3.1. CT Measurements and Reconstruction and FE Analysis for Water Sorption with Experimentally Determined Diffusion Coefficients
Figure 2 shows the reconstructed cross-section of three samples of the dimensions (4 × 2 × 80) mm
3, selected from the center of the sample, aged for different periods in the water bath at room temperature as a greyscale image. The variation in the storage periods in water results in different integral water contents for the samples determined by mass increase, which are given as integer values.
For each sample, a characteristic grey value distribution over the sample cross-section can be recognised, which is evaluated subsequently along the dotted lines inserted in the samples shown in
Figure 3, and which correspond to the reconstruction method of the water concentration distribution described in Equation (3). The dotted lines in
Figure 3 represent the z-direction in the sample coordinate system.
Figure 4 shows the local water saturation along the samples’ z-direction for the integral water contents of 3% (red line), 5% (blue line) and 9% (green line) for the water bath temperature of 80 °C. The water bath temperature of 80 °C was initially chosen to be considerably above the glass transition range
of approx. 60 °C of dry PA6.
The location-dependent saturation profiles shown are measured on three samples each and are then arithmetically averaged so that a standard deviation can be given for each measured value in the form of an error tube.
Figure 4 shows the relative concentration curve reconstructed from the CT measurements over the cross-section of the respective measured samples at a water bath temperature of 80 °C. The samples with an integral water content of 3% are shown with a red-dotted line and those with 5% in blue. The samples shown with a green-dotted line are samples with an integral water content of 9%. It can be seen that the water concentration distribution changes significantly for the different integral water contents.
Furthermore, it can be seen that for the integral water content of 3% for both water bath temperatures, negative water saturations are also determined in the reconstruction. These negative values occur when a higher average X-ray absorption
is measured for the completely dry reference sample than the simultaneously measured gradient sample has in its dry core as a locally determined absorption value
. For this purpose, negative reconstruction values are shifted to the water saturation of 0% on the ordinate in the following, so that a comparison of the CT measurements with the FE calculations can be accomplished. The applied shift parameters are listed in
Table 3 and used according to Equation (14). Here the value
results from the multiplication of the shift factor
with the original measured value
and the subsequent addition of the factor
.
In addition, it can be seen that in the boundary layer of approx. ±1.0 to ±0.8 mm a clearly too-low water content is assumed, although a complete saturation of the boundary layer can be assumed during sorption in a water bath. This insufficient water content can be attributed to two different effects: First, the density gap between air and material is smeared over two voxels so that, due to the geometric measurement arrangement, an area of approximately 110 µm (with a voxel size of 55 µm) is reconstructed with insufficient water content. For this reason, an examination of the water-induced swelling on the basis of the measured CT data is not possible due to the large voxel size and will not be considered further. The required scan time of 1.5 h is sufficiently large for re-drying effects in the water-saturated sample boundary to show a relevant influence, since the surrounding air in the laboratory standard climate (23 °C/50% r.H.) has a lower moisture content than the sample boundary layer. However, re-drying is a continuously changing effect on the moisture content of the sample boundary layer during the scan time, which cannot be sensibly determined with this measurement setup. For the above reasons, the measured maximum values are assumed to show a complete water saturation of 100% at the sample boundary and the shift parameters used for the abscissa are given in
Table 4 and used according to Equation (15). The factors used in Formula (15) are to be seen correspondingly to the factors used in Formula (14) in the z-direction. This enables a direct comparison of the measurement results with the FE calculation model for all samples measured by CT.
For the FE calculation of the concentration distribution within the sample cross-section, the sorption calculation described in
Section 2.4 based on the mass diffusion analysis in Abaqus is used. The concentration-dependent mutual diffusion coefficients are used according to
Table 2 for the values given at a water bath temperature of 80 °C. The analysed sample geometry corresponds to the bar-shaped sample of dimensions (4 × 2 × 80) mm
3 used in the CT scans. Within a Python script, the concentration is normalised to the solubility
(see Equation (11)) from the FE calculation is evaluated as a function of the selected spatial coordinate
and additionally normalised to the respective maximum value of the saturation
, which depends on the water bath temperature, so that it corresponds to the water saturation
in percent.
