METHOD FOR ANALYSING PROCESS STREAMS

Abstract

The invention relates to a method for investigating process streams comprising five or more different hydrocarbon-containing components. In the method at least one process flow line (35) is in operative connection with an online IR spectrometer (2) and an online gas chromatograph (1). The process stream passed through the process stream conduit (35) is subjected to an online characterization which comprises measurements both with the online IR spectrometer and with an online gas chromatograph. The spectral data and the chromatography data are mathematically related to one another by suitable statistical models, thus allowing training of a model used for evaluating the analytical data and for characterizing the process streams. The method according to the invention is characterized by short measurement times in the range of seconds and milliseconds and a high accuracy. The method according to the invention for investigating process streams preferably relates to investigation of process streams deriving from processes proceeding in parallel, the process streams preferably deriving from reaction spaces arranged in parallel.

Claims

1. A method for investigating at least one process stream comprising at least five different hydrocarbon-containing components, wherein the method comprises at least the steps of: a) providing at least one process stream conduit (35) which is in operative connection with at least one online IR spectrometer (2) and in operative connection with at least one online gas chromatograph (1); b) passing at least one process stream through the at least one process stream conduit (35), wherein during this passing of the process stream through the process stream conduit (35) an analytical characterization of the process stream using an online IR spectrometer (2) and an online gas chromatograph (1) is performed.

2. The method according to claim 1, further comprising the steps of: c) evaluating the spectral data obtained in the analytical characterization of the process stream using an online IR spectrometer (2) as a function of the time at which this spectroscopic analysis of the process stream was carried out, d) evaluating the chromatography data obtained during the analytical characterization of the process stream using the online gas chromatograph (1) as a function of a sampling time for samples taken from the process stream, e) machine learning-based training of a model that models a mathematical relationship between spectral data and corresponding chromatography data in respect of an identical process stream by using the evaluation results obtained in steps c) and d) in respect of the process stream passed through the process stream conduit in step b).

3. The method according to claim 2, wherein steps c)-e) are performed and at least one reaction parameter is altered in relation to the same reaction parameter as set in steps a) and/or b), wherein this parameter is preferably selected from: WHSV, temperature, total pressure and/or partial pressure of reactants.

4. The method for investigating at least one process stream according to claim 2, wherein it comprises at least two different phases, wherein one phase is a training phase comprising steps a) to e) and the second phase is an actual measurement phase in which an analytical characterization of a process stream passed through the at least one process stream conduit is carried out using the online IR spectrometer (2) on the basis of the model trained in the training phase.

5. The method for investigating at least one process stream according to claim 1, wherein the process stream passed through the process stream conduit (35) is a gaseous process stream, the temperature of the gaseous process stream preferably being in the range of 20-350 C., more preferably in the range of 50-220 C.

6. The method for investigating at least one process stream according to claim 1, wherein the online gas chromatograph (1) and the online IR spectrometer (2) are serially arranged in respect of the process stream conduit and the temporal offset is in the range of 1 sec to 180 sec.

7. The method for investigating at least one process stream according to claim 1, wherein the analysis units, the online gas chromatograph (1) and the online IR spectrometer (2), are coupled to a common process control unit (4), wherein the process control unit is in operative connection with a process space (11) and controls, governs or regulates the process proceeding in the process space.

8. The method for investigating at least one process stream according to claim 7, wherein the process control unit (4) regulates the process such that the product structure is controlled by adapting a process operating parameter, the product structure preferably being controlled in such a way that octane number is constant, that selectivity is constant or that conversion is constant, regulation more preferably being undertaken by altering parameters from the group of temperature, WHSV.

9. The method for investigating at least one process stream according to claim 1, wherein the process comprises at least one process from the group selected from methanol conversion processes such as MTO (methanol to olefins), dehydrogenation reactions such as propane dehydrogenation, coupling reactions such as methane coupling, naphtha reforming and processes for conversion of aromatics.

