Water analysis from a moving bed separation process simulated by near-infrared spectroscopy

Abstract

A simulated moving bed separation (LMS) method and device suitable for determining the water content of a stream of compounds (e.g. comprising xylenes) flowing in an LMS unit (1) by: recording near infrared (NIR) spectra of a set of samples, the samples being chosen such that the water contents of the samples extend at least partly over a range of water contents of the stream of compounds; measuring the water contents of the set of samples by a reference measurement method; developing a correlative multivariate regression model between the NIR spectra of the set of samples and the water contents of the set of samples; recording an NIR spectrum of the stream of compounds; and applying the correlative multivariate regression model to the NIR spectrum of the stream of compounds to determine the water content of the stream of compounds. Figure 3 to be published

Description

Title of the invention: Analysis of water from a moving bed separation process simulated by near infrared spectroscopy Technical field

[0001] The present invention relates to the field of methods for measuring contents and devices for controlling and regulating simulated moving bed separation (LMS) units, for example for the separation of aromatic compounds. More specifically, the present invention relates to the measurement of the water content of the flow of compounds (e.g. hydrocarbons) circulating in the various separation zones of an LMS unit. A particularly interesting application of the method according to the present invention is the separation of various xylenes, the flows circulating in the unit comprising isomers of aromatic C8 hydrocarbons, a desorbent and water in variable content depending on the measurement point in the LMS unit. The invention also relates to the control and regulation of the LMS unit, as a function of a difference between the measured value(s) of the water contents, and one or more setpoint values. Prior art

[0002] With reference to [Fig.l], an LMS unit 1 typically comprises a multi-stage column comprising a plurality of adsorbent beds A1-A12 arranged in series in a flow direction of the fluid(s) used in the column. The fluid successively passing through the adsorbent beds is called the main fluid to distinguish it from other secondary fluids which may be added to the main fluid via distribution and collection devices, also called trays P1-P12, such a tray generally being located between two successive beds (arranged perpendicular to the flow direction of the main fluid). The LMS unit further comprises a recycling circuit 2, typically provided with a recycling pump 3, for recycling the main fluid leaving the multi-stage column.

[0003] A tray comprises at least one collection zone and a distribution and collection network (system of lines and valves) for collecting the main fluid and / or injecting secondary fluids and mixing these secondary fluids with the main fluid. A tray also comprises at least one distribution zone which has the function of distributing the fluid resulting from the mixing of the main fluid and the secondary fluids over the granular bed located immediately downstream, in the direction of flow of the main fluid.

[0004] An LMS separation process typically comprises the following steps: at least one multi-stage column is fed with at least one feed F and one desorbent D, and at least one extract E and at least one raffinate R are withdrawn from the column. multi-stage, the feed and withdrawal points in the trays of the multi-stage column being offset over time by a value corresponding to an adsorbent bed with a permutation period, and determining a plurality of separation zones of the column, and in particular the following main zones: by definition, each of the operating zones is designated by a number: - zone 1 for desorption of a product to be separated (eg paraxylene) is between the injection of the desorbent and the withdrawal of the extract; - zone 2 of desorption of impurities (eg isomers of the product to be separated) is included between the withdrawal of the extract and the injection of the charge; - adsorption zone 3 of the product to be separated is between the injection of the feedstock and the withdrawal of the raffinate; and - zone 4 is between the raffinate withdrawal and the desorbent injection.

[0005] In this example of [Fig.l], the 12 adsorbent beds are distributed in zones 1 to 4 according to configurations known as type 2 / 5 / 3 / 2, that is to say that the distribution of the beds is as follows: 2 is the number of beds in zone 1; 5 is the number of beds in zone 2; 3 is the number of beds in zone 3; and 2 is the number of beds in zone 4.

[0006] LMS separation processes use zeolite adsorption techniques that allow the replacement of difficult or even impossible separations by distillation. This type of separation requires obtaining an optimum compromise between the separation capacity of the zeolite and the capacity to transfer the products within this zeolite. Indeed, the separation of the products will take place at the nanometric level in the zeolite, but to obtain a continuous and efficient process, the transfer within the bead containing this zeolite must take place in the macro / meso and microporous networks. Of course, these properties depend on the zeolite and the bead itself, also on parameters such as the composition of the mixture to be treated, the operating temperature.

