Determination of polymer blends composed of polycarbonate and rubber entities using near-infrared (NIR) spectroscopy and multivariate calibration

doi:10.1016/j.polymertesting.2013.03.008

Polymer Testing

Volume 32, Issue 4, June 2013, Pages 802-809

In Memory of James DeRudder

https://doi.org/10.1016/j.polymertesting.2013.03.008 Get rights and content

Abstract

This paper presents an application of near-infrared NIR spectroscopy and multivariate calibration for compositional analysis of complex blends of polycarbonate (PC) and three copolymer components (C1, C2, and C3). Each of the copolymers is composed of 2–3 entities (sub-components) consisting of combinations of butadiene, styrene and acrylonitrile. The concentrations of the PC and three copolymer components were varied using a modified D-optimal design with criteria that minimized the inter-component correlations. To minimize non-chemical spectral variations, the acquired NIR spectra were pre-processed using standard normal variate (SNV and 2nd derivative Savitzky-Golay). Spectral range selection was explored in order to identify which optimal spectral regions were required to generate robust partial least-squares (PLS) models for each component. The optimal calibration models for PC, C1, C2 and C3 exhibited RMSEP values of 0.94%, 0.62%, 0.59% and 0.69% respectively. Using a set of external validation samples, the optimized calibration models for PC, C1, C2 and C3 exhibited bias values of −1.07%, 0.28%, −1.21% and −1.00%, and RMSEP of 2.43%, 1.44%, 1.51%, and 2.05%, respectively. Finally, using a set of 3 samples, the optimized model was successfully transferred to a secondary instrument located in a quality control (QC) laboratory.

Introduction

Over the years, there has been increased demand for polymers with superior mechanical, thermal and hydro-aging resistance properties that can be used in automotive, electronic and construction businesses. To achieve this objective, chemical companies either have to develop new polymers or resort to blending several polymers. Most chemical companies resort to the latter because, not only does it combine the individual polymer properties, this approach is less risky from a business standpoint and is easy to scale up and commercialize [1], [2]. Since several components are blended into one product, there is a need to deploy a reliable, robust, precise and accurate measurement system that will determine the relative amounts of all the components within the blend.

Several strategies have been employed in the past to determine the composition of polymer blends with rubber entities [3], [4]. Miller et al. [5] used NIR and classical least squares (CLS) to analyze butadiene and styrene butadiene rubbers. The sample matrix used this study was simple and unique NIR spectral features were identified and used for calibration. In addition, samples were dissolved in carbon tetrachloride (CCl₄) and analyzed by NIR transmission, hence making this approach difficult to implement when dealing with complex blend matrixes that are composed of copolymer entities that are insoluble in any solvent. Guilment and Bokobza [6] used NIR reflectance and chemometrics to determine the microstructures of polybutadiene and styrene-butadiene copolymers composition. This study showed very good reproducible results, however the sample matrix was simple and the calibration set was small (21 samples) and did not utilize design of experiments (DoE). This makes it challenging to apply this methodology to multicomponent samples. Shield and Ghebremeskal [7] examined the use of mid IR and NIR for characterizing blends of styrene-butadiene and acrylonitrile-butadiene. The matrix in this study was a relatively simple system. Vilmin et al. [8] considered using NIR and multivariate analysis for the compositional analysis of butadiene, styrene-butadiene and Isoprene rubber. Again, the sample matrix in this study was simple and the NIR measurements were performed in transmission mode on pressed film. It is well known that pressing films to constant thickness can exhibit variations that in turn affect transmission measurements.

The sample investigated in this research is a complex blend matrix composed of PC and three copolymer components (C1, C2, and C3). Each of the copolymers is composed of 2–3 entities (sub-components) consisting of combinations of butadiene, styrene and acrylonitrile. The powder blend for each component is mixed via a tumble blender and extruded, resulting in pellets as the final product. The relative amounts of the PC and each of the three copolymer components is very critical to key performance attributes of this product such as flow, impact, thermal and hydro-aging resistance. The final pellets are highly insoluble in any solvent, thus making it impossible to use a solution based method.

The current approach used for compositional analysis is mid infrared (IR) on pressed film of one pellet followed by quantitative infrared transmission measurements based on multiple linear regression (MLR). Results of this analysis were then used to determine the composition of several million pounds of material. There are three main issues with this approach. First, the procedures associated with pressing films to constant thickness are known to exhibit variations that in turn affect transmission measurements. Second, there is a huge risk involved in deciding the quality of a high volume product based on the very small sample size (1–2 pellets). Lastly, the turnaround time of pressing a film followed by transmission measurements is approximately 10 min. This duration emanates from warming the heating plates, placing the pellet, removing the film, placing it in an IR transmission card, purging the transmission compartment with nitrogen for 2 min and finally acquiring transmission measurement on the film.

