DE10005130A1 - Control of polymerization, comprising measuring near OH, NH, COOH, NCO and/or epoxide IR absorption bands using device connected to spectrometer by glass fiber cable - Google Patents

Abstract

Control of polymerization involving measuring near OH, NH, COOH, NCO and/or epoxide IR absorption bands to determine the chemical groups in comparison to calibration data using a device connected to a spectrometer by glass fiber cable. Control of polymerization involving measuring near OH, NH, COOH, NCO and/or epoxide IR absorption bands to determine the chemical groups in comparison to calibration data using a device connected to a spectrometer by glass fiber cable. An independent claim is included for the apparatus used comprising a transmission IR source, a Fourier transformation spectrometer including an interferometer or diode array.

Description

The present invention relates to a method for controlling polymerization reactions in which a measurement of infrared absorption bands Determination of chemically functional groups takes place, as well as a device for Performing this procedure.

The chemical processes in the manufacture of polymers, especially those for the use in paints, adhesives and the like are usually by the Controlled measurement of specific indicators. In particular, at Polyaddition reactions or polycondensation reactions the progress of Degree of polymerization by determining the content of functional groups measured. For example, in the manufacture of polyesters from di- / triol- and Di-tricarboxylic acids measure the acid number or with polyadditions like that Polyurethane resin production the measurement of the NCO content.

These methods have the disadvantage in production processes that the key figures in Operating laboratory must be determined. A sample of the Reaction mass pulled out of the reactor, brought to the operating laboratory, there accordingly prepared the regulations for determining the key figures and then wet-chemical measured for their content of functional groups by titration. With this Procedures can occur during sample transport and sample preparation. 50 polyurethane resins are generally by polyaddition of diisocyanates Diols in solvents, for example methyl ethyl ketone or acetone, at temperatures of 60 ° to 100 ° C. Due to the transport and sample preparation it can be too Solvent losses, for example due to evaporation, so that the determination of functional groups no longer refer to the original material from the reactor, but relates to material modified in polymer content. Another disadvantage is that depending on Duration of the salary determination of the value typically 15 to 25 min after the Sampling received so that only a staggered value is determined and none Real-time determination takes place. Are polycondensation or polyaddition reactions led to a narrowly defined target window of the degree of polymerization, can by  the duration of the titrimetric determination already exceeded the target window in the reactor without the operator being aware of it.

EP 0 694 781 A1 discloses a method by which the problems mentioned are solved an online measurement should be avoided. The liquid reaction mass is in a bypass through a measuring cell which contains an ATR crystal (ATR = Attenuated total reflection). According to the method of attenuated total reflection the absorption bands of the typical for the reaction functional groups measured and based on the changes in these bands the Reaction progress determined. The measurement takes place in the middle infrared range (MIR), i.e. at a wavelength of 600 to 4000 cm. The disadvantage of that there The disclosed method is that the reaction mass is conveyed through a bypass must what an additional constructive effort in the design of the reactor means and involves the risk that one for the reaction mixture in the reactor atypical subset is measured. There is also a risk that the bypass clogged and that the measuring cell is contaminated with product from previous batches becomes. Due to the very short penetration depth of the light rays at the ATR, there is a risk of that the measurement is primarily in a liquid layer adhering to the ATR crystal which is not exchanged at the necessary rate. Another disadvantage is that the light rays of the middle infrared range only over short distances between Measuring cell and spectrometer can be directed so that when using the method in Production companies must use very expensive explosion-proof equipment.

The object of the present invention was to overcome the disadvantages of the prior art avoid and a process to control polymerization reactions ready which allows control without delay and is also inexpensive in terms of investment and operating costs.

This object is achieved by a method in which the measurement of infrared Absorption bands in the near infrared (NIR) range is made and in essential first and second harmonics of OH, NH, COOH, NCO and / or Epoxy groups recorded, in which the measured values by comparison with Calibration data sets are evaluated and in which the measuring probe immediately arranged in the reactor and connected to a spectrometer via fiber optic cables.  

With the method according to the invention, polymerization reactions can be measured and observed spectroscopically online, and with the aid of previously determined calibration data, real-time determination of functional groups in the polymer mass can take place. Sampling is no longer necessary. By choosing light from the near infrared range (approx. 4000 cm ^-1 to approx. 14000 cm ^{- 1} ), the light can be transmitted over longer distances (up to 200 m) using suitable glass fiber cables, so that the measuring probe and the Spectrometers for evaluation can be set up separately. It is in particular possible that the probe with the associated glass fiber connections is located in the potentially explosive area of the production system, while the spectrometer is separate from it in a separate laboratory or measuring room. It is therefore not necessary to carry out a cost-intensive effort for the explosion protection of the spectrometer. Rather, standard devices can be used for this. In addition, it is advantageous that the spectrometer can be set up where there is sufficient space and where there are optimal working conditions for the operator.

