Infrared and Thermal Fingerprinting Protocols for Polyolefin Blends
Infrared and thermal fingerprinting verifies polyolefin blend composition, oxidation levels, and phase contamination to enforce supplier compliance and landed part cost.
Partial Least Squares
Meaning
Statistical regression techniques that relate a matrix of dependent variables to a matrix of predictor variables represent a foundational tool in chemometric analysis. In the quality control of polymer blending and extrusion, partial least squares is used to build calibration models from complex infrared or Raman spectra. The algorithm identifies the orthogonal directions of maximum covariance between the spectral data and the physical property of interest, such as copolymer ratio or additive concentration.
The model is effective only within the concentration limits defined by the calibration standard set.
Mathematical Approach
Dimensionality reduction occurs as the algorithm transforms hundreds of spectral wavelengths into a few latent variables. This process avoids the overfitting issues common to multiple linear regression when variables are highly correlated. The method is particularly suited for datasets where the number of variables exceeds the number of samples.
Spectroscopic Application
In-line extrusion monitoring utilizes these models to measure resin melt properties in real time. Spectral fluctuations are instantly converted into concentration values for components like fillers or elastomers. This allows operators to adjust dosing systems without stopping the line.
Calibration Performance
Validation of the model requires testing against an independent set of samples with known properties. High predictive accuracy depends on the representative nature of the calibration dataset. The model must be updated if new raw materials are introduced.