[0007] An important parameter for improving separation remains the control of water content. Indeed, water will play a role of "lubricant" allowing a good transfer of the species to be separated within the zeolite. Furthermore, an excess of water will "clog" the zeolite and thus reduce productivity, that is to say the quantity of separable products per quantity of zeolite. It is therefore necessary, depending on the system studied, to be able to control the water content.

[0008] The Karl Fisher method is the reference method for measuring the water content in flows circulating in LMS units. In particular, the Karl Fisher method is a method for chemically measuring water content by titration. The principle of the Karl Fisher method is based on the oxidation of sulfur dioxide by iodine in the presence of water, in a methanol hydroxide solution following the reactions Math 1 in which imidazole (RN) shifts the equilibrium of equation (1) to the right, resulting in a maximum reaction rate with a stable equilibrium point. Equilibrium is reached when the color changes from yellow to brown (iodine reduction). The titration can be carried out volumetrically or coulometrically, the latter being more suitable for measuring fluxes with low water content. The maximum tolerance deviation of the Karl Fisher method is estimated at plus or minus 10 ppm by weight for a target value at 105 ppm by weight of water of the standard.

[0009] [Math.l] CH3OH + SO2 + RN [RNH]SO3CH3 (1) H2O + b + [RNH]SO3CH3 + 2 RN > [RNH]SO4CH3 + 2 [RNH]t (2)

[0010] Although being the reference method, the Karl Fischer method shows several drawbacks in its use for monitoring LMS processes. Sampling analysis is dependent on the operator performing the analysis. The change of operator affects reproducibility, for example depending on the precision of the weighings and the speed of execution. Sampling can create disturbances in the process during sampling, and sampling is sensitive to weather conditions. Depending on the flow rate of the fluid to be withdrawn, the sample collection time is more or less long and the sample may no longer be representative. Sampling analysis also does not allow real-time analysis because the analysis frequency does not allow advanced control of the unit, which makes the process control unresponsive. The determination of the water content of the flow(s) circulating in the LMS units can thus be improved. Summary of the invention

[0011] In the context previously described, a first object of the present description is to overcome the problems of the prior art and to provide an LMS process and an LMS unit allowing better yield and better productivity without impacting the purity of the separated products.

[0012] According to a first aspect, the aforementioned objects, as well as other advantages, are obtained by a simulated moving bed separation method (for example for the separation of xylenes and in particular paraxylene) comprising steps of measuring or determining the water content of a flow of compounds circulating in a simulated moving bed separation unit, the measuring steps comprising: (a) recording near-infrared spectra of a set of samples, the samples being chosen so that the water contents of the samples extend at least partly over a range of water contents of the compound stream; (b) measurement of the water contents of the sample set by a reference measurement method; (c) the development of a correlative multivariate regression model between the near-infrared spectra of the sample set and the water contents of the sample set; d) recording a near-infrared spectrum of the compound flow; e) applying the correlative multivariate regression model of step c) on the near infrared spectrum of the compound stream to establish / determine the water content of the compound stream.

[0013] According to one or more embodiments, the method comprises a step f) of modifying the water content in the flow of compounds in response to a difference exceeding a threshold value between the water content of the flow of compounds and a set water content in the flow of compounds.

[0014] According to one or more embodiments, the recording of a near infrared spectrum of the flow of compounds is carried out by an online spectroscopic analysis apparatus and / or in which the measurement of the water contents of the set of samples by a Karl Fisher method.

[0015] According to one or more embodiments, developing the correlative multivariate regression model comprises statistical analysis of the data.

[0016] According to one or more embodiments, the correlative multivariate regression model is developed by applying to the near infrared spectra of the sample set one or more analytical methods chosen from partial least squares regression, principal component regression, multiple linear regression, neural networks, direct classical least squares regression, indirect classical least squares regression, inverse least squares regression.

[0017] According to one or more embodiments, the correlative multivariate regression model is developed by applying to the near infrared spectra of the sample set one or more analytical methods chosen from partial least squares regression, multiple linear regression and artificial neural network.

[0018] According to one or more embodiments, the correlative multivariate regression model is developed by applying the partial least squares regression method to the near infrared spectra of the sample set.

[0019] According to one or more embodiments, the recording step(s) a) and / or d) and / or the measuring step b) is or are carried out under temperature and pressure conditions representative of the flow of compounds considered.