Infrared spectroscopy can be used to measure the natural frequencies of molecular vibrations, especially the vibrations of functional groups compared to the rest of the molecule. For quantum mechanical reasons, the vibrational energy of a molecule can only assume discrete values, which, for example in the case of a twelve-atom molecule, can be approximated by the formula based on the model of an anharmonic oscillator

E (n) = hν (n + ½) - h ² / 4D (n + ½) ²

(h = Planck's quantum of action, ν = 1 / T = frequency, D = spring constant, n = 0, 1, 2, ... quantum number)
can be described. The fundamental vibrations of a molecule relate to transitions between the quantum numbers n = 0 and n = 1, in which the highest degree of occupation is present. In contrast, transitions from n = 0 to higher energy levels n <1 are referred to as "harmonics" which only occur with the anharmonic oscillator. Changes in the quantum number n by ± 2, ± 3,. , , allowed. In contrast to a measurement in the medium infrared range (MIR), measurements in the NIR range almost exclusively record the harmonics. The absorption bands have a lower intensity in the NIR range than in the MIR. This has the advantage that the layer thickness of the sample to be examined can be increased to an easier to handle size of typically 1 to 10 mm, which simplifies the use of NIR probes. Furthermore, this rather guarantees the representativeness of the measurement result.

Another advantage of a measurement in the NIR range is that it extends there Options for materials for optical windows and optical fibers consist. For example, glass, quartz and, for high pressures, sapphire are also due usable for their permeability. In addition, these materials have compared to the substances used in MIR such as NaCl and KBr better chemical and mechanical Properties on. Finally, allow instrumental advantages with regard to the light source and the detector in the near infrared record measurement signals with a Signal to noise ratio of well over 10,000, which is generally in the mid-infrared is not available.

The NIR light used for the measurement is preferably taken from a range of 4000 cm ^-1 to 11000 cm ^-1 . This corresponds to a wavelength of light from 2500 nm to approx. 900 nm.

To analyze the measured absorption spectra, a Fourier Transformation (FT) spectrometer used. The main difference of this Compared to previously developed instruments, device types are of the type Light decomposition. With classic IR spectrometers, this is done by Grid monochromators. The FT-IR spectrometer is used instead of the monochromator a modulator is used. Each wavelength of the incident light is with their own frequency modulated. The resulting interferogram is based on absorption demodulated in the sample by a Fourier transformation, that is, in a spectrum converted. This allows for increased light intensity and speed Measure spectra with resolution and accuracy that are not with grating spectrometers is achievable. The Fourier transformation is preferably carried out using the method of fast Fourier transform (Fast Fourier transform FFT).  

Quantitative analysis using NIR spectroscopy is required for all applications a calibration must first be carried out. This is based on the assignment of the spectra derived quantities to a measured value by a reference method (for example measurement analysis determination) was determined. When choosing the Calibration samples should be taken into account that the entire concentration range covered by the Analysis of unknown samples can occur is covered by calibration measurements. Calculating a quantitative relationship between spectra and Concentrations of substances using a mathematical model are preferably carried out with Help of a computer. Chemometric methods are preferably used here, which are various multivariate methods of numerical mathematics acts, all based on the evaluation of a variety of wavelengths / extinction Pairs of values are based. The advantage of chemometry is that it is independent of the Matrix in which the component to be determined is located.

A suitable chemometric method is the factor analysis of full spectra. The The calculation is preferably carried out using the least squares method (PLS = Partial Least Square). This method can be used, for example, with the Computer program "Opus" can be carried out. The procedure is a so-called inverse method, in which the concentrations of the individual Components are decoupled. This means that those who are interested in analysis are interested Have components calibrated without knowing the rest of the composition. The Selection of the wavelengths or factors that result in the optimal calibration takes place automatically using the least squares method. Doing so will be statistical Calculated quantities that allow the quality of a calibration to be assessed.

