The present invention relates to a method for monitoring and controlling the manufacturing process of polyhydroxy alcohols, or polyols. More specifically, the invention is directed to a process for producing polyols from formaldehyde and another aldehyde by aldolization and hydrogenation, and to the monitoring and control thereof by means of infrared spectroscopy. In this method, the production of polyols is controlled by analyzing the composition of one or several process streams with infrared spectroscopy, and by using the difference between the predicted concentration and the desired concentration of a substance to control the process. In addition, the invention relates to a method for producing polyols, the method being monitored and controlled with infrared spectroscopy.

Polyhydroxy alcohols such as neopentyl glycol are important intermediates in the production of synthetic resins, including acryl resins, polyester resins, polyuretane resins, alkyd resins, and polycarbonate resins. Polyols are also used in the production of plasticizers, synthetic lubricants, surface active agents and the like. As is known, the aldolization reaction between formaldehyde and another aldehyde may be carried out continuously in the presence of a catalyst. For instance neopentyl glycol is obtained by reacting formaldehyde with isobutyric aldehyde in the presence of an amine resulting in hydroxypivalic aldehyde as the main product that is further hydrogenated, giving the desired neopentyl glycol as the final product. This neopentyl glycol may be further purified for instance by distillation. The process for producing polyols generally comprises three steps, that is, the aldolization step, the hydrogenation step, and the separation step. Each process step may comprise several units, arranged for instance in series.

Traditionally in the process for producing polyols, the compositions of the starting materials, solvents, aldolization product, by-products, and the final product in different process streams are monitored by taking samples from different points of

the process several times a day. Such frequent sampling is necessary, since the operation of the aldolization reaction typically changes with time. For instance, if an ion exchange resin is used as a catalyst, the operation of the reactor is changed by the deactivation of the catalyst. Organic compounds in the sample are analyzed with gas chromatography (GC), or high pressure liquid chromatography (HPLC). Water determinations are carried out with Karl Fischer titration. All these analytical methods are time consuming, require several instruments, and occupy laboratory personnel. Said chromatographic methods are generally precise, but the instruments need regular maintenance and calibration to function properly. With respect to the process control, the most significant disadvantage is the delay between the determination step and the control operation due to sampling, as well as preparation and analysis of the sample. The minimum delay between the sampling and the obtained result is always at least about one hour. Since the operation of the aldolization reactors changes relatively rapidly during operation, it is difficult to keep the yields of the reactors at predetermined values by using said analysis methods. In conventional chromatographic methods, samples must be removed from a process stream, and thus analysis errors due to sampling and sample preparation are possible.

In addition, the sampling may be dangerous to the person taking the sample, since some of the substances appearing in the process are toxic.

The patent US 5 349 188 discloses a continuous analysis of hydrocarbon mixtures with a near infrared spectroscopic method in the production of fuel mixtures. The patent US 5 712 481 presents a method based on near infrared spectroscopy for the determination of aromatic hydrocarbons as well such as benzene, toluene, xylene and alkyl benzenes in applications of the petroleum industry such as in catalytic cracking, alkylation, and isomerization, as well as in the production of fuel mixtures. Near infrared spectroscopic methods for monitoring and controlling a process in the production of polyesters are disclosed in publications EP 0 839 848 and WO 96/10009. The publication WO 98/29787 discloses the monitoring and controlling of the process for producing halogenated butyl rubber, and a method for controlling the properties of the product by means of a continuous, real-time analyzer system comprising a near infrared spectrometer and a viscosimeter. The publication

EP 0 801 299 presents a method for controlling of a process wherein the absorption is measured with a near infrared spectroscopic method using a wavelength varying between 600 and 2600 nm. The signals of said absorptions or mathematical functions thereof are compared with the corresponding values of standard samples.

