TAGS: PVC, Plasticizers and Sustainability Science-based Formulation Polymer Reinforcement
New research use non-destructive spectroscopic analysis combined with Machine Learning (ML) to obtain rapid information on the identity and content of plasticizers in PVC objects of heritage value.
For the first time, a large and diverse collection of more than 100 PVC objects in different degradation stages and of diverse chemical compositions were analysed by chromatographic and spectroscopic techniques to create a dataset used to construct classification and regression models.
Improves Mechanical Properties
Plasticizers are used to improve and adjust the mechanical properties of the material by increasing flexibility, reducing viscosity, and decreasing friction during manufacture. Plasticizers can account for up to 50% of the total mass of
PVC objects.
Loss of plasticizers in
PVC is a significant process during an object’s lifetime. It can occur by evaporation into the surrounding air, extraction into liquids, or migration into another polymeric material. Non-destructive methods of analysis are frequently required in the field of heritage science. NIR spectroscopy was combined with PLS analysis to date fiber-based gelatine silver photographic papers.
Machine learning (ML) algorithms are used to make predictions or decision models based on training data, using a number of different approaches, from simple to sophisticated. As proof of concept, six common but distinctly different approaches were used and compared their efficacy in solving the problem of identification and quantification of plasticizers in
PVC.
Destructive and non-destructive methods of chemical analysis with machine learning approaches using spectroscopic data were used to obtain information previously available only in a destructive manner.
This information was used to create a publicly available dataset, to be expanded as the collection continues to grow. The currently available data was used together with the ATR FTIR and NIR spectra to create classification and regression models for the most common plasticizers.
Classification Model for Identification
In this study, a classification model was presented to identify DEHP, DOTP, DINP, DIDP, a mixture of DINP with DIDP, and unplasticized
PVC. The model can identify separate plasticizers apart from a mixture of two plasticizers.
Only the combination of DINP and DIDP was investigated because objects containing other combinations of plasticizers were too rare. Successful regression models were built for DEHP and DOTP, the most common plasticizers found in our collection of modern and historical
PVC objects.
Numerical differentiation proved to be particularly useful for NIR spectra, as it increased the classification accuracy. Both types of spectra can be used for spectral quantification using PLS regression, but NIR spectra result in less favourable RMSEP values.
Overall, the machine learning classification and regression models built with ATR FTIR spectra are more accurate and more robust than with NIR spectra. We hypothesize that the reason is that ATR FTIR spectra contain more and better-resolved information about the chemical composition and molecular structure of the object compared to NIR spectra. This study demonstrates that robust classification and regression models can be built based on collections of varied real-life objects.
Recent research on long-term
PVC degradation suggests that knowing the identity of a
plasticizer is important in studies of
plasticizer loss and associated conservation challenges. Knowing the
plasticizer also helps as a rough estimate of age: In the EU, objects with DBP and DEHP tended to be manufactured before 2008, while those with DOTP, DINP, DIDP, and DINCH mostly after 2008.
Accounting for this variety makes the model more robust and reliable for the analysis of objects in museum collections. Six different machine learning classification algorithms were compared to determine the algorithm with the highest classification accuracy of the most common plasticizers, based solely on the spectroscopic data.
A classification model capable of the identification of di(2-ethylhexyl) phthalate, di(2-ethylhexyl) terephthalate, diisononyl phthalate, diisodecyl phthalate, a mixture of diisononyl phthalate and diisodecyl phthalate, and unplasticized
PVC was constructed.
Additionally, regression models for quantification of di(2-ethylhexyl) phthalate and di(2-ethylhexyl) terephthalate in
PVC were built. This study of real-life objects demonstrates that classification and quantification of plasticizers in a general collection of degraded
PVC objects is possible, providing valuable data to collection managers.
Source: Nature