Method and device for determining a permeability of a sample

Abstract

The present invention relates to a method for determining a permeability of a sample (1), in which the sample (1) is acted upon by a permeate (10) and a variation over time of a concentration and / or a quantity of the permeate (11) permeated by the sample (1) is measured and the permeability is determined from the concentration and / or the quantity. The measurement is terminated before a stationary state is reached, in which a temporally constant value of the permeability is present, or after a predetermined time interval has elapsed, and a final value of the permeability is calculated by a machine learning algorithm based on previously measured training data of a variation over time of the concentration and / or the quantity of the permeated permeate (11) and from the previously determined measurement data.

Description

[0001] Method and device for determining the permeability of a sample

[0002] The present invention relates to a method, a device and a computer program product for determining a permeability of a sample.

[0003] Many products are sensitive to atmospheric influences such as oxygen and water. This includes food products, but especially organic electronics such as organic light-emitting diodes (OLEDs) or organic photovoltaics. To protect such products,

[0004] Barrier films are therefore used in products to prevent the penetration of water and oxygen. In order to determine the quality of these barrier films, their barrier effect must be objectively measured. To measure the water vapor transmission rate (WVTR), a number of measurement methods exist in the state of the art, such as a

[0005] Calcium testing, a coulometric measurement, and absorption spectroscopy in the form of tunable diode laser absorption spectroscopy (TDLAS). All of these methods directly or indirectly measure the water vapor permeated through the film. For example, patent DE 10 2013 002 724 B3 discloses a method for determining the permeation rate of barrier materials. Especially for films with very low water vapor permeability, extremely long measurement times (days, weeks, or even months) are often required for just one measured value of the water vapor permeability.

[0006] The extremely high time expenditure is essentially due to the fact that the water vapor permeability can only be calculated from the measured data once a stationary equilibrium has been established (reaching the "steady state" condition). This equilibrium is largely determined by two kinetically controlled processes: the establishment of a stable water vapor concentration gradient in the barrier sample to be investigated and the establishment of an equilibrium between the water vapor molecules in the gas phase and those absorbed on the surfaces of the measuring cell.

[0007] The resulting extremely long measurement times prevent rapid and thorough quality control of barrier films during production, as well as during incoming goods inspection during use. Furthermore, the development processes for barrier films and processes that utilize these barrier films are prolonged and complicated.

[0008] To measure the WVTR, the water vapor permeated through the film is always measured directly or indirectly. The methods used are based on various physical principles. The following three methods are primarily used:

[0009] • Calcium test: optical or electrical measurement of the decomposition of a calcium layer by the permeated water vapor and calculation of the water vapor permeability based on the transmission or current

[0010] • Coulometric measurement: direct, coulometric measurement of the amount of water vapor permeated through the film and calculation of the water vapor permeability

[0011] • TDLAS spectroscopy: Measurement of the concentration of water vapor permeated through the film using tunable diode laser absorption spectroscopy and calculation of the water vapor permeability.

[0012] To obtain the measured water vapor permeability more quickly, for example, films are currently measured at elevated temperatures. However, this method is disadvantageous in that the temperature dependence of the permeation rate of the film under investigation must be known, the measurement conditions usually differ from the application conditions, and damage to the film is possible. Alternatively, a substitute molecule with a significantly higher permeation rate (e.g., helium) is used instead of water vapor. However, this involves additional costs. In addition, the functional relationship between the water vapor permeability and the permeability of the substitute molecule must be known, although this relationship is sample-specific and not generic.

[0013] The present invention is therefore based on the object of proposing a method and a device for determining a permeability which avoids the disadvantages mentioned, and which thus allows the permeability of a sample to be determined quickly, reliably and with high reproducibility.

[0014] This object is achieved according to the invention by a method according to claim 1 and a device according to claim 9. Advantageous embodiments and further developments are described in the dependent claims.

[0015] In a method for determining the permeability of a sample, the sample is exposed to a permeate, for example water vapor, and a temporal progression of the concentration and / or amount of the permeate permeated through the sample is measured, and the permeability is determined from the concentration or amount. The measurement is terminated before a steady state of permeability is reached, which is characterized in that the amount of permeated permeate is equal to the amount of permeate transported away and thus the change in concentration is zero and a constant permeability value exists, or after a predetermined time interval has elapsed. A final permeability value is calculated using a machine learning algorithm based on previously measured training data of a temporal progression of the concentration and / or the amount of permeated permeate and the measurement data determined so far.