In the case of absorption measurements, one can assume that there is a proportionality between the spectral intensity and the concentration of the ingredients of the substance mixtures examined (Lambert-Beersches law). The PLS decomposition assumes the following dependency on component values and spectral intensities:

y = [X] b

The vector y contains the component values determined by measurement analysis and the matrix [X] the row vectors of the value pairs (wavenumber / absorption) of the calibration spectra. The aim of the calibration is to determine the vector b so that an unknown value y _n can be calculated using a spectrum X _n .

In the polymerization reaction controlled by the process according to the invention it can in particular be a polyaddition or a polycondensation. Through the process, both the change in the educt bands (decrease) and also the change in product bands (increase) measured.

The polymer can in particular be a polyurethane, the NCO- and / or NH and / or urethane groups are measured. Experiments have shown that the inventive method in the polymerization of polyurethanes in particular shows good results.

In addition, the polymer can be a polyepoxide, with epoxy and / or OH and / or NH groups are measured.

If polycondensation is controlled, the polycondensate may be a polyester, a polyamide, a polyether acrylate and / or a polyester acrylate.

The invention further relates to an apparatus for performing a method of Above described type, which is a transmission infrared probe made of ceramic or Contains quartz, which can be arranged in a polymerization reactor, and a Fourier transform spectrometer, which is connected to the NIR probe via fiber optic cable connected and an interferometer (e.g. Michelson interferometer or Prism interferometer) or a diode array. With such a device the advantages of the method described above can be achieved. In particular is It is advantageous that the measuring probe and spectrometer are separated using the glass fiber cable can be set up so that the spectrometer outside the hazardous area can be accommodated.

According to the invention, it is advantageous if the NIR probe is covered with a protective jacket Metal, such as stainless steel, is surrounded to protect it from mechanical damage protect. To prevent the formation of deposits on the NIR probe during resin synthesis prevent, with an advantageous configuration of the protective jacket is still one on Sapphire or quartz window of the NIR probe directed rinsing nozzle available. This will flowed through with a gentle stream of nitrogen during the resin synthesis to the  Prevent clogging of the nozzle. Immediately after draining the resin from the The reactor becomes hot or hot from a storage container located outside the reactor pumped warm solvent and sprayed through the nozzle directly onto the probe window, whereby any existing topping is washed away. Devices of this type are common and known and are for example from the company O.K.Tec Optik + Ceramics + technology sold.

NIR spectroscopy not only allows the progress of the reaction to be measured Determination of the appropriate functional groups after calibration, but Special programs can also make comparisons between a running batch of a product and the previous batches of the product. One finds chemical identification instead of a fingerprint of the reaction mass instead. This is only possible in the NIR range, since the direct vibrations of CO, CH, CN, NH, etc. bands are measured, but the corresponding harmonics. With the Fourier transformation, a complete spectral overview is obtained about the reaction mass and can by comparing with historical data Deviations, for example due to faulty raw materials or incorrect weights, immediately determine and take necessary corrective measures.

Examples

The following studies should assess the applicability of NIR in-line spectroscopy for the reaction tracking of resin synthesis. Since the changes in the NIR Spectrum on the different chemical changes in the different Resin formation reactions are due, it is necessary to different from Select product classes syntheses and the key figures to be checked with the to correlate changes obtained from the NIR spectra. In preliminary experiments in Laboratory has shown that good NIR spectroscopy for polyurethane synthesis Results are expected. Therefore, the focus of the experiments in The pilot plant scale also applies to this reaction class.

Another important aspect for the quality of the calibration is the number of measured values, on which the calibration is based. Due to the sampling and the time consuming  Measurement intervals of less than 15 to 30 minutes are not carried out for wet chemical analysis realizable. To provide a sufficiently large number of measurements determined by measurement Calibration values were therefore reactions with the slowest possible course of the reaction selected.

Based on the criteria outlined above, a. the one shown in Table 1 Test plan, which contains four products from the field of polyurethanes (PUR), compiled:

Table 1

Explanation

Target range = target range for the measured value. There are two for PUR syntheses Target areas: The first specification is the target area before adding trimethylolpropane (TMP) and the second is the target range after implementation with TMP. Each attempt was made four times carried out (three reproductions).

The reaction conditions are outlined in the following. Both In an addition reaction, polyurethane (PUR) syntheses become a diisocyanate (e.g. Isophorone diisocyanate) with various polyhydric alcohols (e.g. neopentyl glycol, Trimethylolpropane), a dihydroxy carboxylic acid and a polyester resin in one Temperature from 80 ° C to 90 ° C implemented. The reaction takes place in two stages, whereby in the second stage a complete implementation of the remaining free isocyanate groups is sought.