[0020] According to one or more embodiments, the method comprises a step g) of measuring the water contents of the flow of compounds considered by the measurement method of reference.

[0021] According to one or more embodiments, the recording of near infrared spectra of the set of samples and / or of the flow of compounds considered is carried out at at least one recording point chosen from: - an area between the injection of a desorbent and the withdrawal of an extract; - an area between the extraction of the extract and the injection of a charge; - an area between the injection of the charge and the withdrawal of a raffinate; - an area between the withdrawal of the raffinate and the injection of the desorbent; - a main fluid recycling circuit; - a recycling pump; - a charging circuit; - a raffinate circuit; and - an extract circuit.

[0022] According to one or more embodiments, in which the recording of near infrared spectra of the set of samples and / or of the flow of compounds considered, is carried out at least on the circuit of the extract or the raffinate.

[0023] According to one or more embodiments, the sample set comprises at least 10 samples, preferably at least 20 samples, very preferably at least 50 samples.

[0024] According to one or more embodiments, the sample set comprises at least in part samples prepared in the laboratory or at least in part samples originating from at least one sampling point of the simulated moving bed separation unit.

[0025] According to one or more embodiments, the samples have a water content of between 10 ppm by weight and 200 ppm by weight, preferably between 30 ppm by weight and 150 ppm by weight, very preferably between 40 ppm by weight and 130 ppm by weight.

[0026] According to a second aspect, the aforementioned objects, as well as other advantages, are obtained by a simulated moving bed separation unit (for example for the separation of xylenes and in particular paraxylene) comprising a device for measuring or determining the water content of a flow of compounds circulating in the simulated moving bed separation unit, the measuring device comprising: - at least one immersion probe or flow cell placed at a recording point of a flow of compounds considered from the simulated moving bed separation unit; - a near infrared spectrometry analysis system adapted to generate a near infrared spectrum of the compound flow at the recording point; - an operating system adapted to develop a correlative multivariate regression model between near infrared spectra of a set of samples and water contents of the sample set, and apply the correlative multivariate regression model on the near-infrared spectrum of the compound stream to establish the water content of the compound stream; - a device for controlling and regulating the water content in the flow of compounds in response to a deviation exceeding a threshold value between the water content of the flow of compounds and a set water content; and - optionally, a sampling system comprising sampling equipment adjacent to the immersion probe or flow cell; - optionally, a reference measuring device for analyzing samples; - optionally, an operating conditions sensor adjacent to the immersion probe or the flow cell.

[0027] Embodiments of the method and the unit according to the aforementioned aspects, as well as other characteristics and advantages, will appear on reading the description which follows, given solely for illustrative and non-limiting purposes, and with reference to the following drawing. List of figures

[0028] [Fig.l] shows a schematic view of an LMS unit in which a main fluid circulates and in which a feed and a desorbent are supplied and an extract and a raffinate are withdrawn.

[0029] [Fig.2] shows a schematic view of a device according to the invention for controlling and regulating an LMS unit, the device comprising in particular an immersion probe, a near infrared spectrometry analysis system, an operating system and a device for controlling and regulating the water content.

[0030] [Fig. 3] shows the correlation line between the water content determined by the method according to the invention and the value obtained by the Karl Fisher method. Description of the embodiments

[0031] Embodiments of the invention will now be described in detail. In the following detailed description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced without these specific details. In other instances, well-known features have not been described in detail to avoid unnecessarily complicating the description.

[0032] In the present application, the term "comprise" is synonymous with (means the same as) "include" and "contain", and is inclusive or open and does not exclude other elements not recited. It is understood that the term "comprise" includes the exclusive and closed term "consist". In the present description, the terms "essentially" or "substantially" correspond to an approximation of ± 5%, preferably preferably ±1%, very preferably ±0.5% (eg ±0.1%). For example, an effluent essentially comprising or consisting of compounds A corresponds to an effluent comprising at least 95% by weight of compounds A. Furthermore, in the present description, the term "content" corresponds to the concentration by weight of the species considered.

