Non-destructive Detection Of Complex Solution In Flexible Container Based On Spectral Analysis Technology

The complex mixed solution is a common material form in production and life,and it is widely used in industries,agriculture,environmental protection,food,and biomedicine.With the continuous development of the plastic industry,flexible containers have gradually become the basic packaging form of "liquid products".More and more researchers are beginning to explore suitable methods to directly perform non-destructive analysis of complex solutions in flexible containers.Among them,the spectral analysis technology is widely used in the analysis of complex solutions in flexible containers due to its fast and non-destructive advantages.However,this method still has many technical challenges,such as the difference of flexible containers,the strong scattering of complex solutions,and the overlap of the absorption spectrum of its components.Effectively suppressing the interference of the above factors is the key to improving the accuracy of spectral analysis.This paper starts from three stages of spectral data collection,spectral data processing and spectral data modeling,and studies the non-destructive spectral analysis of complex solutions in flexible containers.The specific work is as follows:(1)Starting from the Beer–Lambert law,the multi-path spectra acquisition model of the complex solution in the flexible container is analyzed,and a differential opticalpath spectrum method is to suppress the influence of the flexible container difference on the spectral analysis.(2)Aiming at the characteristics of visible-near-infrared spectroscopy,a feature extraction method based on Gaussian-wavelet transform is proposed.This method extracts the most effective information related to the target substance in the spectrum data and suppresses the influence of external factors such as flexible containers and instrument noise on the spectrum analysis.(3)Based on the traditional ant colony wavelength optimization method,an improved ant colony optimization algorithm is proposed to select the wavelength of the spectral data.Obtain the wavelength that has the greatest correlation with the target component and is insensitive to flexible containers and non-target components,to suppress the influence of the differences in flexible containers and non-target components on the spectrum analysis.(4)In order to suppress the influence of nonlinearity caused by the strong scattering of complex solutions on spectral analysis,a convolutional neural network model suitable for visible-near infrared spectroscopy regression analysis is proposed to improve the non-destructive analysis accuracy of complex solutions in flexible containers.And the model was applied to the non-destructive evaluation of the hemolysis degree of bagged blood,which verified the superiority of the nonlinear model.Several specific methods were proposed in this thesis from the aspects of suppressing the difference of flexible containers,reducing instrument noise and nontarget component interference.The effectiveness of the proposed methods was verified by using liquid phantoms and actual samples as the research objects in succession.Finally,combining the above methods with the convolutional neural network model proposed in this thesis,correct the nonlinearity between the spectral data and the concentration of the measured substance,and improve the accuracy of spectral analysis.Taking bagged blood containing free hemoglobin as the research object,the content of free hemoglobin was quantitatively analyzed,and the root mean square error of the prediction set was as low as 73.5 mg/L.The research in this thesis not only provides references for solving the problems in the spectral analysis technology of complex solutions in flexible containers and also lays a foundation for the non-destructive evaluation of the hemolysis degree of bagged blood into clinical applications.