The preselected temperatures were accurate to ± 0.5 ° C during the measurement phase adhered to. To reduce the reaction rate, the 1st reproduction  Synthesis 1 deviated from the standard temperature 90 ° C with 80 ° C. This temperature difference had a great influence on the NIR spectroscopic Measurement so that this synthesis is not shared with the others Reproduction syntheses could be calibrated. The temperature of the synthesis must be for a calibration data set can be kept constant, otherwise the spectroscopic determined concentrations and their reference values are no longer comparable.

All of the PU syntheses carried out showed a high viscosity curve Level with values (original plate-cone viscometer at 23 ° C) of well over 100 dPas. The viscosities in dissolved form (50% in N-methylpyrrolidone) reached values up to 20 dPas.

The following table contains the detailed reaction conditions:

Table 2

Reaction conditions of polyurethane synthesis

Reaction apparatus for synthetic resin synthesis

Fig. 1 shows the reactor line 430 in schematic form. In the upper part of the figure, the three storage containers (R 431, R 432 and R 433) are shown. The storage container R 431 has a nominal capacity of 400 l, the R 432 of 40 l and the R 433 of 250 l. The templates R 431 and R 433 are used to weigh in monomers or solvents, while the R 432 is used to introduce initiators. The templates R 431 and R 433 can be cooled, the template R 433 can also be heated. The pressure and temperature in the monomer storage tanks (R 431, R 433) are monitored by the process control system (PLS). This ensures that autopolymerization is displayed in the templates at an early stage (alarm when the pressure or temperature increases).

The masses of all three templates are recorded using mass sensors. The mass flows which are dosed from the templates into the R 430 reactor can be created with the help of the PLS and controllable valves are regulated.

In the middle of the figure, the actual reactor can be seen along with the inputs for the in-line NIR probe. It is equipped with a vapor tube and a horizontal condenser, as well as with a separation template and the associated collecting container. With this arrangement, multi-phase mixtures that have been removed by distillation can be separated and recycled or discharged (e.g. in the case of rectifier rectifications). The reactor has a nominal volume of 400 l and can be inertized with nitrogen. It is equipped with a two-stage INTERMIG® agitator. In order to generate additional turbulence in the reactor and thus to ensure better mixing and an improved heat transfer, four flow breakers are attached to the reactor wall. The speed of the agitator is continuously adjustable from 21-126 min ^-1 . At the maximum speed of n = 126 min ^-1 , the peripheral speed is 5.2 m / s. The R 430 reactor can be heated (max. Product temperature approx. 240 ° C) and cooled. The heat is transferred using thermal oil, which is tempered in a heat exchanger system.

To the right below the R430 is the release boiler R434, which in turn has a Burner tube and a condenser. The temperature in the dissolving kettle is determined by a Bavarian mixer regulated by mixing water and steam.

The equipment for carrying out the spectroscopic investigations consisted of the NIR spectrometer, the connected probe and a personal computer. In addition to the high demands on the optical properties of the measuring equipment, the quality of the hardware and software used for data processing is of great importance. A personal computer (486 DX 2 ) with a clock frequency of 66 MHz and 8 MB RAM was therefore used to provide the required computing power.

The NIR probe from O.K.Tec (type TS- EK1) was installed in the bottom of the reactor and via a glass fiber cable (Fa. Bruker, length: 50 m) with an outside of the hazardous area connected FT-NIR spectrometer IFS 28 N from Bruker. The for the  Materials used in the probe are aluminum oxide for the stem of the probe and sapphire for the windows. The probe is for a wavelength range of 400-2400 nm usable. It is designed for temperatures up to 350 ° C and for pressures up to 20 bar.

Single-channel spectra are recorded with the spectrometer, so that the Comparison with a background spectrum registered before each measurement is necessary. This serves as a reference, so that long-term changes such as for example a change in intensity of the radiation source or a permanent one Change in the sapphire window of the probe (dirt or damage) be taken into account.

After recording the background spectrum and the sample single-channel spectra, that remains Preserved the background spectrum in the working memory until the end of the program. The division of sample single-channel generated by background single-channel spectrum Transmission spectrum, which in turn is converted into an absorption spectrum can be.

Storage of the data obtained in this way and the interferogram can be optional be made. Because for the later evaluations only the Absorption spectra and their background spectra were only needed Data saved.