[0033] In particular, the present invention relates to an LMS method comprising steps of measuring or determining the water content of the flow of compounds (e.g. hydrocarbons) (main fluid or secondary fluids) circulating in the various separation zones of the LMS unit. The measurement of the water content is obtained from the spectrum obtained by near infrared spectroscopy (NIR) of the flow of compounds considered, and from a method of processing said spectrum. In particular, the measurement steps according to the present invention comprise: (a) recording NIR spectra of a set of samples, the samples being chosen so that the water contents of the samples extend at least partly over a range of water contents of a stream of compounds considered circulating in the LMS unit; b) measurement of the water contents of the sample set by a reference measurement method (e.g. Karl Fisher method); c) the development of a correlative multivariate regression model between the NIR spectra of the sample set and the water contents of the sample set (eg by statistical analysis of the data); d) recording a NIR spectrum of the flow of compounds considered circulating in the LMS unit; e) applying the correlative multivariate regression model of step c) to the NIR spectrum of the compound stream considered to establish the water content of the compound stream considered; (f) the modification of the water content in the compound stream in the event that a deviation between the water content of the compound stream and a set water content in the compound stream exceeds a threshold value; and g) optionally, the measurement of the water contents of the flow of compounds considered by the reference measurement method (optional verification step). Steps a), b) and c) are so-called model calibration steps; steps d), e) and f) are routine steps; step g) is a quality step (e.g. one-off) for model validation.

[0034] Advantageously, NIR spectroscopy makes it possible to study the phenomena resulting from the interaction between matter and light, which in the case of NIR spectroscopy concerns the range between 0.8 pm and 2.5 pm of the electromagnetic spectrum. The resulting absorption spectrum exhibits harmonic and com vibrational bands. combination. Chemical groups exhibiting absorption in the near infrared region are mainly of the form XH where X corresponds to carbon, nitrogen or oxygen atoms, and H is the hydrogen atom. NIR spectroscopy thus makes it possible to detect the presence of water in a sample.

[0035] Furthermore, NIR spectroscopy is a rapid and non-destructive analysis method which allows for online implementation on an LMS unit, due to the possibility of using an analysis system (e.g. comprising a flow cell or immersion probe) which can be fibered (i.e. online analysis; remotely controllable). Thus, the present invention makes it possible to obtain the water content in the compound flows of the LMS unit continuously, unlike a one-off measurement requiring laboratory processing. NIR spectroscopy allows for monitoring variations over time and also allows for the control of the water injection flow rate.

[0036] According to one or more embodiments, the recording step(s) a) and / or d) and / or the measuring step(s) b) and / or g) are carried out under temperature and pressure conditions representative of the flow of compounds considered. Advantageously, the accuracy of the development steps c) and application steps e) of the correlative multivariate regression model is thus improved.

[0037] According to one or more embodiments, the sample set comprises at least 10 samples, preferably at least 20 samples, very preferably at least 50 samples. According to one or more embodiments, the sample set comprises at least in part samples prepared in the laboratory. According to one or more embodiments, the sample set comprises at least in part samples from at least one sampling point of the LMS unit. According to one or more embodiments, the samples have a varied water content, for example between 10 ppm by weight and 200 ppm by weight, preferably between 30 ppm by weight and 150 ppm by weight, very preferably between 40 ppm by weight and 130 ppm by weight. The samples are analyzed by a NIR spectroscopic analysis technique, preferably online, and analyzed in a laboratory with the reference measurement method.

[0038] Any NIR spectroscopic technique is acceptable for acquiring the spectra of the samples. For example, a NIR spectroscopic apparatus suitable for the method according to the present invention is a so-called "process" spectrometer for which no sample preparation is necessary. For online recording, an immersion transflection probe or a flow cell can for example be used on the sampling loops for the acquisitions. The acquisition can be carried out on various optical paths, such as 5 mm and 10 mm paths, over a spectral range of 800 nm to 2200 nm.

[0039] According to one or more embodiments, the reference measurement method of the water content is the Karl Fisher method.

[0040] The correlative multivariate regression model makes it possible to correlate the NIR spectra of the sample set with a water content of the sample set, and the application of the correlative multivariate regression model on a NIR spectrum of the stream of compounds considered makes it possible to establish the water content of the stream of compounds considered. Multivariate regression is the application of at least one mathematical equation to the NIR spectra of the samples to output, for each sample and the stream of compounds considered, a value representative of the water concentration. In particular, the correlation between the NIR spectra of the sample set and the reference water contents (obtained for example by Karl Fischer) is obtained from statistical analyses. Preferably, the spectra are pre-processed beforehand, in order to avoid parasitic random variability.The most common pre-processing methods are, for example, centering by the mean of the data, normalization, and derivation. The data thus pre-processed can then be exploited by one or more multivariate methods which make it possible to reduce the dimensionality of the data and to extract the relevant information correlatable to the physicochemical properties of interest.