One of the problems relating to the use of methods of prior art based on near infrared spectroscopy (NIR) is that it is difficult to provide an adequate correlation between the spectrum and the propery being monitored. Moreover, it is difficult to control positive or negative synergisms between properties brought about by different components. In polymerization reactions, such as for HD-polyethylene, a multilinear regression model based on density may give a false factor or correlation, causing problems in the polymerization process based thereon. Methods based on infrared spectroscopy are generally reliable as far as pure hydrocarbons are concerned.

The above teaching suggests that there is an obvious need for a continuously operating and real-time method for controlling the production process of polyols, as well as a processs for producing polyols that is continuously controlled in real time with a method based on infrared spectroscopy.

The object of the present invention is to provide a method for controlling the production process of polyols, and a process for producing polyols.

The characteristic features of the method of the invention for controlling the production process of polyols, and a process of the invention for producing polyols are presented in the claims.

It was discovered that the above objcts may be attained and the disadvantages of the solutions of prior art may be avoided or substantially reduced by proceeding as follows. According to the invention, the process for producing polyols may be controlled in almost real time with a process monitoring and/or controlling system based on infrared spectroscopy. With the method of the invention, the compositions of the starting materials, solvents, intermediates, impurities, by-products, and the

final product may be monitored almost simultaneously at different process points of the process streams.

In the aldolization step of the process for producing polyols, formaldehyde and another aldehyde are reacted in the presence of a suitable catalyst, preferably an amine catalyst such as a basic anion exchange resin, to form an aldolization product.

The other aldehyde may be selected from the group consisting of ethanal, propanal. butanal, 2-methylpropanal (isobutyric aldehyde), 2-methylbutanal, 2-ethylpentanal, 2-ethylhexanal, 2-isopropylbutanal, 2-phenylpropanal, 2-cyclohexylpropanal, 2-phenylbutanal, 2,3-diphenylpropanal, cyclopentyl aldehyde, and cyclohexyl aldehyde. When an ion exchange resin is used as an aldolization catalyst, the characteristic feature of the process is that the catalyst is deactivated by the acids in the process stream. Depending on the process conditions, the aldolization catalyst is regenerated once in 2-10 days. The properties of the aldolization product may be influenced by adjusting the process conditions, such as feed rates of the starting material, temperatures of the reactors, flow rates and the timing of the regeneration of the catalyst. For the optimal control of the process, it is very useful to know the composition of the aldolization product in different reactors in real time. Based on this information, the process conditions may be adjusted in such a manner that the composition of the aldolization product will remain as desired. The process conditions must be continuously adjusted during the run, since the acitivity of the catalyst declines constantly until a state is attained wherein the catalyst should be regenerated to restore the original level of its activity. In the hydrogenation step, a catalyst suitable for the process is used. The hydrogenation is preferably carried out at an elevated temperature and pressure. In the separation step after the hydrogenation, the desired alcohol is separated from the reaction mixture with a suitable method, for instance by distillation. In the separation step, the solvents may be recycled back to the hydrogenation or aldolization step, if necessary.

According to a substantial aspect of the invention, absorbance measurements from process streams are carried out with an infrared spectrophotometer based on reflectance method or flow-through method or a combination thereof. To analyze the

process stream, an optical cell or another measuring head, which is placed directly in the process stream or to a separate analyzer line carrying a stream from the process stream. The optical cell or the measuring head is preferably connected to the spectrophotometer with fibre optical cables, or it may be connected directly to the infrared spectrophotometer. Several optical cells or measuring heads may be connected to a single spectrophotometer with fibre optical cables through a multiplexer, allowing the compositions of different process streams to be monitored alternately. Alternatively, several infrared spectrophotometers may be used, each of them monitoring one or several sample points. Figure 1 shows the principle of a process wherein several sample points are monitored with a single infrared spectrophotometer. This figure shows a block diagram of the principle of the process for the preparation of alcohols wherein the starting materials are fed to the aldolization step and the final product is recovered from a separation step. This process comprises several measuring cells for monitoring the composition of the process stream in different steps. The measuring cells are connected to a spectrophotometer with optical fibres through a multiplexer that alternately transmits a signal from each of the cells to the spectrophotometer. Alternatively, each cell has its own spectrophotometer. The spectrophotometer measures the absorbance spectrum of the process stream, and then, on the basis of this spectrum, a computer determines the composition of the process stream according to a previously programmed calibration model. The difference between the measured, and the desired compositions of the process stream is used for the control of the process manually, or this information is entered to a process computer controlling the temperatures, flow rates and the timing of the regeneration of the aldolization catalyst on the basis of this information.