To mitigate the small sample size coupled with variations in sample preparation and turnaround time, near-infrared (NIR) spectroscopy was examined. Near-infrared spectroscopy (NIR) is an analytical measurement tool used in chemical, pharmaceutical, petroleum and agricultural industries [9], [10], [11], [12]. The technique requires little or no sample preparation and is nondestructive, reagentless, simple and fast. In addition, NIR spectroscopy exhibits the capability to extract quantitative information of several species within a sample from a measured spectrum, thereby making this approach ideal for multicomponent determination of complex matrixes. The principal drawback to this method is the occurrence of broad and highly overlapping spectral features. In a complex sample, it is highly unlikely that selective quantitative information can be made on the basis of a single wavelength. Consequently, quantitative information must be based on information emanating from multiple wavelengths, thereby requiring the use of multivariate calibration techniques such as partial least-squares (PLS) regression [6], [7], [8], [13].

In the research presented here, the application of NIR and multivariate calibration for multicomponent analysis of a complex polymer mixture composed of PC and three components (C1, C2, and C3) will be examined. The approach will evaluate the development of calibration models, validating the models and finally translating the models to another instrument in a quality control (QC) laboratory.

Section snippets

Materials

Ten independent validation samples and 40 calibration samples were prepared at the Mt. Vernon, IN site. The concentration ranges for PC, C1, C2, and C3 were 64 to 78, 8 to 22, 0 to 14, and 0 to 14%, respectively. A D-optimal mixture design was created in Design Expert 7 (Stat-Ease, Inc., Minneapolis, MN) in which the polymeric ingredients were forced as a mixture to 100%. The individual components were tumble blended and extruded, resulting in cylindrical pellets with an average length and

Spectral data characterization

Fig. 2, displays the FT-NIR pure component spectra for all 4 major components and a representative manufacturing sample. Visual inspection of the pure spectra portrays extensive spectral overlap which in turn represents a unique challenge in elucidating and quantifying the perturbations imposed by changes in concentration of components. The complexity of this overlap demands the use of multivariate calibration techniques such as partial least-squares (PLS) regression to correlate the acquired

Conclusions

The results presented in this study clearly demonstrate the capability of NIR and chemometrics in determining the composition of complex polymer blends consisting of PC and three other components. Robust PLS calibration models were developed for all 4 major components. The performance of the calibration models were successfully validated using a set of independent validation and manufacturing samples. Finally, the optimized models were successfully transferred to a secondary instrument in a QC

Acknowledgements

The authors would like to acknowledge Roger Hurst and Jeff Morlock for their assistance.

References (16)

S. Bertin et al.
Study and characterization of virgin and recycled, LDPE/PP blends
Euro. Polym. J.
(2002)
F. Findik et al.
Investigation of mechanical and physical properties, of several industrial rubbers
Mater. Des.
(2004)
J. Guilment et al.
Determination of polybutadiene microstructures and styrene-butadiene copolymers composition by vibrational techniques combined with chemometric treatment
Vib. Spectrosc.
(2001)
D. Lafrance et al.
Vib. Spectrosc.
(2004)
F. Goddu
Determination of unsaturation by near-infrared spectrophotometry
Anal. Chem.
(1957)
A.J. Durbetaki et al.
Near infrared and nuclear magnetic resonance spectrometry in analysis of butadiene polymers
Anal. Chem.
(1965)
C.E. Miller et al.
Determination of microstructure and composition in butadiene and styrene-butadiene polymers by near infrared spectroscopy
Anal. Chem.
(1990)
S.R. Shield et al.
Use of mid- and near-infrared techniques as tools for characterizing blends of copolymers of styrene-butadiene and acrylonitrile-butadiene
J. Appl. Polym. Sci.
(2003)

There are more references available in the full text version of this article.

Cited by (18)