When looking at the spectra, it is noticeable that the baseline is not zero, but is shifted by an offset into the negative range. However, negative absorption is not possible. This parallel offset of the spectra is due to the different media in which the background and sample spectrum are measured. The background spectrum is recorded in a gas (nitrogen or air) and the sample single-channel spectrum in a liquid. The refractive index of liquids is a multiple of the refractive index of gases and is many times lower than that of solids. The NIR light beam passes through the following media in the measuring arrangement:

• 1. in the background measurement:
  Sapphire Window (Solid) - Nitrogen (Gaseous) - Sapphire Window (Solid)
• 2. during sample measurement:
  Sapphire Window (Solid) - Sample (Liquid) - Sapphire Window (Solid)

The difference in the refractive indices is larger in the background measurement than in the Sample measurement so that when recording the background spectrum the light is stronger is broken than in the sample measurement. This stronger refraction is with a higher one Equal loss of light and the intensity recorded by the detector decreases, so that in the generation of the absorption spectrum (measurement of the resin samples) in comparison an "apparently negative" absorption occurs.

The main task of the personal computer is to calculate a quantitative relationship between spectra and substance concentrations using a mathematical model. Chemometric methods are used for quantitative analysis. The great advantage of chemometry is that it is independent of the matrix in which the component to be determined is located. The calibrations created were carried out using a method based on factor analysis of full spectra. The calculation was carried out using the method of least squares (partial least square). The Opus program used for this has the following properties:

• 1. It is a so-called inverse method, in which the concentrations of individual components are decoupled, d. H. those of interest for analytics Components can be made without knowing the remaining composition calibrate.
• 2. The selection of wavelengths or factors that provide the optimal calibration result is done automatically (method of least squares).
• 3. Statistical quantities are calculated that allow the quality of a To assess calibration.
• 4. The chemometric evaluation process is carried out with different ones Problems always in the same way, so that the calibration procedure does not always has to be redeveloped.

For quantitative analysis, a calibration is carried out with the following five steps carried out:  

1. Set Calibration List

A calibration list is created, which from the recorded in time sequence Absorption spectra and the associated reference values (determined by measurement analysis) consists. There is also the option of converting the calibration standards into a set of Calibration and to divide into a set of test spectra. So the goodness of the Calibration can be checked using the test data set.

2. Set Frequencies

Under this menu item a spectral range relevant for calibration can be found selected or an area overlaid with noise can be eliminated.

3. Data preprocessing

Data preprocessing enables a mathematical pretreatment of the Calibration spectra. So z. B. different offsets with differentiation be eliminated.

4.PRESS (Predictive Residual Error Sum of Squares)

The PRESS method serves as an aid for determining the optimal one Calibration parameters, d. H. Here is the possibility of the quality of the PRESS to determine the calibrations performed. When calculating the PRESS can choose between two types of validation of the quant model. This is for one is cross-validation (internal validation) and the other is the verification of Calibration with the help of a control data record with the so-called external validation. After each PRESS calculation, a PRESS report is issued.

5. Calibrate

With the parameters optimized during validation, the "final" calibration carried out. The data important for the analysis of future samples (calibration spectra with the respective measured values, the frequency range (s) and the selected Data preparation) are stored in the Quant II method file.

Results of measurements for polyurethane synthesis

Four different polyurethanes (Product Nos. 1, 2, 3, and 4) were synthesized according to the experimental plan shown in Table 1 above. FIGS. 2-5 each show 4 spectra for these 4 products, which reflect the appearance of the bands over the entire reaction period.

In general, a significant change in a band at 6750 cm ^-1 can be seen in all polyurethanes, which is attributable to the NH valence vibration (sec. Amine: fundamental vibration at 3300 cm ^-1 to 3500 cm ^-1 ) of the synthesized urethane. The band size increases in the course of the synthesis. Other changes that are difficult to see visually occur in the range from 5000 cm ^-1 to 6000 cm ^-1 . Gangs are represented here, the intensity of which increases and decreases as the synthesis progresses. No exact assignment can be made, but these bands are of great importance for the calibration.

The spectra in the figures show an offset of the baseline. This is on the Change in the turbidity of the reaction mixture in the course of the syntheses. At the beginning, when the turbidity is present, more NIR radiation is absorbed than at the end the reaction when the turbidity is no longer present. It follows that the baseline the absorption spectra in the course of the synthesis to lower absorption values "wanders". One can distinguish between this offset of the spectra and the NCO values establish a mathematical connection. The offset based on turbidity change does not occur with the same intensity in every synthesis. Therefore a calibration is the exploiting this parallel offset of the spectra does not make sense.