The reconstructed results of the CT measurement method, shifted according to Equations (14) and (15) and
Table 3 and
Table 4, are compared with the calculated concentration distributions in
Figure 4 for a water bath temperature of 80 °C. Since both the concentration distributions and the integral water contents are determined from averages of the values of several samples, the calculated concentration distributions are shown for the time step directly below and directly above the integral water contents of the experimentally determined data. The error tubes shown in
Figure 4 are omitted in the comparison with the FE analysis in
Figure 5 for readability.
Figure 5 shows the water saturation distribution of the reconstructed CT measurements combined with the concentration distribution calculated by FEM over the sample thickness for a water bath temperature of 80 °C. The measured integral water contents of 3%, 5% and 9% are compared with the calculated curves of the time steps directly below or above, which result in integral water contents directly below or above the measured values considered. For the values shown in red, which belong to 3% integral water content, it can be seen that the calculated concentration gradient tends to slope too steeply in comparison to the measured gradient, so that the calculation initially slightly overestimates the concentration gradient for low water contents at high moisture gradients. For the values shown in blue for an integral water content of 5%, the calculated concentration courses are in good approximation with the reconstructed CT measurement data and in particular the gradient change in the water distribution curve is reproduced qualitatively very well by the FE model. For the reconstructed values of the samples with 9% integral water content, a complete water saturation and thus constant concentration distribution can be recognised, which is reproduced by the FE calculation with the higher integral water content of 8.85%. The calculation with a lower integral water content of 8.37% predicts that, according to the calculation, there is no complete water saturation yet at this time step. However, both calculated concentration distributions have a lower water content than the integral 9% water content assumed in the measurement. The lower integral water contents result from the experimentally determined, temperature-dependent maximum saturation of the material, which corresponds to a saturation of 8.9% for a water bath temperature of 80 °C. The stated 9% water content of the CT reconstruction results from the mean value of the measured samples rounded to integer values. Overall, however, it can be summarised that the FE calculation of the concentration distribution corresponds in a good approximation to the reconstructed concentration distribution from CT measurements, and thus validates it.
3.2. CT Measurements and Reconstruction and FE Analysis for Water Sorption with Diffusion Coefficients Determined Using a Factor Comparison
In addition to assessing the water distribution at different times over the cross-section of samples saturated at a water bath temperature of 80 °C, the variation in the water bath temperature is also of interest. Here, water saturation at a room temperature of 23 °C is considered in order to be able to evaluate diffusion processes and concentration distributions at room temperature. Similar to the previous series of measurements at a water saturation of 80 °C, CT measurements are conducted on samples with different integral water contents after storage in a water bath at 23 °C. The results of the CT measurements are presented as a reconstructed relative concentration distribution over the cross-section of the respective measured samples in
Figure 6. Samples with an integral water content of 3% are shown with a red-dotted line, and those with 5% in blue. For a water bath temperature of 23 °C, in contrast to the previously shown results at 80 °C, measurements are also carried out on samples with an integral water content of 7% instead of 9%; these are shown with a yellow-dotted line. Here, it was decided to select 7% integral water content, since a pronounced moisture gradient is to be expected due to the lower water content and thus not an (almost) constant water distribution as in the previously analysed 9% integral water content. It can also be seen for this measurement series that the concentration distribution for the different integral water contents differ significantly from each other and the higher the integral water content of the samples, the lower the concentration gradients. For the two integral water contents of 3% and 5%, respectively, constant values can be seen in the area of the sample centre; this cannot be seen for the integral water content of 7%.
Analogously to
Figure 4, it can be seen that the water concentration distribution changes significantly for the different integral water contents and a delay of the water saturation to negative values occurs both in the boundary area of the samples and in the middle area of the concentration curves. Based on the previously mentioned reasons in
Section 3.1., these can be shifted to a maximum saturation of 100% water saturation in the boundary region and to 0% saturation in the sample core of the 3% and 5% samples, respectively, by using shift factors. The shift factors are listed in
Table 5. They are applied similarly to the 80 °C water bath temperature results using Equations (14) and (15) and the reconstruction results thus obtained are shown in direct comparison with the calculated concentration curves in
Figure 7.