[0041] According to one or more embodiments, the correlative multivariate regression model is developed by applying to the NIR spectra of the sample set one or more analytical methods chosen from partial least squares regression (PLS), principal component regression (PCR), multiple linear regression (MLR), neural networks (ANN, CNN), direct classical least squares regression (DCLS), indirect classical least squares regression (ICLS), inverse least squares regression (ILS). This list is not exhaustive. According to one or more embodiments, the analytical method or methods are chosen from partial least squares regression (PLS), multiple linear regression (MLR) and artificial neural network (ANN). According to one or more embodiments, the method is partial least squares regression (PLS).

[0042] The correlative model is applied to the NIR spectrum of the compound flow considered to determine the water content of the compound flow considered. Once applied, the value given at the output of the model corresponds to the property of interest, i.e. the water concentration of the compound flow considered.

[0043] According to one or more embodiments, the recording of NIR spectra of a set of samples and / or of the flow of compounds considered circulating in the LMS unit is carried out online. Preferably, the method according to the invention comprises a method for continuous measurement / determination of the water content present at one (or more) point(s) considered in the LMS unit by means of an immersion probe, in which - a signal is sent to the immersion probe, - a PIR signal is recovered from the immersion probe to a PIR spectrometer, - the PIR spectrum of the PIR signal considered is recovered at the output of the PIR spectrometer, - the correlative multivariate regression model is applied to the PIR spectrum considered to establish the water content of the flow of compounds at the point(s) considered in the LMS unit.

[0044] With reference to [Fig.l], according to one or more embodiments, the recording of NIR spectra of the sample set and / or the flow of compounds considered circulating in the LMS unit, is carried out at at least one measurement point (eg 2, 3 or 4 measurement points), such as in zone 1, 2, 3, and / or 4 of the multi-stage column, the recycling circuit 2 (eg recycling pump 3), the charging circuit (eg charging pump), the raffinate or extract circuit (eg distillation column). Advantageously, the online recording of the NIR spectra makes it possible to obtain data under the temperature and pressure conditions representative of the flow of compounds considered.

[0045] According to one or more embodiments, the LMS unit is fed with a feed containing C8 aromatic hydrocarbons and producing a raffinate and an extract. According to one or more embodiments, the LMS unit is a hybrid unit implementing an LMS step and a crystallization step.

[0046] According to one or more embodiments, for example when the LMS unit is a unit for separating xylenes into LMS, when there is a single measuring point on the unit, this is preferably located on the extract or raffinate circuit. When there are two measuring points on the unit, the first is preferably located on the extract circuit, and the second is preferably located on the raffinate circuit. According to one or more embodiments, when there are two measuring points on the unit, the first is preferably located on the extract or raffinate circuit, and the second is preferably located on the feed or desorbent circuit, or on the recycling circuit (e.g. in the vicinity of the recycling pump). Optionally, a measuring point may be arranged in the rectification zone of the raffinate distillation column.Furthermore, when a crystallization step is implemented, a measuring point can be located on the liquid flow leaving the crystallization unit, i.e. at the level of the production line of the mother solution depleted in the product of interest to be separated (e.g. paraxylene).

[0047] Advantageously, the method according to the invention can allow the control and regulation of the LMS unit. For example, when the difference between the water content of the flow of compounds determined by said method and a profile of the set water content present in the flow of compounds is greater than a threshold value (e.g. 20 ppm by weight, preferably 15 ppm by weight, very preferably 10 ppm by weight), a control and regulation step is applied to one or more operating variables of the process, for example the flow rate of a valve, to modify the water content of the flow of compounds considered. According to one or more embodiments, a control and regulation step is applied to at least one operating condition chosen from the group consisting of: a main fluid flow rate, a feed flow rate, a desorbent flow rate, an extract flow rate, a raffinate flow rate, a water injection flow rate (e.g. in the feed or the desorbent), a permutation time, a temperature in the LMS unit.

