Effect of Extrusion Parameters on Properties of Powder Coatings Determined by Infrared Spectroscopy

In polymer extrusion, compounding is a continuous mixing process that is also used to produce highly reactive powder coatings. A premixed batch of powder coating is added to the feeding section and extruded, preferably by a co-rotating twin-screw extruder. One essential parameter in the processing of highly reactive materials is the melt temperature: If it is too high, pre-reactions occur during the extrusion process, which may cause high rejection rates. We studied the melt temperature of an epoxy/carboxylbased powder coating using a retractable thermocouple at 3 different axial positions along the barrel of a ZSK34 co-rotating twin-screw extruder. The influence of different processing conditions on the reactivity of a highly reactive powder coating was examined by infrared spectroscopy and differential scanning calorimetry. Furthermore, the specific energy input and the color change in the finished powder coating at different processing points were investigated. Multivariate data analysis was used to correlate mid-infrared spectra, melt temperatures, specific energy inputs, enthalpies of reaction and changes in color. Received on 18-05-2017 Accepted on 05-09-2017 Published on 05-10-2017


INTRODUCTION *
Optical measurement techniques are widely used to analyze chemical reactions and polymer compositions.Near-infrared (NIR) and mid-infrared (MIR) show different details of the chemical structure in different spectral ranges.NIR spectra (12500 to 4000 cm -1 , corresponding to 0.8 to 2.5 µm) are often used to investigate polymer blends and curing kinetics or to predict mechanical properties [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17].In the NIR range, A. Kelly et al. [1] measured the drug and plasticizer content in a polymer melt.A sensor was placed in the die, and measurements were performed in transmittance.Other reactions such as the graft copolymerization of maleic anhydride on polypropylene were studied by L. Moghaddam [2]: A NIR spectrometer was connected via a fiber to a laboratory-scale extruder.To determine the effect of chain scission, NIR spectra and apparent viscosity were measured simultaneously.A. R. McLauchlin et al. studied the contamination of recycled polyethylene terephthalate by polylactid acid [3].D. F. Barbin investigated the potential of VIS/near infrared spectroscopy as a tool for analyzing biodegradable films [4].The spectra were analyzed by partial least squares (PLS) regression and the results used to predict mechanical properties such as elongation, tensile strength and Young's modulus.T. Rohe et al. [5] investigated different compositions of polypropylene-polyethylene blends, measuring them inline in transmittance.A further field of application is quality control in hot melt extrusion [6,7].T.  [15,16].
In polymer extrusion, the same properties have also been determined using light in the MIR spectral range (4000 to 400 cm -1 ) [21][22][23][24][25][26]. P.D. Coates et al. employed an online MIR setup to detect polymer blends based on polypropylene and polyethylene [21].The blends were measured inline using near infrared and Raman and online using MIR and NIR in a single-screw extruder.Y. Kann demonstrated the possibility of analyzing the crystallinity of poly(3-hydroxybutyrate-co-4hydroxybutyrate) polymers and compared the results of the method to DSC measurements [19].Z. Cui studied the application of attenuated total reflectance (ATR) Fourier transform infrared spectroscopy (FTIR) in a reductionpolymerization route for the synthesis of conducting poly (3,4ethylenedioxythiophene) polymers [20].
A commonly applied multivariate data analysis method used to predict material or process parameters on the basis of chemical information contained in infrared spectra is partial least squares (PLS) regression [24][25][26][27][28][29].PLS is basically a combination of principal component analysis (PCA) and multiple linear regression (MLR).We used the PLS regression method incorporated into the OriginPro 2015G software (OriginLab Corporation, Northampton, USA).PLS is often used as a prediction tool and for reducing highly correlated variables, which is usually the case for NIR and MIR spectra, to a set of independent variables.The goal is to distinguish between dependent (endogenous) variables and independent (exogenous) variables and build a linear model.Latent variables are extracted to estimate the covariance of dependent and independent variables.The regression of the values is predicted from the dependent variables by separating them from the independent variables.To define a quantitative relation between latent variables, each latent variable must be defined by one or more indicators [31].
In this case study, MIR spectroscopy in combination with PLS regression was used to predict the influence of screw speed, mass throughput, specific energy input (SEI) and melt temperature on the reactivity and color of epoxy-based powder coatings.The color was evaluated using !E (i.e., the change of color in the CIE Lab color space) relative to sample 1, calculated according to: where L, a and b were 40.88, 47.36 and 25.81, respectively.CIE Lab describes the lightness of a color, where L is between black and white, a is between red and green and b is between blue and yellow.
The SEI describes the amount of energy needed for material processing and is defined as: where M is the drive torque of the screws, N is the screw speed, and !m is the mass throughput.
Models for the linear regressions were evaluated using the root mean square error (RMSE) and the coefficient of determination R 2 in equations ( 3) to (6).RMSE and R 2 are defined as: S y = y i !y ( ) where E is the expectation of an estimator ! with respect to an unknown parameter !, S û 2 is the residual sum of squares, and sy is the total sum of squares [32].

Infrared Spectroscopy
The epoxy/carboxyl-based thermosets (Table 2) were measured by infrared spectroscopy.Figure 1 shows the samples for the measurements.
The samples were studied in the MIR range by means of ATR measurements.Figure 2 shows the principle of an ATR measurement setup.The emitted light passes the crystal and arrives at the crystal-sample interface, where it is totally reflected.If light passes from a material with a high refractive index to one with a low refractive index, total reflection will occur at a critical angle of incidence.However, at the interface between crystal and sample, light penetrates the sample and creates an evanescent field.The penetration depth is usually in the range of a few microns [33].The incident angle ! and the penetration depth dp can be defined as: and where ns is the refractive index of the sample, nc is the refractive index of the crystal, and !0 is the wave length.