10. The method for investigating at least one process stream according to claim 1, wherein the online IR spectrometer (2) operates in the mid IR range (MIR) and the wavenumbers are in the range of 400 cm.sup.1-3500 cm.sup.1.

11. The method for investigating at least one process stream according to claim 2, wherein the model or the method as employed in step e) comprises a statistical method selected from the group of multivariate analyses such as principal component analysis (PCA), partial least squares (PLS) regression, principal component regression (PCR), multi-linear regression (MLR) analysis, discriminant analysis or neural networks.

12. The method for investigating at least one process stream according to claim 2, wherein additional data for training the model, provided for according to steps c) to e), are obtained by variation of process parameters such as temperature, pressure, partial pressure of the reactants or WHSV.

13. The method for investigating at least one process stream according to claim 1, wherein the mass of the catalyst employed for the process is in the range of 0.1-200 ccm, preferably between 0.2-20 ccm.

14. The use of the method according to claim 1 in high-throughput testing of at least four, preferably at least eight, more preferably at least 12, catalysts arranged in parallel reactors.

Description

EXAMPLES

A.1 Conversion of Methanol to Olefins

[0086] Investigations into the catalytic conversion of methanol to olefins were performed to illustrate the method according to the invention. The investigations were performed using a high-throughput apparatus for catalyst testing which was assemblable with up to 16 reactors arranged in parallel.

[0087] The schematic construction of the apparatus is shown in FIG. 9, wherein the apparatus shown in the figure comprises five reactors arranged in parallel (11)-(15). The output conduits (21)-(25) of the reactors are connected to a multiport valve (33) leading to a process stream conduit (35) which is in operative connection with an online IR spectrometer (2) and an online gas chromatograph (1). Both the online IR spectrometer (2) and the online gas chromatograph (1) are connected to the apparatus process control means. Testing employed two different zeolite-containing catalyst samples which are commercially available and are referred to as sample C1 and sample C2 in the context of the description. It is apparent from FIG. 9 that the process control unit (4), connected to the online IR (2) and the online GC (1), is in operative connection with each of the individual reaction spaces (11)-(15). The operative connection to the process control unit (4) for the reactant supply, which is preferably also connected to the process control unit (4), is not shown.

[0088] Sample C1 contained a catalyst based on SAPO-34 and sample C2 contained a catalyst based on ZSM-5. To prepare the catalytic investigations the individual catalyst samples were each mixed with quartz powder and introduced into tubular reactors in the form of dumped powder beds. The amount of catalyst samples employed was 0.92 g or 1.85 g. A total of four catalyst-containing sample mixtures were accordingly prepared. For comparative purposes two reactors were laden with quartz powder free from catalyst as inert material.

[0089] The employed reactors had a tube length of 30 cm and an internal diameter of 15 mm.

[0090] Employed as the online gas chromatograph (1) was a Hewlett Packard HP 5890 chromatograph equipped with a fused silica column (having a length of 20 m) and an FI detector. Employed as online IR spectrometer (2) was an FTIR spectrometer from BRUKER which was optimized for the MIR spectral range in a wavenumber range of 7000-400 cm.sup.1. The optics of the instrument were controlled with an HeNe laser. The beam path of the spectrometer was equipped with a heatable gas measuring cell having an internal volume of 500 mL and an optical beam length of 75 cm. During the investigations the measuring cell was heat treated at a temperature of 180 C.

[0091] When supplying the methanol-containing reactant stream a distinction was made between the following two metering procedures:

[0092] Metering procedure 1. Continuous (quasi-continuous) metering of the methanol-containing carrier gas stream over a duration in the range of 3-30 min and

[0093] Metering procedure 2. Pulsed metering of the methanol-containing carrier gas stream over a duration in the range of 5-60 seconds.

[0094] After termination of the reactant stream supply an oxygen-containing gas stream was supplied to the reaction space in order to burn off the coke formed and to regenerate the catalyst.