ATR-FTIR spectroscopy combined with machine learning for classification of PVA/PVP blends in low concentration
2022, Vibrational Spectroscopy
Citation Excerpt :
Near-infrared (NIR) and Partial Least Squares (PLS) were shown to be more effective than thermal analysis to determine the fraction of polypropylene below 1 wt% into polypropylene/high-density polyethylene, and polypropylene/ acrylonitrile‐butadiene‐styrene blends [9]. Similar results were obtained to evaluate polycarbonate blends with different amounts of butadiene, styrene, and acrylonitrile [10]. NIR, Fourier transform infrared spectroscopy (FTIR) and Raman spectroscopy was tested to analyze the concentration of polylactic acid, propylene polycarbonate, and poly (butylene adipate co-terephthalate) into biodegradable blends.
The use of Fourier transformed infrared spectroscopy (FTIR) and multivariate analysis for sample classification has been extensively explored in the literature. The overall accuracy obtained in most of the studies raises questions about their reliability due to technical limitations and misuse of the analysis driving to overfitting/underfitting of data. There are established procedures to avoid overfitting/underfitting, but there is a lack of studies to understand the relationship between the limitation of the data provided by the FTIR and the associated method of analysis for sample classification. In this study, FTIR spectra were obtained from thick films of poly (vinyl alcohol) (PVA)/poly (vinyl pyrrolidone) (PVP) polymeric blends, with PVP concentrations below 3% (wt./wt.). We analyzed the FTIR spectra by using simple algorithms for sample classification to evaluate the predicting model accuracy at low concentrations of PVP into the PVA matrix. The raw data were submitted to pre-treatment by standard normal variate (SNV), and different spectral ranges were explored by principal component analysis (PCA). Then, PCA and FTIR-SNV data were used for supervised analysis by using linear discriminant analysis (LDA); k-nearest neighbors (KNN); and support vector machines (SVM) algorithms. The method was able to classify samples with 0.1 wt% of PVP with an overall accuracy of 100% with quadratic SVM by the proper choice of the spectral range and number of PCs. Finally, we show that the FTIR associated with multivariate analysis can be used for sample classification at low concentration changes.
Quali-quantitative monitoring of chemocatalytic cellulose conversion into lactic acid by FT-NIR spectroscopy
2021, Spectrochimica Acta - Part A: Molecular and Biomolecular Spectroscopy
Citation Excerpt :
Nevertheless, after a long initial step to build the model, FT-NIR analysis can disclose, in few seconds, multiple information about the analyte, without sample pre-treatment or degradation. FT-NIR spectroscopy has gained popularity as a fast and inexpensive analytical technique in multiple different fields ranging between the determination of physicochemical characteristics of inorganic [15] or polymeric materials [16–18], the quality control of food [19–23], pharmaceutical [24,25] and agricultural products [26–28]. Moreover, FT-NIR instrumentations are oriented to online monitoring, by accessing the connection of the spectrophotometer with the reaction vessel through suitable fiber optic probes.
Chemocatalytic conversion of cellulose into lactic acid is a valuable alternative to simple sugar fermentation. Nevertheless, the procedures still need optimization to be translated to the industrial scale. Such translation would benefit by on-line monitoring of reaction parameters by fast, inexpensive, direct spectroscopic techniques. In this work, we propose the application of FT-NIR spectroscopy as a suitable analytical tool for monitoring the chemocatalytic conversion of cellulose into lactic acid. Comparison between different FT-NIR spectra at different reaction temperatures and times was exploited to qualitatively indicate the feasibility of the reaction. Besides, an FT-NIR prediction model was proposed for rapidly estimating the molar distribution of cellulose catalytic degradation components in the reaction mixtures. The calibration model was based on reference samples analysed by HPLC. The model was validated by an external validation set. Relevant statistical values of Ratio Performance to Deviations (RPD) referred to both calibration and external validation were obtained, thus demonstrating the potential of such analytical technique in process monitoring.
Development of a non-destructive method for determining protein nitrogen in a yellow fever vaccine by near infrared spectroscopy and multivariate calibration
2018, Spectrochimica Acta - Part A: Molecular and Biomolecular Spectroscopy
Citation Excerpt :
Since the 1980s, near infrared (NIR) spectroscopy has been used as a fast non-destructive method, in a quasi-universal application, using few reagents and with reduced analytical analysis times [16]. When associated to multivariate calibration involving molecules containing CH, NH, SH or OH bonds [17,18], it is widely applied for the determination of protein content [19–21] in agricultural activities and in environmental scenarios [22–24], as well as in the food [20,25–30], polymer [31–33], pharmaceutical [17,18,34–37] and fuel [38–41] industries. The near infrared region ranges from 700 and 2500 nm in the electromagnetic spectrum [17,18].
Near infrared spectroscopy (NIR) with diffuse reflectance associated to multivariate calibration has as main advantage the replacement of the physical separation of interferents by the mathematical separation of their signals, rapidly with no need for reagent consumption, chemical waste production or sample manipulation. Seeking to optimize quality control analyses, this spectroscopic analytical method was shown to be a viable alternative to the classical Kjeldahl method for the determination of protein nitrogen in yellow fever vaccine. The most suitable multivariate calibration was achieved by the partial least squares method (PLS) with multiplicative signal correction (MSC) treatment and data mean centering (MC), using a minimum number of latent variables (LV) equal to 1, with the lower value of the square root of the mean squared prediction error (0.00330) associated with the highest percentage value (91%) of samples. Accuracy ranged 95 to 105% recovery in the 4000–5184 cm⁻¹ region.