EXPERIMENTAL
A highly reactive epoxy resin was extruded using a ZSK34 co-rotating twin-screw extruder from Coperion.The melt temperatures were measured at two positions in the barrel (Figure 3) and at one position in the die (Figure 4).The processing points are listed in Table 1, and the schematic screw configurations are shown in Figure 5.
As mentioned above, all samples were produced at different screw speeds and mass throughputs.The reaction enthalpies of the extruded flakes and the final powder coating were measured by differential scanning calorimetry using a DSC822e from Mettler Toledo.The heating rate was 5 K/min from 90°C to 200°C.The color measurements were performed using a Datacolor 400 spectrophotometer.The data listed in Table 2 and the IR spectra of three samples at each processing point were used for PLS regression analysis.As the maximum melt temperatures of screw configurations T1 and T2 were similar, PLS regression was applied to samples 1 to 9.

RESULTS AND DISCUSSION
The influence of mass throughput on the melt temperature is illustrated in Figure 6.Due to the shorter residence time at higher mass throughput, the maximum melt temperature decreased from approximately 145°C to 135°C.The centers         An increase in screw speed of 300 rpm resulted in a temperature increase of between 6°C and 9°C (Figure 9) in the die, whereas increasing the throughput by 60 kg/h decreased the melt temperature by about 4°C in the die (Figure 6).The mixing elements used slightly increased the maximum melt temperature and led to a more homogeneous melt temperature (Figure 10).The difference between minimum and maximum melt temperature !T decreased.The reaction enthalpy was measured by DSC before and after milling.The results are illustrated in Figure 11 (flakes) and Figure 12 (powder).High melt temperatures resulted in a low reaction enthalpy, and inhomogeneous temperature profiles led to increased variability in the samples.Further, milling decreased the enthalpy of reaction.Figure 13 shows the difference in the infrared spectra between the premix and the extruded samples (flakes and pressed powder tablets).Due to the high energy input of the mill, the chemical structure of the powder coating changed.On the one hand, homogenization improved, but on the other a decrease in the enthalpy of reaction was observed, indicating pre-reaction in the mill.Figure 14 shows the difference in the infrared spectra between the premix and the extruded samples (flakes and pressed powder tablets).Due to the high energy input of the mill, the chemical structure of the powder coating changed.On the one hand, homogenization improved, but on the other a decrease in the enthalpy of reaction was observed, indicating pre-reaction in the mill.
Since different processing points led to different melt temperatures and residence times in the extruder, we expected different reaction enthalpies and thus different chemical states of the finished powder coatings.We investigated, whether this information can be extracted from MIR absorption spectra.Relevant sections of the corresponding absorptions spectra are plotted in Figure 14.
The spectra shows a significant difference in absorbance between premix and extruded samples.To determine the influence of different processing conditions, the PLS method was used to examine the spectra.
The multivariate linear PLS regression models (calculated from the ATR absorbance spectra of the samples) of throughput, screw speed, maximum melt temperatures at positions 1 to 3, SEI and !E value are illustrated in Figure 15.A Savitzky-Goly filter and the 1 st derivative were used for preprocessing the recorded spectral data.Cross-validation of the PLS models was performed using leave-one-out cross validation.The resulting PLS regression models show a coefficient of determination above 0.97 and high predicted value accuracy.

CONCLUSION
We have shown that infrared spectroscopy is well suited to quantify the influence of different processing conditions of extruded epoxy-based powder coatings.Combining ATR with multivariate data analysis to determine color change, specific energy input and the reactivity of the powder coating yielded results that showed good correlation with measurements.Further, it is possible to verify the influence of the processing conditions on the melt temperature at different axial positions  along the barrel.The importance of color as a parameter in powder coatings makes the ability to examine !E values before the coating process of great potential interest.

Figure 2 :
Figure 2: Principle of ATR measurements; ns and nc are the refractive indices of the sample and the crystal, respectively.

2 .
Mass throughputs, screw speeds, reaction enthalpies, maximum melt temperatures and !E values are shown in Table All melt temperatures from positions 1 to 3 are the maximum values measured during the production time of about 10 min per processing point.The !E values are given relative to sample 1.

Figure 3 :
Figure 3: Measurement position of the retractable temperature sensor in the barrel.

Figure 4 :
Figure 4: Measurement position of the retractable melt thermocouple in the die.

Figure 5 :
Figure 5: Screw configurations and measurement positions of the temperature sensor; length of spacer element: 14 mm; total screw length: 24 L over D.

Figure 6 :
Figure 6: Temperature profile in the die -relationship between temperature and mass throughput at 900 rpm.Center of the die indicated by a dash-dotted line.

Figure 7 :Figures 7
Figure 7: Relationship between temperature and screw speedmeasuring position 1 at 900 rpm.The center of the barrel is indicated by a dash-dotted line.

Figure 8 :
Figure 8: Relationship between temperature and screw speedmeasuring position 2 at 900 rpm.The center of the barrel is indicated by a dash-dotted line.

Figure 9 :
Figure 9: Relationship between temperature and screw speedmeasuring position 3.The center of the barrel is indicated by a dashdotted line.

Figure 10 :
Figure 10: Relationship between temperature and screw configuration.

Figure 13 :
Figure 13: ATR measurements of premix and the extruded samples at processing point 1.

Figure 14 :
Figure 14: ATR measurements of pressed powder tablets at different wavenumbers.