Independent Component Analysis In The Application Of Infrared Spectrum Data In Research

Energetic materials synthesis process is complex,if we can separate the pure spectral signal,the relative concentration information from the spectral data of the mixture (the mid-infrared spectral data),the separated spectral pure signal and the relative concentration of information as independent as possible,it may be more accurate to infer the unknown compounds to deduce the reaction mechanism of energetic materials synthesis process.Chemical data mining is one of the key issues for cheminformatics research,therefore data mining techniques can be applied to organic chemistry synthetic reaction mechanisms,organic synthesis process,optimization analysis and high-throughput screening to solve a typical knowledge discovery problem.In order to perceive the energetic materials synthesis process,we employ the data mining method-independent component analysis to investigate the synthesis process of the mid-infrared spectral data.On this basis,this paper presents fast fixed-point algorithm which based on principal component analysis and joint approximate diagonalization of eigenmatrices algorithm which based on principal component analysis.We applied fast fixed-point algorithm which based on principal component analysis to analyze the mid-infrared spectral data,and separated concentration information is able to reflect the changes of substances,but its spectral signal can’t provide the accurate peak place value of substances.Then,we applied joint approximate diagonalization of eigenmatrices algorithm which based on principal component analysis to analyze the mid-infrared spectral data,and separated spectral signal can provide the accurate peak place value of substances,but concentration information is not able to reflect the changes of substances.On the analysis of the merits of these two algorithms,we presented a new hybrid algorithm-fast fixed-point algorithm and joint approximate diagonalization of eigenmatrices algorithm hybrid algorithm,and hybrid algorithm used to separate the infrared spectral data,ultimatly the simulation results verified the the feasibility and effectiveness of the proposed method,compared to traditional multivariate curve resolution-alternating least squares algorithm more substances can be obtained form this new hybird algorithm.