[0016] By aborting or ending the measurement after a predetermined time interval or before reaching the steady state or equilibrium, the measurement duration is reduced compared to prior art methods, whereby the algorithm used allows the final value of the permeability, i.e. the value that would occur as a steady state of permeability during continuous measurement, to be predicted with sufficiently high accuracy. Advantageously, no additional equipment is required, i.e. no additional measuring devices are necessary, and the method can in principle be carried out with any of the measuring methods already known and used. The actual measurement can be carried out with the actual target molecule, e.g. water vapor or oxygen, and under the actual target conditions, in particular the intended temperature.

[0017] In this case, it can be provided that the permeability or the final value of the temporal progression of the permeability is calculated once before the steady state is reached or once after the expiry of the specified time interval. Alternatively, it can also be provided that the permeability is calculated several times at different times before the steady state is reached or at different times after the expiry of the specified time interval, with a previous calculation result for the permeability being taken into account for each new calculation. This makes it possible to achieve an early availability of the final result, whereby at the same time the accuracy of the prediction can be successively increased. In particular, the method is designed to determine a transmission rate as permeability. In particular, it can be provided to determine a water vapor permeability and / oror to determine oxygen permeability using the method mentioned.

[0018] The machine learning algorithm can be selected from an integrated autoregressive moving average model, a multi-layer perceptron, a convolutional neural network, a recurrent neural network, and / or a neural network with time embedding. By using artificial intelligence-based prediction models, valid predictions about the final measurement result can be made quickly and with high accuracy based on the training data or training measurement results obtained from previous, conventional measurements, which are used to train the respective model. Time series analysis methods and models are preferred for this purpose.

[0019] The predetermined time interval is preferably between 2 hours and 48 hours, particularly preferably between 12 hours and 24 hours, particularly preferably 24 hours. This means that the time course of the concentration or the amount of permeate permeated through the sample is recorded for a period of time predetermined before the start of the measurement, the duration of which is between 2 hours and 48 hours (or preferably has a value between 12 hours and 24 hours), before the measurement is terminated and the permeability is calculated or predicted from the measured data.

[0020] Typically, the time course of the concentration or amount of permeate permeated through the sample is determined using a calcium test, a coulometric measurement or diode laser absorption spectroscopy, i.e. a TDLAS method.

[0021] A device for carrying out a method with the described properties has a measuring chamber for carrying out the measurement and an evaluation unit for calculating the final value. The evaluation unit can also generally function as a computing unit and control the measuring chamber. The measuring chamber has a sample with a charged permeate, wherein the sample is designed to allow the permeate to permeate, wherein the sample divides the measuring chamber into a gas supply chamber and a permeate chamber. In addition, a gas supply unit is designed to introduce the permeate into the gas supply chamber. In addition, an inlet valve and an outlet valve are designed to introduce a carrier gas into the permeate chamber and to discharge it into the measuring chamber.In addition, a laser radiation source designed to emit a laser beam into the permeate chamber and a detector unit designed to measure the concentration of the permeated permeate are provided.

[0022] The sample is typically a film, preferably a barrier film, since permeability is an essential sample property.

[0023] A computer program product comprises a computer program having software means for carrying out a method having the described properties and / or for controlling a device having the described properties when the computer program is executed in an automation system such as the aforementioned evaluation unit.

[0024] Embodiments of the invention are illustrated in the drawings and are explained below with reference to Figures 1-6.

[0025] They show:

[0026] Fig. 1 is a schematic view of an apparatus for determining a water vapor permeability of a sample;

[0027] Fig. 2 is a schematic representation of the prediction of the course of a measurement after two different measurement times;

[0028] Fig. 3 shows a time course of water vapor permeability at different gas flows;

[0029] Fig. 4 a prediction of a final value of the water vapor permeability after 2 hours of measurement;

[0030] Fig. 5 a representation corresponding to Figure 4 after 12 hours of measurement and

[0031] Fig. 6 shows a representation corresponding to Figures 4 and 5 after 24 hours of measurement.