In the following plots of the spectra are the times when the spectra were determined from the respective figure. The times are given here to be interpreted in hours (h) in relation to the first sampling. A label without unit specifies the order of the spectra acquisition.

The spectra of FIG. 2 (product No. 2) show a lower absorption band at 6750 cm ^-1 than the other polyurethanes at the same concentration. One reason for this could be the different chemical composition of these polyurethanes. As a result of this difference, a joint calibration of product No. 2 with the other polyurethanes is not possible.

While the direct comparison of the spectra in FIGS. 3-5 shows the great similarity of the bands in the range from 5000 cm ^-1 to 5500 cm ^-1 , and from 6100 cm ^-1 and 7000 cm ^-1 , are in the range of 8000 cm ^-1 to 9000 cm ^-1 different band shapes can be seen. The bands of product No. 4 show a rounded course in this area, while the bands of polyurethanes No. 1 and No. 3 show sharper contours. These contours are most pronounced in the No. 1 polyurethane ( FIG. 5). The spectra of the different polyurethanes at 6750 cm ^-1 shown in FIGS. 3 and 4 show approximately the same absorption at the same isocyanate ^contents , so that the basic prerequisite for the common calibration is given.

In the following, special changes in the spectra during the course of the polyurethane syntheses are exemplified for the product No. 1 on 6 spectra in FIGS. 5 and 6. FIG. 5 shows four spectra during the reaction of the diisocyanate with the first polyol. With a wave number of approx. 7000 cm ^-1 , a decrease in the intensity can be observed, which is due to the decrease in the concentration of the alcohol (1st harmonic of the OH functionality). This change at 7000 cm ^-1 is used to create calibration algorithms for determining the NCO content.

Since TMP is added to bind the remaining free isocyanate when a certain degree of polymerization is reached, there is a correlation between the OH band (at approx. 7000 cm ^-1 ) and the NCO value only up to the addition of TMP. After the addition, this correlation no longer exists, which is illustrated by the two spectra shown in FIG. 6. The alcohol gang increases and at the same time the NCO-Wen decreases. The effect that the calibration is based on such opposing or more or less "accidental" or non-reproducible relationships has to be eliminated by choosing a suitable calibration data set. In this case, the calibration data set must contain spectra before and after TMP addition, so that the entire PUR synthesis can be monitored with a calibration function in later measurements.

Various data pretreatments were used to determine the calibration function tested. For the calibrations in the context of the PUR syntheses, the 1st derivative resulted as Data pretreatment gives the best results, so all calibration functions for this  Resin class were created using this procedure. It has already been so Correlation coefficients of 98 and more achieved, which with a maximum, absolute Error of the NCO values in the order of magnitude of ± 0.1% NCO is equivalent.

An additional improvement in the calibration could be achieved by restricting the Frequency range around 0.5% can be achieved. The standard deviation was in same size (reduced by 0.5% relative). Another way to get one You can influence the calibration by selecting the number of PLS Vectors (number of ranks) given. If the above mentioned changes It should be noted that when modifying a parameter (e.g. the data preparation) the settings of the other parameters (e.g. the Frequency range) must be re-optimized.

After evaluation using the variation options described above, the values shown in FIG. 7 were obtained for the syntheses examined.

In the following, the quality of the calibration function for individual products discussed. The figures (True Predict Mapping and No Differ Mapping) are taken directly from the OPUS software. The True Predict figures show one Comparison of the measurement-analytically determined reference values (True) and those of the PRESS calculation calculated values (Predict). The diagonal in this diagram there the location of the concentration values again, which is based on the measurement analysis determined calibration function can be calculated. Points that are above the diagonals, stand for lower measured values, Points below the diagonal gave larger values in the dimension analysis than in the Calculation.

The no-difference figures are also taken from the OPUS software. You show that Deviations that exist between the reference values and calculated values. To For this purpose, the individual measured values were numbered. The numbers (No) can be found on the abscissa. The differences mentioned are on the ordinate (Differ) plotted.

For an easier interpretation of the True Predict diagrams, an overview of the the actual course of the reaction is in addition a typical concentration-time course  pictured. In addition to illustrating the course of the reaction, this is also intended to Estimate the error of the dimensional analysis.

Fig. 8 shows the characteristic of a two-step synthesis of PUR-reaction. It can be seen that in the case of large NCO values, there are relatively large absolute scatterings of the NCO contents determined by measurement around the interpolated concentration-time curve. In comparison, this scatter decreases significantly with small NCO values.

