Unconsolidated composite material component verification via non-contact chemical sensing

A material chemistry verification system comprising:

• a computer system;
• a scan classifier configured to classify material chemistries in unconsolidated composite materials using Fourier transform infrared scans of the unconsolidated composite materials;
• an analyzer in the computer system, wherein the analyzer is configured to:
• receive a Fourier transform infrared scan for a section of an unconsolidated composite material in real time during manufacturing of the unconsolidated composite material by a composite material manufacturing system;
• and
• verify whether a material chemistry for the section of the unconsolidated composite material is correct in real time during manufacturing of the unconsolidated composite material by the composite material manufacturing system using the Fourier transform infrared scan and the scan classifier, wherein the scan classifier is a machine learning model system trained to classify the material chemistries in the unconsolidated composite materials using the Fourier transform infrared scans of the unconsolidated composite materials;
• and
• a machine learning model trainer configured to:
• identify sample Fourier transform infrared scans of sample materials;
• calculate a mean of the sample Fourier transform infrared scans;
• determine correlations of the sample Fourier transform infrared scans to the mean;
• remove a selected sample Fourier transform infrared scan from the sample Fourier transform infrared scans having a lowest correlation to the mean in response to a number of sample Fourier transform infrared scans having a correlation that is less than a threshold;
• repeat calculating the mean, determining the correlations, and removing the selected sample Fourier transform infrared scan in response to the number of sample Fourier transform infrared scans having the correlation that is less than a threshold;
• and
• train the machine learning model system using the remaining sample Fourier transform infrared scans in response to the remaining sample Fourier transform infrared scans having the correlation greater than the threshold, wherein the remaining sample Fourier transform infrared scans are standardized Fourier transform infrared scans.

A method for verifying a material chemistry, the method comprising:

• receiving a Fourier transform infrared scan for a section of an unconsolidated composite material in real time during manufacturing of the unconsolidated composite material by a composite material manufacturing system;
• verifying whether the material chemistry for the section of the unconsolidated composite material is correct in real time during manufacturing of the unconsolidated composite material by the composite material manufacturing system using the Fourier transform infrared scan and a scan classifier, wherein the scan classifier is a machine learning model system trained to classify material chemistries in unconsolidated composite materials using Fourier transform infrared scans of the unconsolidated composite materials;
• identifying sample Fourier transform infrared scans of sample materials;
• calculating a mean of the sample Fourier transform infrared scans;
• determining correlations of the sample Fourier transform infrared scans to the mean;
• removing a selected sample Fourier transform infrared scan from the sample Fourier transform infrared scans having a lowest correlation to the mean in response to a number of sample Fourier transform infrared scans having a correlation that is less than a threshold;
• repeating calculating the mean, determining the correlations, and removing the selected sample Fourier transform infrared scan in response to the number of sample Fourier transform infrared scans having the correlation that is less than a threshold;
• and
• training the machine learning model system using the remaining sample Fourier transform infrared scans in response to the remaining sample Fourier transform infrared scans having the correlation greater than the threshold, wherein the remaining sample Fourier transform infrared scans are standardized Fourier transform infrared scans.