[0048] With reference to [Fig.2], the invention also relates to a simulated moving bed separation unit comprising a measuring device for determining the water content of a stream of compounds (preferably comprising xylenes), the stream of compounds circulating in said simulated moving bed separation unit, said measuring device comprising: I) at least one immersion probe (or flow cell) 4 placed at a recording point of a flow of compounds considered 5 of the LMS unit; II) a NIR 6 spectrometry analysis system; optionally, a sampling system 7 comprising sampling equipment adjacent to the immersion probe or the flow cell; III) optionally, a reference measuring device 8 for analyzing samples IV) an operating system 9; V) a device for controlling and regulating 10 the water content in the flow of compounds considered; and VI) optionally, an operating conditions sensor 11 adjacent to the immersion probe or the circulation cell 4.

[0049] With reference to [Fig. 2], the immersion probe 4 is immersed in a PIR spectrum recording point of the LMS unit 1, a point for which it is desired to know the water concentration in a stream of compounds considered 5. In response to a control signal, a PIR signal emitted at the recording point is collected by the immersion probe 4 and is transmitted to the PIR spectrometry analysis system 6 which generates and records a PIR spectrum corresponding to the recording point. The PIR spectrum is sent to an operating system 9 (e.g. a computer) for analysis.

[0050] Optionally, in an area close to the recording point, an operating conditions sensor 11 (e.g. temperature and / or pressure) is immersed in the LMS unit 1 making it possible to send the operating conditions of the recording point to the operating system 9.

[0051] The operating system 9 analyzes the NIR spectrum and optionally the operating conditions, and determines the water content of the stream of compounds considered 5 at the recording point by applying the correlative multivariate regression model on the NIR spectrum.

[0052] By comparing the water content of the flow of compounds considered thus obtained with one or more set water content values, the control and regulation device 10 acts on one or more operating variables of the process, for example the flow rate of a valve, to modify the water content of the flow of compounds considered.

[0053] Optionally, in an area close to the recording point, the sampling system 7 is arranged to take one or more samples which are analyzed by the reference measuring device 8. Advantageously, the sampling system 7 makes it possible to provide samples for the model calibration steps. Advantageously, the sampling system 7 makes it possible to provide samples for the optional verification step. Examples

[0054] A set of more than 50 samples is selected. These samples are mainly two of the four effluents from the xylene separation process. 15 samples are laboratory-prepared samples acquired on the bench to increase sample variability, all other data are from the acquisition of online NIR spectra during dedicated tests. The same spectrometer was used for the analyses. The spectra were acquired on a Metrohm Process XDS spectrometer.

[0055] The modeled property is the water content of the effluents from the xylene separation process. For the acquisitions carried out in situ, the water content varies over time at stable pressure and temperature conditions during the test, and for those carried out in the laboratory, the data are acquired at room temperature and atmospheric pressure. A sample was taken for each sample and analyzed immediately on the Karl Fischer 756 coulometer equipped with a Metrohm diaphragm generator electrode.

[0056] The establishment of the multivariate regression model between the NIR spectra of the sample set and the water contents of the sample set, and the preprocessing of the NIR spectra, were carried out as follows: normalization on the total area of ​​the spectrum, extended multiplicative signal correction (EMSC) and centering by the mean. The validation criterion of the chosen model is the cross validation error RMSECV (for Root Mean Square Error of Cross Validation according to the English terminology) (example of PLS ​​regression). The lower the value, the better the prediction of the property of interest. For the model built to predict the water content of the xylenes mixture, the RMSECV is 9.13, with 4 latent variables. Thus, the average error of the model in cross-validation compared to the error of the reference method is in less than 10 ppm. In this example, statistical criteria were used to validate the model. The regression coefficients R2, biases and the ratio RMSECV / RMSEC (where RMSEC corresponds to an average calibration error; the closer the values ​​are, the closer this ratio is to 1, which is the target) were taken into account in the validation of the model. The correlation line (solid line) obtained with respect to the experimental data (hatched line) is presented in [Fig.3] (abscissa: measurement; ordinate: prediction).