[0095] The method for converting the methanol-containing carrier gas stream was performed according to the method of the invention. Initially a training phase was performed. During this training phase the reactor spaces were consecutively subjected to three different reactant gas streams, each having a lower reactant content than the reactant gas stream intended for the target parameter space. The reactant content in the reactant gas stream was adjusted by varying the gas loading defined by the WHSV (weight hourly space velocities). The gas loadings chosen for the training phase were characterized by the following three values of the WHSV: 0.2 h.sup.1, 0.3 h.sup.1, 0.4 h.sup.1.

[0096] During the different metering procedures performed during the training phase, both gas chromatographic analyses using online gas chromatograph (1) and spectroscopic analyses using online spectrometer (2) were performed on the individual process streams successively discharged from the reaction spaces into the process gas conduit (35) via the multiport valve (33).

[0097] The individual analytical results obtained by means of the different methodsi.e. by spectroscopy and by gas chromatography as a function of time and experimental test parameterswere related to one another by mathematical models. For example the GC analyses were used quantitatively to determine the respective amount of individual substances in the process stream (for example the content of methanol, dimethyl ether, methane, ethane, ethene, . . . ) and the amount of substance groups in the process stream was also determined (for example the content of aromatics, olefins, . . . ).

[0098] In the present case a proprietary model based on the PLS (partial least-squares) method was used for quantification for 9 individual substances, wherein 15 components sufficiently elucidated the variance. The wavenumbers were reduced to the ranges 423-1040 cm.sup.1 and 1244-2704 cm.sup.1 since especially the range between 1040-1244 cm.sup.1 and 2700-3100 cm.sup.1 reached the absorption limit at the employed concentration range, thus preventing further data evaluation. A finer or more precise adjustment of the optical path length to the concentration range should also allow these ranges of wavenumber to be integrated into the evaluation model.

[0099] The training phase was followed by investigations in the context of the production phase which were carried out in the presence of a reactant gas stream having a higher gas loading compared to the training phase, namely under conditions in the target parameter space.

[0100] The investigations during the production phase/measurement phase were here characterized by a gas loading at which the WHSVs were in the range of 2-20 h.sup.1, wherein the analytical characterization of the process stream (or the process streams) was by means of an online IR spectrometer (2) (presently optimized for the MIR range). Measurement signals were recorded at time intervals of 5 seconds to characterize the process stream. The quantitative evaluation of the band regions using the model established during the training phase made it possible to draw conclusions about the concentration of individual substances using the spectra. Trained models accordingly made it possible to predict the concentration of individual substances using the IR spectra.

[0101] To check the data analytical characterizations of the process stream were furthermore performed simultaneously with the IR measurements at intervals of in each case 20 minutes using the gas chromatograph (1) during the measurement phase. When performing so-called pulse experiments or pulse metering the additional performance of gas chromatographic analyses was completely dispensed with. An overview of the tests performed is shown in Table 1. Three experiments were performed during the training phase and five experiments were performed during the measurement phase.

[0102] When converting methanol to olefins the gas loading during the training phase was chosen such that the WHSV values were 0.2 h.sup.1, 0.3 h.sup.1 and 0.4 h.sup.1. During the measurement phase the WHSV values were 5 h.sup.1, 8 h.sup.1, 12 h.sup.1, 16 h.sup.1 and 20 h.sup.1. The process streams were thus generated and then also characterized as a function of time during the training phase and are characteristic of the three different parameter spaces and selected process parameters. Whether these three different parameter spaces are sufficient depends, among other things, on the process being studied and the temporal behavior of the process being studied.

[0103] It is preferable in connection with the method according to the invention that a number of process parameters spaces which is preferably 2 is investigated during the training phase. The number of process parameter spaces employed for generation and characterization of process streams when performing the training phase is more preferably 3, the number of process parameter spaces yet more preferably being 4.