Characterization of PLA-talc films using NIR chemical imaging and Multivariate Image Analysis techniques
2018, Polymer Testing
Citation Excerpt :
Near infrared imaging spectroscopy (NIR) is a fast, powerful, and non-invasive technique allowing to scan the entire film surfaces (2-D inspection). It has been used in a number of investigations to characterize plastic films [18–21]. NIR hyperspectral imaging coupled with the Multivariate Image Analysis (MIA) technique was effectively used by Ghasemzadeh-Barvarz et al. to classify multilayer polymer films according to the absence or presence of functional layers [19].
A fast and non-invasive near-infrared (NIR) imaging system is proposed to characterize biodegradable polylactic acid (PLA) films containing talc, and produced by extrusion film-blowing for packaging applications. The ultimate goal is to use the system on-line for quality control and after a post-processing thermal treatment (annealing). A set of PLA-talc films with different talc contents (0–3% wt.) were produced and annealed under different conditions. NIR images of the films were collected after which the samples were characterized for their physical, mechanical, and gas barrier properties. Multivariate imaging techniques were then applied on the hyperspectral images. The results showed that the films with various talc loadings can be discriminated, and the information contained in the NIR spectra allowed predicting the annealed film crystallinity and O₂ permeability with good accuracy.
Use of laser-induced breakdown spectroscopy for the determination of polycarbonate (PC) and acrylonitrile-butadiene-styrene (ABS) concentrations in PC/ABS plastics from e-waste
2017, Waste Management
Citation Excerpt :
PLS multivariate calibration is strongly used in combination with near-infrared (NIR) spectroscopy. Sulub and DeRudder (2013), for example, determining the concentration of polycarbonate and rubber in blends of these materials using PLS models NIR. Thus, the present study shows and discuss a new application of laser-induced breakdown spectroscopy (LIBS) using PLS for fast and direct determination of the amount of PC and ABS in blends of these materials obtained from samples of e-waste.
Due to the continual increase in waste generated from electronic devices, the management of plastics, which represents between 10 and 30% by weight of waste electrical and electronic equipment (WEEE or e-waste), becomes indispensable in terms of environmental and economic impacts. Considering the importance of acrylonitrile-butadiene-styrene (ABS), polycarbonate (PC), and their blends in the electronics and other industries, this study presents a new application of laser-induced breakdown spectroscopy (LIBS) for the fast and direct determination of PC and ABS concentrations in blends of these plastics obtained from samples of e-waste. From the LIBS spectra acquired for the PC/ABS blend, multivariate calibration models were built using partial least squares (PLS) regression. In general, it was possible to infer that the relative errors between the theoretical or reference and predicted values for the spiked samples were lower than 10%.
Determination of diethyleneglycol content and number of carboxylic end groups in poly(ethylene terephthalate) fibers using imaging and conventional near infrared spectroscopy
2016, Polymer Testing
In this work, multivariate empirical models using conventional near infrared (NIR) spectroscopy and hyperspectral images were employed, respectively, to determine and map the diethyleneglycol (DEG) content and the number of carboxylic end groups (TCG) in poly(ethylene terephthalate) fiber samples. NIR spectroscopy is a fast, clean and efficient technique. NIR spectral data were obtained using two conventional spectrometers and a hyperspectral camera. The regression models were built using partial least squares. The models showed satisfactory results for the evaluated parameters, with Root Mean Squared Error of Prediction (RMSEP) values equivalent to repeatability of the reference methods. The distribution maps of the DEG and TCG content in the fibers samples were obtained by using the models built with the average spectra of the images to estimate the pixel content of each species, allowing us to identify areas with high content of these parameters.

View all citing articles on Scopus

View full text

Short communication: Analysis methodDetermination of polymer blends composed of polycarbonate and rubber entities using near-infrared (NIR) spectroscopy and multivariate calibration

Abstract

Introduction

Section snippets

Materials

Spectral data characterization

Conclusions

Acknowledgements

Euro. Polym. J.

Mater. Des.

Vib. Spectrosc.

Vib. Spectrosc.

Determination of unsaturation by near-infrared spectrophotometry

Anal. Chem.

Near infrared and nuclear magnetic resonance spectrometry in analysis of butadiene polymers

Anal. Chem.

Determination of microstructure and composition in butadiene and styrene-butadiene polymers by near infrared spectroscopy

Anal. Chem.

Use of mid- and near-infrared techniques as tools for characterizing blends of copolymers of styrene-butadiene and acrylonitrile-butadiene

J. Appl. Polym. Sci.

Short communication: Analysis method
Determination of polymer blends composed of polycarbonate and rubber entities using near-infrared (NIR) spectroscopy and multivariate calibration