The optical cell or measuring head is used for irradiation of the process stream with mid infrared or near infrared radiation having a wavelength between 800 and 5000 nm. Using this measuring head, the absorbance spectrum of at least one process stream is measured in this mid infrared and/or near infrared region of 800-5000 nm.

More specifically, in this method, the absorbance of the reaction mixture is measured several times during the process with one or more measurements by using a single

wavelength or several different wavelengths selected from this mid infrared and/or near infrared region. The wavelengths used are selected with statistical methods to obtain the most favorable correlation ratio between known and measured analysis results in calibration.

In the method of the invention, a suitable multivariant method is used to collect the necessary composition data from the spectrum, since the absorbances of different components overlap in mid and near infrared range. Especially the spectrum of water acting as a solvent partly overlaps the absorption of other components. Figure 2 shows the absorption spectra of different components in near infrared range with the equipment settings of Table 1. The absorption spectrum of Figure 2 shows that the spectra of the starting materials, products and solvents are very similar. The spectral information is analyzed with mathematical and computational methods to obtain the necessary composition data. The mathematical calibration model is determined with calibration samples covering the variation of all the components to be measured found in actual process conditions. Based on this model that correlates the absorbance data to the composition data, in these varying process conditions such as at different temperatures and compositions, precise predictions of the concentrations of substances may be computed. The difference between the predicted composition and the desired composition is then used for the control of the process. Infrared spectrophotometry may thus be used to analyze the process stream, and then, with a computer software, the spectrum of the process stream is compared to the background spectrum and the concentrations of the substances of interest in the process stream in question are calculated with a preprogrammed calibration model.

During the monitoring of the process, the measuring head of the infrared spectrophotometer may be connected to the spectrophotometer with optical fibres, or it may be connected directly to the spectrophotometer. The spectrophotometer is connected to the controlling computer carrying out the necessary computation of the absorbance spectra and calculation of the concentrations of the substances according to the calibration model. This computer may opionally be in direct contact with the system for controlling the process that adjusts the temperatures and flow rates of the

process, and controls the regeneration of the catalyst. Alternatively, the analysis data from the computer may also be used to control the process manually.

There are several different mathematical computation methods for modelling the correlation between the composition of the sample and its mid and/or near infrared spectrum. Examples of suitable computation methods include the multivariant methods MLR, PCA, PCR and PLS, and several computation methods based on neural networks. The computation is typically carried out with a computer software applying one of the said mathematical methods.

The correctness of the calibration model created may be determined with cross validation or a separate test group. In cross validation, one or several samples are omitted from the calibration group, the calibration model is created with the remaining samples and then, the correctness of the model is tested for the samples omitted from the calibration group.

According to the present invention, it was found that the analytical method based on infrared spectroscopy used for controlling the production of polyhydroxy alcohols is accompanied with significant advantages over methods of the prior art. The invention makes a continuous monitoring of a process possible in almost real time. the analysis delay being only about one minute. Moreover, it allows a process to be controlled continuously by using the difference between the composition measured with infrared spectrophotometer and the desired composition to control the process according to Figure 1. The method of the invention neither requires highly skilled personnel nor is it laborious in comparison with traditional analysis methods. The analysis is carried out directly from the process stream and thus, no sampling from the process is needed. Due to the low energy level of near infrared radiation compared for instance to LJV radiation, the analyzer neither affects nor harms the substances being analyzed. The physical properties of the sample have no influence on the analysis result, since the absorption spectrum is used for the analysis.

