| Complex solution is a very common material form in production and life,involving food safety,human health,environmental protection and other aspects.The component detection of complex solution has always been one of the hot issues concerned by scholars,but it is difficult to achieve accurate measurement because of the variety of its components.At present,the common method used in component analysis of complex solution is spectral analysis technology,and how to achieve more accurate component detection is the main direction of research.The"M+N"theory is a method about the source of error in spectral analysis,and its elimination and utilization.The multi-dimensional multi-mode method provides a new way to improve the accuracy of spectral analysis by increasing the content of spectral information.A multi-band spectrua acquisition device was built and used for clinical data acquisition.After centrifugation of the collected blood samples,the plasma was taken out and its ultraviolet,visible and near-infrared transmission spectra were collected.Independent modeling of each band and joint modeling of multi-band are carried out.The experiment results show that the prediction set correlation coefficient of the latter is about 0.1 higher than the single band with the best effect,and the root mean square error is reduced up to 87%.On the basis of the multi-band spectral device,a 365nm ultraviolet LED light source was added and dual-mode spectral data was collected.The fluorescence and transmission spectra were modeled separately,and the dual-mode spectra were joint modeled.The results show that the correlation coefficients of the latter predicted set were about 0.2 and 0.1 higher than those of the former,and the root mean square error of the predicted set was reduced by 31.98%and 23.49%respectively.The fluorescence spectra have good selectivity,but also have strong nonlinearity.Based on the"M+N"theory,a method of two-position excitation fluorescence spectrum modeling is proposed by using the characteristics of fluorescence spectrum.The fluorescence spectra of the mixed solution of fat milk and C28H20S2O6Na2 were measured at two locations,and the fluorescence spectra at two locations were respectively modeled and the two-position spectra were jointly modeled.The results show that the root-mean-square error of dual-position fluorescence spectrometry is 9.1%lower than that of traditional single-position method.Based on the multi-dimensional multi-mode method of"M+N"theory,this paper puts forward the methods of multi-band spectrum combined modeling,dual-mode spectrum modeling and dual-position fluorescence spectrum modeling.It is proved by experiment that these method can effectively improve the accuracy of complex solution component analysis,and the"M+N"theory is of great significance to the spectral analysis of complex solutions. |