Raman spectroscopy and machine learning for quality control

1. A method for determining the quality of a polymer based on a Raman spectrum of the polymer, the method comprising:

obtaining, by a polymer property computing device, a Raman spectrum of a polymer sample from a Raman spectrometer, wherein the Raman spectrum comprises a plurality of peaks and valleys in chemical and structural fingerprints that correspond to one or more polymer properties or features;

computing, by the polymer property computing device, the one or more polymer properties or features by comparing the obtained chemical and structural fingerprints to stored polymer property and feature data, wherein the computing step is carried out by performing, by the polymer property computing device, one or more trained machine learning algorithms; and

determining, by the polymer property computing device, based on one or more computed polymer properties or features, whether the quality of the polymer sample meets a predetermined quality threshold,

wherein the one or more trained machine learning algorithms are trained by:

receiving, by the polymer property computing device, both a Raman spectrum and measured polymer properties and features for each of a plurality of known polymer samples;

correlating, by the polymer property computing device, one or more peaks and/or valleys in the structural and chemical fingerprints of the received Raman spectrum with the measured polymer properties and features to form at least part of the polymer property and feature data;

executing, by the polymer property computing device, one or more machine learning algorithms to learn to recognize the correlated one or more peaks and/or valleys in the structural and chemical fingerprints and to compute the polymer properties and features;

adjusting, by the polymer property computing device, one or more hyper-parameters to minimize any errors between the computed polymer properties and features and the measured polymer properties and features; and

repeating, by the polymer property computing device, at least the executing step and the adjusting step to form the one or more trained machine learning algorithms.

2. The method as set forth in claim 1 , further comprising, before the computing step, classifying, by the polymer property computing device, the peaks and/or valleys in the obtained chemical and structural fingerprints as either a relevant polymer property or feature or a non-relevant polymer property or feature.

3. The method as set forth in claim 1 , further comprising, before the computing step, transforming, by the polymer property computing device, the obtained Raman spectrum of the polymer sample to a format readable by one or more trained machine learning algorithms.

4. The method as set forth in claim 1 , further comprising, after the determining step, reporting, by the polymer property computing device, the one or more computed polymer properties or features.

5. The method as set forth in claim 1 , wherein the polymer sample comprises a polyolefin.

6. The method as set forth in claim 5 , wherein the polyolefin is polypropylene.

7. The method as set forth in claim 1 , wherein the polymer sample is in a form of a pellet.

8. The method as set forth in claim 1 , wherein the one or more polymer properties is one or more properties selected from the group consisting of molecular weight, melt flow rate, lamellar thickness, crystallinity, xylene solubles, one or more tensile properties, and one or more compressive properties.

9. The method as set forth in claim 1 , further comprising adjusting, by the polymer property computing device, a production process for the polymer based on the one or more computed polymer properties or features.

10. The method as set forth in claim 1 , further comprising enhancing, by the polymer property computing device, the one or more trained machine learning algorithms with one or more ensemble techniques.

11. The method as set forth in claim 1 , further comprising refining, by the polymer property computing device, the one or more trained machine learning algorithms at least by repeating the executing step and the adjusting step.

12. The method of claim 1 , wherein the one or more polymer features is one or more features selected from the group consisting of an amount of one or more additives and an amount of one or more comonomers.

13. A non-transitory computer readable medium having stored thereon instructions for determining the quality of a polymer based on a Raman spectrum of the polymer comprising machine executable code which when executed by at least one processor, causes the processor to perform steps comprising:

training one or more machine learning algorithms, wherein the training step comprises:

receiving both a Raman spectrum and measured polymer properties and features for each of a plurality of known polymer samples;

correlating one or more peaks and/or valleys in the structural and chemical fingerprints of the received Raman spectrum with the measured polymer properties and features to form at least part of the polymer property and feature data;

executing one or more machine learning algorithms to learn to recognize the correlated one or more peaks and/or valleys in the structural and chemical fingerprints and to compute the polymer properties and features;