Accordingly, the process stream may be used as such for the analysis without any preparation. Further, the analysis will not change the process stream, nor produce

waste to be disposed. Because of the fact that no removal of samples from the process stream is necessary, there is no occupational safety risk during sampling, and no errors influencing the analysis result due to incorrect sampling or incorrect sample preparation are possible.

The method of the invention allows an overall economic control of the process for producing polyhydroxy alcohols, and an overall economic production of polyhydroxy alcohols with this process, maintaining the quality of the product constant despite of varying process conditions. By using an on-line control, the aldolization and hydrogenation steps are constantly under control and the process is not allowed to drift to an inacceptable state leading to an impure product that would necessitate several distillation steps to separate the by-products. Accordingly, thanks to this on-line control process, a simpler purification unit may be designed, and product obtained is purer and has a better quality. These facts also make the process economically favourable.

The method of the invention is useful in various steps of the production of polyhydroxy alcohols, as also shown by the following examples 1-4. The method of the invention is particularly preferable in the process for producing polyhydroxy alcohols comprising an aldolization step wherein an ion exchange resin is used as the catalyst. The regeneration cycle of the resin catalyst may then be optimized by monitoring the conversion level or the composition of the product. Because in the polyol process, and particularly in the production of neopentyl glycol, trimethylol propane, 2-butyl-2-ethyl-1,3-propanediol and pentaerythritol, an ion exchange resin as the catalyst is used, which is quickly deactivated by the acids and must thus be regenerated once in every 24-120 hours, continuous on-line analysis is essential because of process adjustment. In this way, the state of the catalyst and the need for its renegeration may be monitored continuously, and if necessary, the catalyst may be transferred quickly to regeneration. Said polyol process will never be in a steady state, thus requiring continuous control of the process conditions. The real-time control of the process gives intermediates and products with improved purity, and

thus no separate purification steps are needed. Accordingly, the overall economy of the process is improved.

The immediate advantages of the invention are the facts that in this analysis method, the analysis delay is shorter and the method of analysis is less laborious when compared with traditional methods of analysis. The shorter analysis delay allows a more precise process control and makes possible to connect the analysis method directly to the process control system. As an on-line control method, the versatile NIR method of the invention is by far superior over other methods. All analyses may be carried out simultaneously, quickly and reproducibly and all components of the process may be determined at the same time. The following examples illustrate the analytical precision of the method, the working time needed therefor, and the reproducibility thereof. The advantages of the method are also evident from these examples.

The invention is now illustrated in more detail with reference to the following examples without wishing to limit the scope thereof thereto.

Example 1 A calibration model based on near infrared spectroscopy was drawn up for the aldolization step of the process for producing neopentyl glycol (NPG). In this aldolization step, hydroxypivalic aldehyde is obtained by allowing formaldehyde to react with isobutyric aldehyde in the presence of an amine catalyst. In this process, formalin may be used as the formaldehyde source containing besides formaldehyde water and methanol as solvents.

As the infrared spectrophotometer, a model based on Fourier transformation was used, connected to a flow-through cell wherein near infrared radiation travels through the liquid to be analyzed. The cell was connected to the spectrophotometer with fibre optical cables. The apparatuses and operating settings are shown in Table 1.

