Apparatus for interactive self-modeling mixture analysis

Claim 2

Apparatus for controlling a process wherein starting material undergoes a reaction over a period of time to produce reaction products, said apparatus comprising: a) process spectrometer means operatively associated with said process for providing spectrographic data from said process; b) programmed computer means operatively connected to said process spectrometer means for analyzing said spectrographic data to determine pure variables relating to the reaction products and for comparing the determined pure variables to predetermined pure variables; and c) control means operatively associated with said process and operatively connected to said computer means for controlling the process in a manner determined by the result of the comparison.

Claim 3

Apparatus according to claim 1, wherein said computer means is programmed to analyze said spectographic data by performing the following steps: a) providing said spectrographic data of the material in the form of compound variables which have contributions from the components of the material; b) determining the purity values of said compound variables and presenting said purity values in the form of a first purity spectrum; c) finding the maximum of said purity values which is the first pure variable which has a contribution from only a first one of the components of said material; d) calculating a determinant-based weight function with respect to said first pure variable to eliminate the effect of said first pure variable; e) multiplying the first purity spectrum by said determinant-based weight function and presenting the results in the form of a second purity spectrum; f) finding the maximum of said purity values in said second purity spectrum which is the second pure variable which has a contribution from only a second one of the components of said material; and g) repeating steps d), e) and f) for successive determinant-based-weight functions with respect to preceding pure variables and presenting the results in the form of successive purity spectra for finding successive maxima to yield successive pure variables until the product of a determinant-based weight function and a purity spectrum does not have spectral features thereby indicating that no additional pure variables exist.

Claim 4

Apparatus according to claim 2, wherein said step of determining purity values comprises calculating the mean and the standard deviation of the magnitudes of said compound variables and calculating the ratio of the standard deviation to the mean for each of said compound variables.

Claim 5

Apparatus according to claim 3, further including a step of displaying the calculated standard deviation values in the form of a spectrum.

Claim 6

Apparatus according to claim 3, wherein a noise correction is applied to the purity calculation.

Claim 7

Apparatus according to claim 3, wherein a noise correction term is added to the denominator in the ratio of standard deviation to the mean calculated for each of said compound variables

Claim 8

Apparatus according to claim 6, further including a step of displaying the noise corrected purity values.

Claim 9

Apparatus according to claim 2, further including a step of displaying selected ones of said purity spectra.

Claim 10

Apparatus according to claim 2, wherein said step of determining purity values includes correcting for noise.

Claim 11

Apparatus according to claim 2, wherein said determinant-based weight function has a value of zero for the first pure variable and a high value for the other pure variables.

Claim 12

Apparatus according to claim 2, wherein said repeating steps d), e) and f) said successive determinant-based weight functions have zero values for preceding pure values and high values for remaining pure variables.

Claim 13

Apparatus according to claim 2, wherein a noise correction is applied to the calculation of the determinant-based weight function.

Claim 14

Apparatus according to claim 2, wherein said pure variables determined are used with said spectral data of the mixture for obtaining the components corresponding to the pure variables.

Claim 15

Apparatus according to claim 13, wherein said step of obtaining spectra of pure components is performed by a least squares method.

Claim 16

Apparatus according to claim 13, further including calculating the concentration of the components of said mixture using the spectra of the pure components with the spectral data of the mixture.

Claim 17

Apparatus according to claim 15, wherein said step of calculating the concentration of the components is performed using a least squares method.

Claim 18

Apparatus according to claim 1, wherein said computer means is programmed to analyze said spectrographic data by performing the following steps: a) providing said spectrographic data of the material in the form of compound variables which have contributions from the components of the material and which have intensities or magnitudes; b) placing said spectrographic data in a data matrix D having size v*c wherein v is the number of compound variables comprising the rows of the matrix and c is the number of spectra comprising the columns of the matrix and wherein the elements of the matrix are the intensities or magnitudes of the compound variables; c) calculating for said data matrix D the mean ##EQU14## and compound variable length ##EQU15## wherein d is each data value in said matrix D in row i column j; d) calculating noise correction terms α=(N/100)*Max(μ i ) and β=(N/100) * Max(λi) where N is a predetermined noise correction value; e) calculating a first determinant-based weight function W i ,1 =(λi/(λi+β)) 2 ; f) calculating purity spectrum values p i ,j=w i ,j *σ i /(μ i +α) and standard deviation spectrum values S i ,j =W i ,j *σ i where j=P; g) determining the variable with maximum purity value using Pk=i for Max i P i ,j where k=1 which is the first pure variable; h) displaying a selected one or both of the first purity spectrum and standard deviation spectrum; i) calculating a correlation around the original matrix COO by first scaling the data matrix D by d i ,j (λ,β)=d i ,j /(λ i +β) and then calculating the matrix C=(1/c) D (λ,β) D (λ,β) T thereby giving all the compound variables an equal contribution in subsequent calculation of a determinant-based weight function; j) calculating a second determinant-based weight function ##EQU16## where j=P; k) calculating purity spectrum values p i ,j =W i ,j *σ i /(μ i +α)l and standard deviation spectrum values S i ,j =W i ,j * σ i using said second determinant-based weight function; l) determining the variables with maximum purity value using P k =i for Max ipi ,j ; and m) displaying selected ones or both of additional purity spectra and standard deviation spectra.

Claim 19

Apparatus according to claim 17 wherein said step of utilizing the pure variables comprises the steps of: a) calculating resolved spectra by a least squares method using pure variable information and the spectral data of the mixture; b) normalizing the resolved spectra for total intensity; c) calculating resolved concentrations by a least squares method using the normalized resolved spectra, the pure variable information and the spectral data of the material; and d) displaying the normalized resolved spectra and the resolved concentrations.