On-line quality control in polymer processing using hyperspectral imaging

The use of plastic composite materials has been increasing in recent years in order to reduce the amount of material used and/or use more economic materials, all of which without compromising the properties. The impressive adaptability of these composite materials comes from the fact that the manufacturer can choose the raw materials, the proportion in which they are blended as well as the processing conditions. However, these materials tend to suffer from heterogeneous compositions and structures, which lead to mechanical weaknesses. Product quality is generally measured in the laboratory, using destructive tests often requiring extensive sample preparation. On-line quality control would allow near-immediate feedback on the operating conditions and may be transferrable to an industrial production context. The proposed research consists of developing an on-line quality control tool adaptable to plastic materials of all types. A number of infrared and ultrasound probes presently exist for on-line composition estimation, but only provide single-point values at each acquisition. These methods are therefore less adapted for identifying the spatial distribution of a sample’s surface characteristics (e.g. homogeneity, orientation, dispersion). In order to achieve this objective, a hyperspectral imaging system is proposed. Using this tool, it is possible to scan the surface of a sample and obtain a hyperspectral image, that is to say an image in which each pixel captures the light intensity at hundreds of wavelengths. Chemometrics methods can then be applied to this image in order to extract the relevant spatial and spectral features. Finally, multivariate regression methods are used to build a model between these features and the properties of the sample. This mathematical model forms the backbone of an on-line quality assessment tool used to predict and optimize the operating conditions under which the samples are processed.