Table 1 Spectrophotometer FT-IR/NIR Cell Flow-through cell Length of light path 5 mm Wavelength region 800-2500 nm Background spectrum Empty cell and air Calibration temperature 50-70 °C Number of spectra for a sample 20 The calibration model was drawn up for all main components found in the aldolization step of the NPG production, that is, the aldolization product hydroxypivalic aldehyde, the starting materials isobutyric aldehyde and formaldehyde, and the solvents methanol and water. For the calibration, 80 calibration samples were prepared, part of them by mixing pure substances found in the process, and part of them by taking as samples from bench scale aldolization test equipment. These samples were analyzed with gas chromatography and high pressure liquid chromatography to determine the compositions thereof. In the calibration, a PLSplus/IQ statistical software based on PLS multivariant method was used. The number of the main components included in the PLS model was selected with cross validation. The calibration samples were prepared to cover the whole concentration range found in the aldolization step for all the substances to be calibrated. There was no significant correlation between the concentrations of the substances to be calibrated. The correctness of the calibration model was determined with cross validation by using the selected amount of the main component. The calibration results for each of the substances calibrated are shown in Table 2 below.

The table shows the concentration range used in the calibration, the number of the main components in the calibration model, and the correlation (Q2) of the test group from the cross validation, illustrating the correctness of the calibration with samples not belonging to the calibration model.

Table 2 SubstanceConcentrationrange MaincomponentsQ2 wt-%ofthePLSmodel Isobutyricaldehyde0-50 70.997 Formaldehyde0-25 70.994 Methanol8-20 80.999 Water10-40 70.999 Hydroxypivalic0-65 60.996 aldehyde

The calibration results show that with the method of the invention, the near infrared spectrophotometer may be calibrated to monitor the composition of the aldolization product in all process conditions found. Accordingly, for instance in a system of reactors in series, the composition of the aldolization product in the starting material feed, in all reactors and in the final aldolization product may be monitored with the analysis method of the invention. The calibrated concentration range covers all concentration variations of all substances to be calibrated in this range.

Example 2 A calibration model based on near infrared spectroscopy was drawn up for the hygrogenation and separation steps of the process for producing neopentyl glycol.

In the hydrogenation step, neopentyl glycol is obtained by allowing hydroxypivalic aldehyde to react with hydrogen in the presence of a catalyst. Solvents used in the process are typically water and methanol. In the separation step, the neopentyl glycol obtained from the hydrogenation is purified and the solvents are removed.

The calibration of the hydrogenation and purification steps was carried out with the specifications shown in Table 1. The calibration was carried out in such a way that the calibration samples contained all substances found in the hydrogenation and separation steps. Further, there was no significant correlation between the

concentrations of the substances to be calibrated. Table 3 below shows the substances present in the calibration samples and the concentration ranges thereof in the calibration.

Table 3 Substance Concentrationrange wt-% Neopentylglycol 0-90 Isobutyricaldehyde 0-10 Formaldehyde 0-10 Methanol 10-60 Water 0-30 Hydroxypivalicaldehyde 0-65 Isobutanol 0-10 Hydroxypivalylhydroxypivalate 0-4

In the calibration, the concentration ranges used were so wide that in addition to distinct hydrogenation steps, the same calibration model is suitable for several substance streams of the purification step wherein the solvents are separated and recycled to the process and the impurities are removed from the product. The calibration was carried out for neopentyl glycol and for the solvents water and methanol found in hydrogenation and separation steps. The calibration results are presented in Table 4 showing the concentration range used in calibration, the number of the main components of the calibration model, and the correlation (Q2) of the test group from the cross validation, illustrating the functionality of the calibration with samples not belonging to the calibration model.

Table 4 SubstanceConcentrationrange MaincomponentsQ2 wt-%ofthePLSmodel Neopentyl glycol 0-90 6 0.995 Methanol10-60 50.991 Water0-30 5 0.994

The calibration results show that by using the method of the invention, the near infrared spectrophotometer may be calibrated to monitor in the hydrogenation step and in the purification step of the product, the purity of the final product (neopentyl glycol) and the concentrations of the solvents in different process conditions.

Accordingly, for instance in a system of reactors in series for hydrogenation, the hydrogenation conversion in all reactors may be monitored with the method of analysis according to the invention, since the total concentration range of the product, that is from 0 to 90 wt-%, may be covered with this calibration model. Due to the wide concentration range, the concentrations of the product and the solvent may also be monitored in several process streams of the purification step.