[0104] When performing the measurement phase the parameter space of the WHSV is preferably in the range of >1 h.sup.1 to 20 h.sup.1. By contrast, during the training phase the parameter space of the WHSV is preferably in the range of 0.1 h.sup.1 to 1 h.sup.1. During performance of the measurement phase the employed WHSV range may also be regarded as a target parameter space.

[0105] In example A.1 the WHSV is significantly outside the target parameter space during the training phase.

[0106] Having regard to the lower limit of the target parameter space (i.e. a WHSV of 5 h.sup.1) during the measurement phase the deviation of the process parameter WHSV during the training phase is only 8% compared to the measurement phase. This gives rise to the requirement that the process parameters, or at least one process parameter, are outside the target parameter space during the training phase. The term outside the target parameter space is to be understood as meaning thatwhen comparing the training phase and the measurement phasethe deviation of at least one process parameter is 10%, the deviation preferably being 20%, the deviation more preferably being 50%, the deviation especially preferably being 75%.

[0107] Having regard to A.1 the lower limit of the WHSV in the target parameter space was given by a WHSV of 5 h.sup.1. The WHSV during the training phase was 0.4 h.sup.1. The deviation was therefore 4.6 h.sup.1 which corresponds to a percentage deviation of 92%. Having regard to the upper limit of the target parameter the deviation is given by a WHSV difference of 19.6 h.sup.1 which corresponds to a percentage deviation of 98%.

[0108] Table 1 gives an overview of the test numbers and the associated process parameters.

TABLE-US-00001 Experiment WHSV Temp. Operating number [h.sup.1] [ C.] mode Phase C1_01 0.2 400 continuous training phase C1_02 0.3 400 continuous training phase C1_03 0.4 400 continuous training phase C1_04 5 400 continuous measurement phase C1_05 8 400 continuous measurement phase C1_06 12 400 continuous measurement phase C1_07 16 400 continuous measurement phase C1_08 20 400 continuous measurement phase

[0109] Table 2 shows the percentage of the variance covered by the model according to the components (factors).

TABLE-US-00002 A. 2 Method for upgrading of aromatics mixtures Components ethene propene butene methane ethane propane butane methanol DME 1 0.405 7.72 59.059 28.49 89.47 52.08 47.89 22.06 70.82 2 75.92 92.01 79.86 28.91 89.49 89.06 86.22 22.22 91.81 3 94.5 94.05 96.05 40.55 89.95 97.31 96.56 25.23 96.66 4 94.78 97.52 98.9 94.36 89.96 98.69 97.86 36.79 99.22 5 96 98.65 99.24 95.75 90.27 98.86 97.88 94.43 99.23 6 98.94 99.09 99.5 96.08 97.36 99.08 99.03 95.55 99.43 7 99.12 99.11 99.67 98.06 97.38 99.09 99.38 95.58 99.66 8 99.13 99.19 99.72 98.18 97.88 99.53 99.47 95.65 99.66 9 99.13 99.32 99.73 98.42 98.7 99.53 99.47 95.67 99.72 10 99.13 99.34 99.75 98.65 98.79 99.62 99.55 95.67 99.73 11 99.17 99.35 99.75 98.69 98.84 99.69 99.55 95.8 99.73 12 99.17 99.37 99.77 98.89 99.33 99.71 99.63 95.8 99.73 13 99.17 99.38 99.78 98.89 99.34 99.71 99.65 95.8 99.73 14 99.19 99.38 99.78 98.89 99.34 99.71 99.65 95.83 99.73 15 99.22 99.4 99.78 98.89 99.34 99.72 99.7 95.98 99.73

[0110] Catalytic investigations into the upgrading of aromatics mixtures were also performed to illustrate the method according to the invention. For comparability of catalysts it is in many cases of interest to operate them under process conditions selected such that the different catalysts and processes result in identical target parameters. In numerous processes this target parameter is conversion. In naphtha reforming the target parameter is octane number.