[0032] Figure 1 schematically shows a device for determining the water vapor permeability of a sample 1. Using the device shown, diode laser absorption spectroscopy, more precisely tunable diode laser absorption spectroscopy (TDLAS), can be carried out. The sample 1, in the illustrated embodiment a barrier film made of Aclar® PCTFE from Honeywell, is held in a measuring chamber 8. Water vapor 10 is introduced from a gas supply 3 arranged above the measuring chamber 8 into an upper part of the measuring chamber 8, which is also referred to as the gas supply chamber 4. This water vapor 10 permeates through the sample 1, with the sample 1 separating the measuring chamber 8 into the upper part and the lower part.The concentration or quantity of permeated water vapor 11 contained in the lower part of the measuring chamber 8, also referred to as the permeate chamber 5, is detected by a laser beam 9 emitted by a laser radiation source 6 with a wavelength of 1.38 pm via a detector 12. Nitrogen is supplied as a carrier gas into the lower part of the measuring chamber 8 via an inlet valve V3 and an outlet valve V4 from a nitrogen source 7. In addition, nitrogen lines are provided in the embodiment shown in Figure 1 for operating a ring purge, in which nitrogen is purged through the permeate chamber 5 in the direction of the arrow, thus shielding it from the ambient atmosphere.

[0033] The device is connected to a computer as a computing unit or evaluation unit 2. The inlet valve 3 and the outlet valve 4 can also be controlled by the computer as a control unit, i.e. the computer is set up to control the device shown in Figure 1 and to evaluate the obtained measurement data as described in more detail below. A computer program runs on the computer which, on the one hand, controls the device, but which can also further process the determined measurement values ​​and display them on an output unit, for example a screen. With the device shown in Figure 1, during a measurement in which the sample 1 is exposed to water vapor 10, a time profile of a concentration of the water vapor 11 permeated through the sample 1 is measured at the optical detector 12 and the measurement values ​​are transmitted to the computer as an automation unit.The water vapor permeability is determined by computer from the concentration, with the measurement itself terminating before a steady state is reached or after a specified time interval. A machine learning algorithm stored in the computer program calculates the water vapor permeability value that would result in a steady state based on the measured temporal variation of the concentration of the permeated water vapor 11 and the measurement data obtained so far.

[0034] The otherwise time-consuming measurement of quality parameters such as the permeability of sample 1 is drastically accelerated through the use of artificial intelligence, resulting in time savings of more than 90 percent. The basis of this measurement method is that the entire temporal progression of the measured value, i.e., the water vapor concentration, is also determined by the sample properties and thus also by the water vapor permeability of sample 1. This is achieved using predictive models that, after a relatively short measurement time, can predict the further temporal progression of the water vapor concentration in the TDLAS measurement shown as an example in Figure 1, and ultimately also the water vapor permeability of the sample 1 under investigation, based on the existing concentration progression.This enables the temporal behavior of a complex physical quantity to be predicted over the measurement duration in order to accelerate the determination of the final value of this quantity. This results in a drastic reduction in the time required to determine the water vapor permeability of the respective sample 1 and thus a significantly higher sample throughput per measuring device. The measurement can be carried out using existing measuring equipment; only the computer needs to be adapted using an appropriate computer program. The prediction models used are independent of the sample and also independent of the measuring device used (of an identical device class). A termination criterion can be flexibly specified by the user, i.e. the measurement can end after a certain time, usually predetermined before the measurement begins, or when a certain value is reached.This allows a constant choice between the desired accuracy and time savings. Furthermore, the method can be continuously improved in terms of accuracy and robustness by incorporating additional measurement data into the training data. The measured water vapor permeability is influenced by various aspects, whose contribution to the measurement result changes as the measurement progresses. At the end of the measurement, the measured concentration depends solely on the water vapor permeability of sample 1, and thus the water vapor permeability can be determined analytically.

[0035] A transfer of the generic principle to other quality parameters of barrier films (e.g., the oxygen transmission rate) and other measurement principles (coulometric, calcium test, image data, electrical conductivity) is possible, ie in further embodiments, a coulometric measurement or a calcium test can be carried out instead of a TDLAS measurement (the computer program with the training data used would have to be adapted accordingly in these cases).

