irradiation,in the mean time,it implements online detection.The advantages of this method includes strong operability,simple analysis platform build and quality optimization.The paper researched on several mainstream algorithms applied for each step in spectral analysis and proposed a new PLS-VIP-ACO wavelength screening method.The method I proposed is based on an improved Ant Colony Algorithm,which adds some important coefficients of projection variables into the pheromone update operation and differentiates Wavelength contribution rate on the basis of PLS regression coefficient.The new method is proven to be efficient according to standard data set near-infrared.I also did research on the application of Raman spectroscopy in new field of biodiesel blending oil.Within Raman spectral wave number screening,I applied my PLS-VIP-ACO method and find it suitable for Raman spectroscopy as well.The main contents of this paper can be summarized as follows:1.In this section,I did a full range of reading studies related to spectral analysis and introduced the significance of spectral analysis as a rapid,non-destructive analytical technique.I also introduced theoretical basis,advantages/disadvantages,historical development and research status of the application of near-infrared spectroscopy and Raman spectroscopy in detail.Furthermore,the significance of wavelength selection in spectral analysis and the research status of wavelength selection are also discussed in this section.2.This part introduces main methods of chemical metrology in spectrum analysis.Involved chemical methods are introduced in three explicit parts:data preprocessing,wavelength selection and multivariate calibration.3.First proposed the VIP-ACO algorithm,a method based upon Ant Colony Algorithm and updates the basis of information with the importance of projection variables.This method improves the accuracy and reduce the complexity of the model camparing to full spectrum modeling and other wavelength selection methods.Aiming at the shortcoming of the new method,randomness,I combine VIP-ACO algorithm with partial least squares model coefficients to distinguish advantages and disadvantages of the optimal wavelength combination in each iteration and prevent random doping independent wavelength points disturbing the iterative results.The results of the new PLS-VIP-ACO algorithm show a superior modeling method.4.I researched on application of spectral analysis in biodiesel oil blending ratio,simulated the spectral acquisition and detection device in the laboratory and analyzed spectral data.To test and verify the new method,I choose PLSR as a modeling method,which is stable,simple,and has good prediction effect.By applying new method to PLSR model along with other mainstream wavelength selection methods,I found the new method superior to other methods since it hightens the model accuracy and reduces the complexity.