Example 3 A calibration model based on infrared spectroscopy was drawn up for the aldolization step of the process for producing 2-butyl-2-ethyl-1,3-propane diol (BEPD). In the aldolization step, the BEPD aldol is obtained by reacting formaldehyde with 2-ethylhexanal in the presence of an amine catalyst.

The calibration model was created for the aldolization product, BEPD aldol, and for the other starting material, the 2-ethylhexanal, and for the hydrogenation product, BEPD being formed in this process in minor amounts already in the aldolization step.

For the calibration, 50 calibration samples were prepared, part of them by mixing pure substances, and part of them by taking samples from a laboratory process. The

samples were analyzed with CG and HPLC methods to determine the compositions thereof. The apparatuses and operating conditions are shown in Table 5.

Table 5 Spectrophotometer FT-IR/NIR Cell Flow-through cell Length of light path 1 mm Wavelength range 1670-2700 nm Background spectrum Empty cell and air Calibration temperature 20 °C Number of spectra for a sample 20 In the calibration, a statistical PLS multivariant method was used. The number of the main components included in the PLS model was selected with cross validation. The calibration samples were prepared in a manner to cover the whole concentration range found in the aldolization step for all the substances to be calibrated and that there was no significant correlation between the concentrations of the substances to be calibrated. The correctness of the calibration model was determined with cross validation by using the selected amount of the main component. The calibration results for each of the substances calibrated are shown in Table 6 below. The table shows the concentration range used in the calibration, the number of the main components in the calibration model, and the correlation (Q2) of the test group from the cross validation, illustrating the functionality of the calibration with samples not belonging to the calibration model.

Table 6 SubstanceConcentrationrange MaincomponentsQ2 wt-%ofthePLSmodel 2-Ethylhexanal 0-50 7 0.998 BEPD aldol 8-20 80.990 BEPD10-40 7 0.996

The calibration results show that with the method of the invention, the near infrared spectrophotometer may be calibrated to monitor the composition of the aldolization product in the process for producing BEPD.

Example 4 A calibration model based on near infrared spectroscopy was drawn up for the process for producing trimethylol propane (TMP). In the aldolization step, n-butyric aldehyde and formaldehyde are reacted in the presence of an amine catalyst to form TMP aldol that may further be hydrogenated to TMP. In this process, formalin may be used as the formaldehyde source containing besides formaldehyde water and methanol as solvents.

The calibration was carried out by using the apparatuses and settings shown in table 1. In the calibration, a PLS multivariant method was used. The number of the main components of the PLS model was selected with cross validation. The calibration was carried out in such a manner that the calibration samples contained most of the substances pressent in the process, and that there was no significant correlation between the concentrations of the substances to be calibrated. The correctness of the calibration model was determined with cross validation by using the selected amount of the main component. The calibration results for each of the substances calibrated are shown in Table 7 below. The table shows the concentration range used in the calibration, the number of the main components in the calibration model, and the correlation (Q2) of the test group from the cross validation,

illustrating the functionality of the calibration with samples not belonging to the calibration model.

Table 7 Substance ConcentrationrangeMaincomponents Q wt-%ofthePLSmodel Fonmaldehyde 0-30 6 0.991 n-Butyricaldehyde 0-30 6 0.990 Trimethylolpropane 0-65 6 0.997 Methanol 0-30 5 0.988 Water 0-20 5 0.994 The calibration results show that with the method of the invention, the near infrared spectrophotometer may be calibrated to monitor the process for producing TMP.

Example 5 (comparative) The method of analysis according to the invention, based on infrared spectro- photometry, and the chromatographic (GC and HPLC) analysis methods used traditionally in the production of polyhydroxy alcohols were compared in the aldolization step of the bench scale process for producing neopentyl glycol. The method of the invention was used to analyze hydroxypivalic aldehyde, isobutyric aldehyde, formaldehyde and methanol in the aldolization mixture. The apparatuses and operation conditions, and the calibration model were the same as in Example 1.