[0036] The initial water content of the film as sample 1 and the measuring cell or measuring chamber 8 only play a role at the immediate start of the measurement. Only the absorption-desorption behavior of the surfaces of the measuring cell has a decisive influence until the end of the measurement. This behavior is described by an isotherm, but is unknown due to the complexity of the processes involved. This also means, however, that the course of the isotherm, i.e. the concentration of water vapor in the gas phase compared to the concentration of adsorbed water vapor, must have a cell-specific course, independent of the sample 1 examined. The isotherm of sample 1 itself is included in the water vapor permeability in this consideration, since the release of water vapor is part of the permeation process.

[0037] Figure 2 shows an example of the temporal progression discussed above in a schematic diagram. The calculated water vapor permeability is plotted against time. The solid middle curve represents the measured progression of the water vapor permeability, while the dashed upper curve shows the predicted progression after the end of the measurement at time t1. The dashed lower curve shows the corresponding predicted progression at the end of the measurement at time t2. It is therefore possible to predict the further progression up to equilibrium based only on the initial phase of the measurement, which enables an early determination of the final value of the water vapor permeability. The prediction can be repeated as often as required as soon as new measured values ​​become available, whereby the quality of the prediction gradually improves and approaches the actual value of the water vapor permeability.Ideally, a reliable result can be obtained after a much shorter time than the actual measurement time required.

[0038] Figure 3 shows a diagram of a measurement process in which different gas flows are used, in a view corresponding to Figure 2. By frequently changing the gas flow, the information density of the measurement for the prediction can be increased. This occurs both when the carrier gas flow is reduced and when the carrier gas flow is deliberately increased. The resulting changes in the gas atmosphere in the measuring cell have a direct influence on the contribution of the measuring cell isotherm to the measured concentration values. Likewise, a defined gas flow change can occur at defined times, or a temporal variation of other measurement parameters such as temperature or relative humidity can be carried out.

[0039] As already explained, the actual prediction of the measurement trend is carried out using artificial intelligence methods, particularly deep learning. Various methods for predicting time series can be used. These include, but are not limited to: auto-regressive integrated moving average models, multilayer perceptrons, convolutional neural networks, recurrent neural networks such as long short-term memory networks, echo state networks, or gated recurrent networks, and / or neural networks with time embedding.

[0040] The models are trained using the results of real measurements of film samples, i.e., the temporal progression of the calculated water vapor permeability as a function of the specified carrier gas flows. The models determine a functional relationship between a known progression of the water vapor permeability (during the initial phase of the measurement) and the progression in the subsequent time phase, up to the constant target value of the water vapor permeability.

[0041] The prediction of the future course can be performed as a "single-shot" prediction, i.e., a prediction of all subsequent time steps in a test, or as an autoregressive prediction, i.e., a step-by-step prediction using the prediction result of the previous step. For example, the initial measurement times between 12 h and 48 h can be selected as input data. In addition to the water vapor permeability curve at the beginning of the measurement, the models also require the corresponding gas flow. This gas flow is also made available to the models for the prediction. For successful model training, scaling of the data (e.g., normalization, standardization, logarithmization) may also be necessary.

[0042] A trained prediction model can then predict the further course of the WVTR at different times (tl, t2, t3, ...) after the start of the measurement and thus allows an early estimation of the value of the sample 1 under investigation.

[0043] Figures 4, 5, and 6 show the values ​​predicted using the described method and the actually measured values. The measurements were obtained as follows: A data set of 30 TDLAS measurements of the water vapor permeability of barrier films in the WVTR range between 10 1 g / m 2 d 1 and 10 -3 g / m 2 d 1used. The barrier films are films with two different layer systems. The TDLAS-WVTR measurements were carried out using devices of the same design, i.e. the permeation cells used were almost identical. The typical measurement time until a constant water vapor permeability was reached was approximately 30 to 200 hours. The measurements were converted to time steps of 10 minutes and logarithmized. The data set was divided into a training data set (25 samples) and a training data set (5 samples). The measurements were also divided into subsequences: an input sequence based on which the target sequence, the further course of the measurement, is to be predicted. For each of the measurements, the course between a measurement time of 2 h and 24 h and the complementary target sequences were selected as input sequences.