The relationship between the calculated and measured values for the synthesis of product No. 1 is shown in FIG. 9. In FIGS. 8 and 9 it can be seen that the end of a reaction step, the temporal change in the NCO content decreases, which makes in Fig. 9 by a reduction in the distance between individual values on the diagonal noticeable.

For a better demonstration of the deviations between measurement-determined and calculated values, the differences from reference values (True) and calculated values (Predict) for the individual samples are plotted in FIG. 10. From the figure it can be seen that maximum deviations between the calculated and the analytically determined value of 0.108% in absolute terms are not exceeded with an NCO content of 1.4%. This maximum deviation is presumably due to errors in the measurement analysis determination, since the measurement analysis value, assuming an almost linear concentration-time curve, should be between 1.44% and 1.07%, i.e. around 1.26%. An NCO of 1.30% was determined using the spectroscopic method. A value of 1.41 was determined using the dimensional analysis. This shows that the errors of the spectroscopic method indicated have a great influence on the errors caused by measurement analysis.

The Fig. 10 can also still be seen that the deviations of the individual values are statistically almost normally distributed. The greatest deviation for values 4 and 17 is of the order of 0.1% NCO (absolute). The other values show a deviation of 0 to 0.05% NCO (absolute). The calculations of the individual values are subject to approximately the same absolute deviations for large and low values. Deviations from 3% to 14% of the NCO value are found in the range from 0.18% to 0.68%.

The individual calibrations of the reproductive syntheses of product No. 1 and the two Polyurethanes No. 3 and No. 4 gave comparable results when calculating the Individual values. The deviations are between 0.06 NCO-% and 0.10 NCO- % approximately in the same range as in the synthesis of product No. 1 discussed above. It is therefore on the graphical comparison of calculated and dimensional analysis determined values waived.

The individual calibrations of product No. 2 showed very different deviations between 0.036 NCO% and 0.2 NCO%. One reason for this is in the influence of the error the measurement analytical determination seen in the individual syntheses in different degrees occurred. Approximately the error of the dimensional analysis Process for this product estimated at 0.1% to 0.3% of the NCO content and is noticeably higher than that of the other polyurethanes. Cause of these great Deviations were severe precipitations during the titration of the product samples. The The end point was not clear or the color change from blue to yellow did not remain consist. According to the test instructions for the NCO determination, the end point is the Titration achieved when the solution remains yellow for 30 s. Therefore titrated the NCO values of product No. 2 until the reaction was complete (Titration time: 20 to 30 min). During this long titration period, one NCO content of 0.8% changes up to ± 0.3%. A calibration of the spectroscopic method is used in this case with the help of here measurement analytical determination not possible or only with large errors.

Another essential part of the studies on polyurethane synthesis is the Testing the applicability of a calibration function for calculating the NCO content a reproduction synthesis. For this purpose the calibration function of the Synthesis of product no. 1 / FN 5995112 to calculate the reference values of the Reproduction syntheses with the fill numbers (FN) 5995110 (product no. 1) and 5995184 (Product No. 1) used. The achieved in the so-called validation Standard deviations for fill numbers 5995110 (product no. 1) were 0.055 NCO% and for the filling number 5995184 (product No. 1) 0.095 NCO%. Therefore the two syntheses with the fill numbers 5995112 (calibration synthesis Product No. 1) and 5995184 (reproduction synthesis product No. 1) in the entire course examined by means of measurement analysis and NIR spectroscopic methods. in the  The difference was the synthesis with the filling number 5995110 (product no. 1) stopped after the 1st reaction stage (before TMP addition), so that in this case none NCO values in the range below 1% are available.

FIGS. 11 and 12 show the deviations of the volumetrically determined from the calculated values for the synthesis FN 5,995,110 (product no. 1) and 5,995,184 (product no. 1) based on the calibration function of the synthesis FN 5,995,112 (product no. 1). The value with the greatest deviation was determined with a measurement-analytically determined NCO content of 1.41% with a spectroscopically determined value of 1.72%, ie a difference of 0.31%. The reason for this large difference is the deviation of the product temperature at this point from the setpoint of 90 ° C. The product was cooled overnight; the exact temperature (80 ° C-90 ° C) cannot be exactly quantified. The influence of the temperature is estimated on the basis of a spectrum in the heating phase, for which both the exact temperature value and the reference value determined by measurement analysis were available. At this time the temperature was 80 ° C (10 ° C below target), the NCO content determined by measurement was 1.44%. Using the calibration function for the synthesis of product No. 1 FN 5995112, on the other hand, an NCO content of 2.14% was calculated, ie a deviation of 0.70%; this results approximately in a temperature influence of approx. 0.07% / ° C for this synthesis and this temperature range. This unexpectedly large influence is probably not only due to temperature-dependent changes in the spectral intensity distribution, but possibly also to inhomogeneities in the reaction mixture during the heating process.

