Method and apparatus for on-line measurement of polymer properties

1. A method for measuring a property of a test sample comprising:

a) obtaining a set of measured spectra for a set of calibration samples, and performing constrained principal spectral analysis on said calibration samples to provide a corrected set of measured spectra,

b) providing said test sample having an unknown value of said property,

c) applying light to said test sample resulting in reflected light,

d) detecting said reflected light,

e) transmitting said reflected light to a spectrometer,

f) measuring a spectrum for said test sample,

g) performing constrained principal spectral analysis on said test sample spectrum to provide a corrected test sample spectrum,

h) performing locally weighted regression analysis by selecting a subset of said corrected calibration sample spectra and building a regression model based upon said selected subset, and

i) predicting a value for said unknown property of said test sample using said corrected test sample spectrum and said regression model.

2. The method of claim 1 , wherein said property is selected from the group consisting of moisture content, ethylene content, Epoxidized Soy Bean Oil (ESBO) content, calcium stearate content, and ethylidene norbornene (ENB) content.

3. The method of claim 1 , wherein said test sample comprises a polymer.

4. The method of claim 1 , wherein said test sample comprises a rubber.

5. The method of claim 1 , wherein said measured spectra for said set of calibration samples comprise reflectance spectra in the near infrared region.

6. The method of claim 1 , wherein said test sample spectrum comprises reflectance spectra in the near infrared region.

7. The method of claim 5 or 6 , wherein said measurement of near infrared spectra is made by a Fourier Transform Near Infrared (FTNIR) spectrometer.

8. The method of claim 1 , wherein said measurement of said spectrum for said test sample is performed at least once every minute.

9. A method for online control of a process to produce a product with a property P and having a desired value D, said method comprising:

a) obtaining a set of measured spectra for a set of calibration samples,

b) performing constrained principal spectral analysis on said set of measured spectra for said set of calibration samples to produce corrected spectra for said set of calibration samples,

c) providing a test sample having an unknown value of said property P,

d) applying light to said test sample resulting in reflected light,

e) detecting said reflected light,

f) transmitting said reflected light to a spectrometer,

g) measuring a spectrum for said test sample,

h) performing constrained principal spectral analysis on said test sample spectrum to produce a corrected test sample spectrum,

i) performing locally weighted regression analysis by selecting a subset of said corrected calibration sample spectra and building a regression model based upon said selected subset,

j) predicting a value for said property P of said test sample using said corrected test sample spectrum and said regression model,

k) adjusting the process parameters based upon the difference in the predicted value of said property P and desired value D.

10. The method of claim 9 , wherein said subset is selected based upon at least one of classification analysis and cluster analysis.

11. The method of claim 9 , wherein said property is selected from moisture content, ethylene content, Epoxidized Soy Bean Oil (ESBO) content, calcium stearate content, and ethylidene norbornene (ENB) content.

12. The method of claim 9 , wherein said test sample is a polymer.

13. The method of claim 9 , wherein said test sample comprises a rubber.

14. The method of claim 9 , wherein said measured spectra for said set of calibration samples comprise reflectance spectra in the near infrared region.

15. The method of claim 9 , wherein said test sample spectrum comprises reflectance spectra in the near infrared region.

16. The method of claim 9 , wherein said measurement of near infrared spectra is made by a Fourier Transform Near Infrared (FTNIR) spectrometer.

17. The method of claim 9 , wherein said measurement of said spectrum for said test sample is performed at least once every minute.

18. A method for online detection of a change from a first polymer grade to a second polymer grade in a polymer finishing process comprising:

a) obtaining a set of measured spectra for a set of calibration samples,

b) performing constrained principal spectral analysis on said set of measured spectra to produce corrected spectra for said set of calibration samples,

c) assigning a grade type to each said corrected spectrum of said set of calibration samples,

d) applying light to a polymer sample in said polymer finishing process resulting in reflected light,

e) detecting said reflected light,

f) transmitting said reflected light to a spectrometer,

g) measuring a spectrum for said polymer sample,

h) performing constrained principal spectral analysis on said polymer sample spectrum to provide a corrected polymer sample spectrum,

i) performing K-nearest neighbor analysis on said corrected polymer sample spectrum comprising selecting a subset of said corrected calibration spectra based upon the similarity of said corrected subset spectra to said corrected polymer sample spectrum, and counting the assigned grade type of each said corrected calibration spectrum in said subset,

j) assigning a grade type to said polymer sample based upon said count, and

k) comparing said assigned grade type to said polymer sample with said first polymer grade.

19. The method of claim 18 , wherein said measurement of near infrared spectra is made by a Fourier Transform Near Infrared (FTNIR) spectrometer.

20. A method for online control of a polymer finishing process resulting from a change from a first polymer grade to a second polymer grade comprising:

a) obtaining a set of measured spectra for a set of calibration samples,

b) performing constrained principal spectral analysis on said set of measured spectra to produce corrected spectra for said set of calibration samples,

c) assigning a grade type to each said corrected spectrum of said set of calibration samples,

d) providing a test sample having an unknown grade type,

e) applying light to said test sample resulting in reflected light,

f) detecting said reflected light,

g) transmitting said reflected light to a spectrometer,

h) measuring a spectrum for said test sample,

i) performing K-nearest neighbor analysis on said test sample spectrum comprising selecting a subset of corrected calibration spectra from said set of corrected calibration spectra based upon the similarity of said subset spectra to said test sample spectrum, counting the assigned grade type of each said corrected spectrum in said subset, and

j) assigning a grade type to said test sample based upon the comparison to said subset.

21. The method of claim 20 , wherein said measurement of near infrared spectra is made by a Fourier Transform Near Infrared (FTNIR) spectrometer.

22. The method of claim 20 , further comprising:

k) altering the conditions of said finishing process for said second polymer grade.