[0044] The resulting sequences were used as data for training, optimization, and validation of a sequence-to-sequence long-short-term memory network model. The model's hyperparameters were optimized, and a final model was trained.

[0045] The resulting model was applied to the test data, and the WVTR curve was predicted after a measurement time of 2 to 24 hours. Based on the predicted curves, the final WVTR value of the films was determined and compared with the WVTR values ​​determined from the TDLAS measurements. The results are shown in the figures for a measurement time of 2 hours (Figure 4), 12 hours (Figure 5), and 24 hours (Figure 6). It can be seen that the accuracy of the prediction increases with increasing measurement time, and after 24 hours at the latest, there is very good agreement between the values ​​predicted by the model and those determined by TDLAS. The model can therefore predict the further course of the measurement and thus the actual water vapor permeability of test sample 1 with high accuracy even after a relatively short measurement time.

[0046] The described method can be used to determine quality parameters such as water vapor permeability or oxygen permeability of barrier films in the food industry, pharmaceuticals, inorganic solar cells, organic electronics such as organic solar cells or organic light-emitting diodes, and vacuum insulation panels.

Claims

Patent claims 1. Method for determining the permeability of a sample (1), wherein the sample (1) is exposed to a permeate (10) and a time course of a concentration and / or quantity of the permeate (11) permeated by the sample (1) is measured, and the permeability is determined from the concentration and / or quantity, wherein the measurement is terminated before reaching a steady state in which a time constant value of the permeability is present, or after the elapsed time interval, and a final value of the permeability is calculated by a machine learning algorithm based on previously measured training data of a time course of the concentration and / or quantity of the permeated permeate (11) and from the measurement data determined so far.

2. Method according to claim 1, characterized in that the final value of the permeability is calculated once before reaching the steady state of permeability or after the expiry of the specified time interval.

3. Method according to claim 1, characterized in that the final value of the permeability is calculated multiple times at different times before reaching the steady state of the permeability or after the expiry of the specified time interval, wherein a previous calculation result for the permeability is taken into account in each recalculation.

4. Method according to one of the preceding claims, characterized in that the machine learning algorithm determines the final value of the permeability by an integrated autoregressive sliding mean model, a multilayer perceptron, a convolutional new- A neural network, a recurrent neural network, and / or a neural network with time embedding is calculated. A method according to one of the preceding claims, characterized in that the measurement is carried out over a period of between 2 h and 48 h before termination. A method according to one of the preceding claims, characterized in that the measurement is carried out over a period of 24 h before termination. A method according to one of the preceding claims, characterized in that the measurement is carried out with water vapor and / or oxygen as permeate (10). A method according to one of the preceding claims, characterized in that the time course of the concentration and / or the amount of the permeate (11) permeated by the sample is determined by means of a calcium assay, by means of a coulometric measurement, or by means of diode laser absorption spectroscopy.Apparatus for carrying out the method according to one of the preceding claims, comprising a measuring chamber (8) for carrying out the measurement, comprising: a sample (1) with a permeate (10) applied to it, wherein the sample (1) is configured to allow the permeate (10) to permeate, the sample (1) dividing the measuring chamber (8) into a gas supply chamber (4) and a permeate chamber (5); a gas supply unit (3) configured to introduce the permeate (10) into the gas supply chamber (4); an inlet valve (V3) and an outlet valve (V4) configured to are a carrier gas for introducing and removing it from the permeate chamber (5); a laser radiation source (6) configured to emit a laser beam (9) into the permeate chamber (5); a detector unit (12) configured to measure the concentration of the permeated permeate (11); and an evaluation unit (2) for performing the calculation of the final value. Device according to claim 9, characterized in that the sample (1) is configured as a film. Device according to claim 10, characterized in that the sample (1) is configured as a barrier film. Computer program product with a computer program comprising software means for carrying out a method according to any one of claims 1-8. Computer program product with a computer program comprising software means for controlling a device according to any one of claims 9-11 when the computer program is executed in an automation system.

Images