[0111] This target parameter is achieved for example by alteration of the process parameters such as temperature, pressure, WHSV etc. Especially when operating the method according to the invention in a parallel arrangement in which a multiplicity of catalysts is investigated simultaneously and in parallel the method according to the invention results in advantages. In a preferred embodiment the method also relates to adjustment of the target parameter to a fixed value or, in connection with an apparatus having a multiplicity of reactors arranged in parallel, the respective adjustment of the target parameters in the processes performed in parallel. In the cases where deactivation of the catalyst occurs the method also relates to readjusting the process to achieve and maintain the target parameter despite the occurrence of deactivation.

[0112] In connection with the method it is important that adjustment and readjustment is carried out faster than the change in the product spectrum and the deactivation of the catalysts over time. One advantage of the method according to the invention is also that performance of the online IR measurements during the measurement phase allows precise and rapid analysis and thus also ensures rapid readjustment.

[0113] In the present example the method according to the invention was illustrated in connection with a catalytic method for converting and upgrading an aromatics mixture. The employed aromatics mixture was a mixture of mononuclear aromatics having different numbers of alkyl substituents and positions of the alkyl substituents. The objective of the method was that of producing a product having a high proportion of p-xylene. The method was performed in such a way that the conversion of aromatics was selected as a target parameter. The investigations were performed using an apparatus for catalyst testing which was equipped with four reactors arranged in parallel. Each of the four reactors arranged in parallel was filled with another catalyst material referred to hereinbelow as catalyst K1-K4.

[0114] In a first method variant the analysis of the process stream was carried out exclusively with an online gas chromatograph (1), wherein the spectra were evaluated automatically and the conversion determined. The duration of the individual gas chromatographic analysis was approximately 30 minutes. In order to sequentially characterize the process streams of the four reactors arranged in parallel using the one online gas chromatograph (1) a duration of 2 hours was required to obtain one measured value per process stream.

[0115] The four reactors arranged in parallel comprising the catalysts K1 to K4 were initially started up isothermally at the same temperature to obtain initial information about the relationship between conversion and temperature. These measurements, carried out at identical temperature, were deepened by additional measurements at different temperatures to obtain a calibration function showing the relationship between conversion and temperature. The results of the investigation are shown in FIG. 7, wherein the lower portion shows temperature as a function of the TOS (TOS=Time On Stream) and the upper portion shows conversion as a function of the TOS. Adjustment to a target value for conversion taking account of this calibration function commences from a TOS of 6 hours, wherein the result from the gas chromatic analysis is automatically transferred to the process control means and the temperature adjusted. It is apparent that the same conversion is achieved after about 3 cycles, each of 2 h in length, and actual measurement of the catalysts may be carried out at identical conversion. In the present case the adjustment process required about 6 hours. Performing the method with an apparatus having a higher degree of parallelization and equipped, for example, with 16 reactors would correspondingly require 1630 min measuring time per cycle which for three cycles would result in a duration of 24 hours.

[0116] Investigations into performing the method in a second method variant are described hereinbelow. FIG. 8 shows the method according to the invention with inclusion of an online IR spectrometer (2). The initial phase, in which the reactors were operated isothermally, was used to record the calibration function. The online gas chromatograph (1) and the online IR spectrometer (2) were operated in parallel. GC chromatograms comprising the conversion information and the accompanying IR spectra were obtained. Accompanying IR spectra is to be understood as meaning that said spectra were recorded simultaneously with the GC chromatograms. The obtained measurement data contained in the GC chromatograms and IR spectra as a function of time were used to train a model by PLS. Subsequently, during adjustment to constant conversion the recordal of GC chromatograms was dispensed with. Investigation of the process streams was carried out based on recordal of online IR spectra while determining conversion using the previously established model. In this example, the measurement time was reduced from 30 minutes to 2 minutes and the target conversion was achieved within 24 minutes. In this example, the method according to the invention showed a significant speed advantage compared to a method performed not in conjunction with an online IR spectrometer. In order to obtain measurement points with a higher information density the online gas chromatograph (1) was switched on upon reaching a constant target conversion. The additional use of the online gas chromatograph (1) to characterize the process streams had the advantage that the determination of selectivities was further improved. Online IR spectroscopy provided the control parameter, wherein in the present case conversion was used as the control parameter. An improved determination of the selectivities made it possible to improve the differentiation of the catalysts.