The aldolization process equipment comprised three reactors in series. The catalyst used in the process was an anion exchange resin. A flow-through cell was connected to an infrared spectrophotometer with fibre optical cables. The cell was joined to the exits of the three reactors in such a manner that the flow could be directed to the cell alternately from any of the three reactors with the analyzer pump.

The method of the invention and the traditional gas chromatographic and high pressure chromatographic analysis methods were compared for five days by monitoring the compositions of the effluents of the reactors. With the near infrared spectroscopic method of the invention, the process stream was continuously monitored by alternately analysing each sampling point. With the traditional method, reference samples, altogether 45, were taken three times a day from the exit of each reactor, and the composition of the samples was determined with GC, HPLC and Karl Fisher methods. Table 8 shows the concentration ranges of the reference samples, and the mutual correlation (R 2) of the results obtained with near infrared spectroscopy and with the chromatographic methods.

Table 8 Comparison Substance Comparative analysis Concentration range R Isobutyricaldehyde GC 3. 3-17. 1 0.990 Formaldehyde HPLC 0.7-6.7 0.982 Methanol GC 11. 8-40. 3 0.995 Hydroxypivalicaldehyde HPLC 39.4-63.3 0.978 Water KarlFishertitration 10. 9-23. 1 0. 983 The comparative results presented in Table 8 show that the method according to the invention based on near infrared spectroscopy gives almost equal measuring results in a wide concentration range where compared to the results obtained with the chromatographic methods used traditionally. The correlation, R2, of the results obtained with these methods for different substances was between 0.978-0.995.

The results of the comparison for individual samples is shown in Figure 3.

During the comparison period, the work load needed to carry out the analyses, and the analysis delay between the sampling and the results were determined. In the method of the invention, the work load consists of the running of the monitoring samples and adjustments of the apparatuses. The work load and the analysis delay in

the chromatographic methods are caused by the sampling, sample preparation and the running of the chromatogram. The results shown in Table 9 are calculated for a single sample at a time. The sample delay increases significantly if samples from all the reactors are determined simultaneously with traditional methods.

Table 9 Comparison MethodWorkload,Sampledelay, min./sample min. Methodoftheinvention2 2 Traditionalchromatography(GCandHPLC)40 50 1 The comparison presented in Table 9 shows that in the infrared spectroscopic method of the invention, the sample delay is clearly shorter and the work load is considerably smaller than in the traditional chromatographic analysis methods based on sampling.

Example 6 (comparison) The infrared spectrophotometric analysis method of the invention and the chromatographic analysis methods (GC and HPLC) used traditionally for the monitoring of the production of polyhydroxy alcohols were compared for the monitoring of the production of hydroxypivalic aldehyde. The reproducibility of said methods in continuous process monitoring was compared by pumping hydroxypivalic aldehyde mixture through a transmission flow-through cell connected with fibre optical cables to an infrared spectrophotometer. A sampling point for taking samples for chromatography was situated immediately after the flow-through cell. In the comparison, the concentration of hydroxypivalic aldehyde was determined 25 times from a mixture containing about 45 wt-%, HPA, with a gas chromatographic method (GC), a high pressure chromatographic method (HPLC), and with the method of the invention (NIR). The standard deviation and maximum deviation of the analysis results is presented in Table 10.

These results show that with the method of the invention, the reproducibility of the measuring results is clearly better than with the traditional chromatographic methods and the poorer reproducibility of the chromatographic methods is for instance caused by the various working steps needed for the analysis. The reproducibility of the results is extremely important if the analysis method is used as an aid in process control, and thus, the method of the invention suits better for the continuous process control than the traditional chromatographic methods (GC and HPLC).

Table 10 Comparison Analysismethod Standard deviation Maximumdeviation NIR 0. 257 0. 93 GC 0. 569 1.53 HPLC 0. 631 2.12