adjusting one or more hyper-parameters to minimize any errors between the computed polymer properties and features and the measured polymer properties and features; and

repeating at least the executing step and the adjusting step to form the one or more trained machine learning algorithms;

obtaining a Raman spectrum of a polymer sample from a Raman spectrometer, wherein the Raman spectrum comprises a plurality of peaks and valleys in chemical and structural fingerprints that correspond to one or more polymer properties or features;

computing the one or more polymer properties or features by comparing the obtained chemical and structural fingerprints to stored polymer property and feature data, wherein the computing step is carried out by performing one or more trained machine learning algorithms; and

determining based on one or more computed polymer properties or features, whether the quality of the polymer sample meets a predetermined quality threshold.

14. The medium as set forth in claim 13 , further comprising, before the computing step, classifying, by the polymer property computing device, the peaks and/or valleys in the obtained chemical and structural fingerprints as either a relevant polymer property or feature or a non-relevant polymer property or feature.

15. The medium as set forth in claim 13 , further comprising, after the determining step, reporting, by the polymer property computing device, the one or more computed polymer properties or features.

16. The medium as set forth in claim 13 , wherein the polymer sample comprises a polyolefin.

17. The medium as set forth in claim 13 , wherein the one or more polymer properties is one or more properties selected from the group consisting of molecular weight, melt flow rate, lamellar thickness, crystallinity, xylene solubles, one or more tensile properties, and one or more compressive properties.

18. The medium of claim 13 , wherein the one or more polymer features is one or more features selected from the group consisting of an amount of one or more additives and an amount of one or more comonomers.

19. A polymer property computing device comprising:

a processor;

a memory, wherein the memory is coupled to the processor configured to execute programmed instructions stored in the memory comprising:

obtain a Raman spectrum of a polymer sample from a Raman spectrometer, wherein the Raman spectrum comprises a plurality of peaks and valleys in chemical and structural fingerprints that correspond to one or more polymer properties or features;

compute the one or more polymer properties or features by comparing the obtained chemical and structural fingerprints to stored polymer property and feature data, wherein the computing step is carried out by performing one or more trained machine learning algorithms; and

determine based on one or more computed polymer properties or features, whether the quality of the polymer sample meets a predetermined quality threshold;

wherein the processor is further configured to execute programmed instructions stored in the memory further comprising, training one or more machine learning algorithms before the obtaining step, wherein the training step comprises:

receiving both a Raman spectrum and measured polymer properties and features for each of a plurality of known polymer samples;

correlating one or more peaks and/or valleys in the structural and chemical fingerprints of the received Raman spectrum with the measured polymer properties and features to form at least part of the polymer property and feature data;

executing one or more machine learning algorithms to learn to recognize the correlated one or more peaks and/or valleys in the structural and chemical fingerprints and to compute the polymer properties and features;

adjusting one or more hyper-parameters to minimize any errors between the computed polymer properties and features and the measured polymer properties and features; and

repeating at least the executing step and the adjusting step to form the one or more trained machine learning algorithms.

20. The device as set forth in claim 19 , wherein the processor is further configured to execute programmed instructions stored in the memory further comprising, before the computing step, classifying, by the polymer property computing device, the peaks and/or valleys in the obtained chemical and structural fingerprints as either a relevant polymer property or feature or a non-relevant polymer property or feature.

21. The device as set forth in claim 19 , wherein the processor is further configured to execute programmed instructions stored in the memory further comprising, after the determining step, reporting, by the polymer property computing device, the one or more computed polymer properties or features.

22. The device as set forth in claim 19 , wherein the polymer sample comprises a polyolefin.

23. The device as set forth in claim 19 , wherein the one or more polymer properties is one or more properties selected from the group consisting of molecular weight, melt flow rate, lamellar thickness, crystallinity, xylene solubles, one or more tensile properties, and one or more compressive properties.

24. The device of claim 19 , wherein the one or more polymer features is one or more features selected from the group consisting of an amount of one or more additives and an amount of one or more comonomers.