For the FE calculation of the concentration curves over the sample cross-section, it should be noted that the concentration-dependent diffusion coefficients previously used for the simulation originate from experimentally measured sorption curves of a temperature of 80 °C. The experimental procedure for determining concentration-dependent diffusion coefficients as well as the maximum saturation is time-consuming due to the required sorption times and becomes even more tedious for lower water bath temperatures and lower surrounding water concentrations due to the reduction in the diffusion rate at lower temperatures. A conversion of the diffusion coefficients with the help of the Arrhenius approach is also not possible, as the range of validity of the formula is only given below the glass transition region
[
34]. The approach presented by Vrentas et al. [
35] for calculating the diffusivity of water based on the free-volume theory with a range of validity above
cannot be used for conversion in this case either. PA6, however, has a
of approx. 60 °C in the dry state but shifts it to −20 °C for completely water-saturated material [
10,
11]. Although the chosen 23 °C of the water bath temperature is below the glass transition for dry material, a
shift is induced locally by the water absorption [
11]. Since both models mentioned are only valid above or below, but not within the glass transition, an alternative approach for determining concentration-dependent diffusion coefficients is developed.
For the approach used, only the diffusion coefficient for saturation of the samples in a water bath at 23 °C is determined experimentally, since in this case sample saturation is still reached within a moderate period of approx.
for the present sample geometry. This is determined from the measured sorption curve using Equation (7) and is
. However, in the FE calculations performed, the diffusion coefficients
converted using Equation (8) are used according to the approach of Inoue et al. [
21], which is based on the principle of mutual diffusion, and are converted from the diffusion coefficients previously determined experimentally at different ambient concentrations. So that FE analyses can still be calculated for saturation at 23 °C with concentration-dependent diffusion coefficients, the correlations between the different concentration-dependent mutual diffusion coefficients must be determined. For this purpose, factors
are determined, which are calculated from the ratio of the respective concentration-dependent, mutual diffusion coefficient
to the experimentally determined diffusion coefficient for the water bath saturation at 80 °C according to Equation (16).
The factors
are subsequently multiplied according to Equation (17) with the experimentally determined diffusion coefficient for the saturation in the water bath at 23 °C
so that the respective concentration-dependent mutual diffusion coefficients
result and the values are listed in
Table 6.
By using the concentration-dependent values for the mutual diffusion coefficients
listed in
Table 6, the concentration curves for the different integral water contents can be calculated by FE. These are shown in
Figure 7 in comparison with the reconstructed CT measurement data shifted by means of
Table 5. For reasons of readability, the error tubes of the concentration curves averaged from three measurements each are omitted.
The results of the calculated and measured concentration distributions shown in
Figure 7 indicate that for sorption in a water bath at 23 °C the FE calculation also agrees with the CT measurements in a good approximation, but in principle slightly overestimates the concentration gradient. For each measured integral water content, a calculated integral water content below and above the measurement is shown corresponding to
Figure 7.
The different concentration gradients from the maximum saturation at the boundaries to the concentration minimum at the centre of the samples are simulated by the FE model for all measurements considered. With an integral water content of 3%, only the constant, dry region within the sample is overestimated, so that a discrepancy occurs here. The results presented in
Figure 7 show that the conversion by factors is a reasonable method to determine concentration-dependent, mutual diffusion coefficients from only one experimentally determined diffusion coefficient at high surrounding concentration. However, a requirement for this is that concentration-dependent, mutual diffusion coefficients have previously been determined for a different surrounding temperature.
The approach described in
Section 2.2. shows that low-energy CT measurements can be used to represent concentration distributions due to sorption processes in PA6. In addition, it can be stated that the FE calculation method introduced in
Section 2.3. based on the mass diffusion analysis provided in Abaqus can resolve the concentration distribution over the sample cross-section in a good approximation. Concentration-dependent, mutual diffusion coefficients are used for the FE calculation. In conclusion, concentration distributions of water in PA6 can be calculated with the developed FE method with sufficient accuracy.