Claims

Claims

1. A method of simulated moving bed separation comprising determining the water content of a stream of compounds flowing through a simulated moving bed separation unit (1), said determining the water content comprising the following steps: - recording near-infrared spectra of a set of samples, the samples being chosen so that the water contents of the samples extend at least partly over a range of water contents of the stream of compounds; - measuring the water contents of the set of samples by a reference measurement method; - developing a correlative multivariate regression model between the near-infrared spectra of the set of samples and the water contents of the set of samples; - recording a near-infrared spectrum of the stream of compounds;- applying said multivariate regression correlative model to said near infrared spectrum of the compound stream to determine the water content of the compound stream;

2. A method according to claim 1, comprising a step of modifying the water content in said compound stream in response to a deviation between the water content of said compound stream and a set water content in the compound stream, said deviation exceeding a threshold value.

3. A method according to claim 1 or claim 2, wherein the recording of a near infrared spectrum of said stream of compounds is carried out by an on-line spectroscopic analysis apparatus and / or wherein the measurement of the water contents of the sample set is carried out by a Karl Fisher method.

4. A method according to any preceding claim, wherein developing the correlative multivariate regression model comprises statistical analysis of data.

5. A method according to any preceding claim, wherein the multivariate regression correlative model is developed by applying to the near-infrared spectra of the sample set one or more analytical methods selected from partial least squares regression, principal component regression, multiple linear regression, neural networks, regression direct classical least squares, indirect classical least squares regression, inverse least squares regression.

6. The method of claim 5, wherein the multivariate regression correlative model is developed by applying to the near infrared spectra of the sample set one or more analytical methods selected from partial least squares regression, multiple linear regression and artificial neural network.

7. Method according to any one of the preceding claims, in which the recording step(s) and / or the measuring step are carried out under temperature and pressure conditions representative of the flow of compounds.

8. A method according to any preceding claim, comprising a step of measuring the water content of the compound stream by the reference measurement method.

9. Method according to any one of the preceding claims, in which the simulated moving bed separation unit comprises a multi-stage column, a main fluid recycling circuit (2), a recycling pump (3), a feed circuit, a raffinate circuit and an extract circuit, and in which the recording of near-infrared spectra of the sample set and / or the flow of compounds is carried out at at least one recording point chosen from: - an area of ​​said multi-stage column between the injection of a desorbent (D) and the withdrawal of an extract (E); - an area of ​​said multi-stage column between the withdrawal of the extract (E) and the injection of a feed (F); - an area of ​​said multi-stage column between the injection of the feed (F) and the withdrawal of a raffinate (R); - a zone of said multi-stage column between the withdrawal of the raffinate (R) and the injection of the desorbent (D);- the main fluid recycling circuit (2); - the recycling pump (3); - the charging circuit; - the raffinate circuit; - the extract circuit.;

10. Method according to claim 9, in which the recording of near infrared spectra of the sample set and / or the flow of compounds considered is carried out at least on the circuit of the extract or the raffinate.

11. A method according to any preceding claim, wherein the sample set comprises at least 10 samples, preferably at least 20 samples, most preferably at least 50 samples.

12. A method according to any preceding claim, wherein the sample set comprises at least partly laboratory-prepared samples or at least partly samples from at least one sampling point of the simulated moving bed separation unit (1).

13. A method according to any one of the preceding claims, wherein the samples have a water content of between 10 ppm by weight and 200 ppm by weight, preferably between 30 ppm by weight and 150 ppm by weight, very preferably between 40 ppm by weight and 130 ppm by weight.

14. A method according to any preceding claim, for the separation of xylenes.

15. A simulated moving bed separation unit comprising a measuring device for determining the water content of a stream of compounds preferably comprising xylenes, the stream of compounds circulating in said simulated moving bed separation unit (1), said measuring device comprising: - at least one immersion probe or flow cell (4) placed at a recording point of said stream of compounds (5) of the simulated moving bed separation unit (1); - a near infrared spectrometry analysis system (6) adapted to generate and record a near infrared spectrum of said stream of compounds at the recording point;- an operating system (9) adapted to develop a correlative multivariate regression model between near infrared spectra of a set of samples and water contents of said set of samples, and apply the correlative multivariate regression model on the near infrared spectrum of said stream of compounds to determine the water content of said stream of compounds; - a device for controlling and regulating (10) the water content in said stream of compounds in response to a deviation between the water content of said stream of compounds and a set water content, said deviation exceeding a threshold value; - optionally, a sampling system (7) comprising sampling equipment adjacent to the immersion probe or the; circulation cell (4); - optionally, a reference measuring device (8) for analyzing samples; - optionally, an operating conditions sensor (11) adjacent to the immersion probe or the flow cell (4).

Images