The calibration function (synthesis of product no. 1 / FN 5995112) comprises a data set with values from 0.18 to 4.41 NCO% and was during the synthesis with the FN 5995110 to the measuring range from 1.40 to 5.65 NCO -% applied. As can be seen in FIG. 12, the greatest deviation of the calculated value from the measurement-analytically determined value can be determined in the upper NCO range, since this value lies outside the calibration range (deviation = 0.15 NCO%). The deviations of the calculated values in the calibration range are all less than 0.1 NCO% (max. Deviation = 0.0996%; min. Deviation = 0.0033 NCO%).

In the following it should be checked whether the NCO contents in samples differ Let polyurethanes be calculated with a common calibration function, d. H. whether too Reaction mixtures of polyurethanes that do not derive from the same starting materials, Have cross correlations. For this purpose, two calibration data sets created; one of which is based on Polyurethane No. 1 and No. 2. This Polyurethanes have different band intensities with the same NCO contents. The second calibration data set is based on the polyurethanes No. 1, No. 2 and No. 4. The Spectra of these products showed approximately the same band intensities with the same NCO Held. The calibration data records of the Single calibrations (without outliers) are used. Furthermore, the entire Frequency range and the 1st derivative used as data pretreatment.

When products No. 1 (fill number 5995110, 5995112, 5995184) and No. 2 (fill number 5995173) were calibrated together, a correlation coefficient R ² of 96 was achieved. The reason for this correlation coefficient lies in the already mentioned differences in the intensities of the NH absorption band of the two polyurethanes at 6750 cm ^-1 . At equal concentrations in the solution, the intensity of the NH band of product No. 2 is about a quarter lower than that of product No. 1. Accordingly, the individual values of product No. 2 tend to be calculated too high. The standard deviation is given as 0.25% absolute. When considering the calculated individual values of product No. 2, a greater deviation from the analytical value is noticeable than for product No. 1. Thus, for product No. 2 with an NCO content of 0.1%, a value of 0, 5% calculated.

The joint calibration of products No. 1, No. 3 and No. 4 resulted in a better correlation coefficient R ² of 99.18. Although this is lower than for individual calibrations, it is within a range that is sufficient for practical use (standard deviation = 0.15 NCO%).

Claims (11)

1. A method for controlling polymerization reactions, in which a determination of chemically functional groups takes place via a measurement of infrared absorption bands, characterized in that the measurement is carried out in the NIR range and essentially first and second harmonics of OH, NH, COOH, NCO and / or epoxy groups recorded that the measured values are evaluated by comparison with calibration data sets, and that the measuring probe is arranged directly in the reactor and is connected to a spectrometer via glass fiber cables. 2. The method according to claim 1, characterized in that NIR light from a range of 4000 to 11000 cm ^-1 is used. 3. The method according to claim 1 or 2, characterized in that the Analysis of the spectra is carried out by a Fourier transform spectrometer. 4. The method according to any one of claims 1 to 3, characterized in that the polymerization is a polyaddition or polycondensation acts. 5. The method according to claim 4, characterized in that the polymer is a polyurethane and NCO and / or NH and / or urethane groups measured become. 6. The method according to claim 4, characterized in that the polymer is a polyepoxide and measured epoxy and / or OH and / or NH groups become. 7. The method according to claim 4, characterized in that the Polycondensate is a polyester, a polyamide, a polyether acrylate and / or Is polyester acrylate.   8. The method according to any one of claims 1 to 7, characterized in that the absorption bands with stored absorption bands earlier Reactions are compared. 9. Device for carrying out a method according to one of claims 1 to 8, containing a transmission infrared probe, which in one Polymerization reactor can be arranged, as well as a Fourier Transformation spectrometer, which is connected to the probe via fiber optic cable is connected and contains an interferometer or a diode array. 10. The device according to claim 9, characterized in that the transmission Infrared probe in a protective jacket made of metal, especially stainless steel, is arranged. 11. The device according to claim 10, characterized in that the protective jacket one on the sapphire or quartz window of the transmission infrared probe has directed nozzle through which hot or warm solvent on the Window can be sprayed.