[0117] The method for upgrading aromatic mixtures presently described by way of example may in the same way be used as a method for transalkylation reactions, for the dealkylation of ethylbenzene, for the disproportionation of toluene, for the isomerization of xylenes and in other processes.

BRIEF DESCRIPTION OF THE FIGURES

[0118] FIG. 1 shows a series of MIR spectra, which were recorded at a WHSV of 0.4 h.sup.1 during the training phase of the method. The method step is characterized with the number Experiment C1_03 in table 1. The time intervals for the recording of the spectra were 5 seconds, wherein the spectra recorded over a period of several minutes are shown.

[0119] FIG. 2 shows the composition of the process stream as a function of time which had been determined by GC at three different gas loads during the training phase and which had been calculated for MIR spectra. The WHSVs were 0.2 h.sup.1, 0.3 h.sup.1 and 0.4 h.sup.1 and were recorded in the experiments having the test numbers C1_01, C1_02 and C1_03. FIG. 2 top panel: Yields from gas chromatographic analysis; FIG. 2 lower panel: predicted yields from MIR spectra.

[0120] FIG. 3 shows data acquired during the training phase at low WHSVs, namely in a superimposed representation of GC and MIR data. The large symbols characterize the yields determined by gas chromatographic analyses; the small symbols characterize the predicted yields determined from the MIR spectra.

[0121] FIG. 4 shows the models obtained by relating GC data and IR data during the training phase. The predicted yields versus the measured yields are shown for 9 different individual substances.

[0122] FIG. 5 shows the RMSEP (root mean square error predicted) of the cross-validation for the trained models of the 9 different individual substances against the number of components, wherein up to 15 components were used.

[0123] FIG. 6 shows the yields from the MIR spectra predicted during the production phase for WHSVs in the target parameter space (i.e. there were high WHSVs of 5, 8, 12, 16 and 20 h.sup.1 in five experiments). The large symbols represent yields from the gas chromatographic analysis, the small symbols represent predicted yields from the MIR.

[0124] FIG. 7 shows values for conversion and temperature as a function of the time-on-stream (TOS) obtained when performing a method for converting aromatics in which 4 catalysts were arranged in a reactor system with 4-fold parallelization. Conversion is shown at the top and reactor temperature at the bottom. Conversion was determined solely by gas chromatograph.

[0125] FIG. 8 shows values for conversion and temperature as a function of the time-on-stream (TOS) obtained when performing a method according to the invention in which four catalysts were investigated in four reactors arranged in parallel. Conversion is shown at the top and reactor temperature at the bottom. Conversion was determined using an online gas chromatograph (1) and adjustment to constant conversion was performed using an online IR spectrometer (2).

[0126] FIG. 9 shows a schematic representation of the apparatus according to the invention in an embodiment equipped with five reaction spaces (11)-(15) arranged in parallel, wherein the online spectrometer (2) and the online gas chromatograph (1) are serially arranged in the process stream conduit (35) and the process control unit (4) is coupled to the reaction spaces.

[0127] FIG. 10 shows a schematic representation of the apparatus according to the invention in an embodiment which corresponds to the embodiment shown in FIG. 9, wherein the process stream conduit (35) is divided into two conduits and the online IR (2) and the online GC (1) are arranged in parallel in these two conduits.

LIST OF REFERENCE NUMERALS

[0128]

TABLE-US-00003 1 online GC, online gas chromatograph 2 online IR, online IR spectrometer or online spectrometer, preferably optimized for the MIR range 4 Process control unit 35 Process stream conduit 34 Waste air conduit 33 Multiport valve 21, 22,-25 Reaction space output-side process gas conduits connected to multiport valve 11, 12,-15 Process spaces, preferably reaction spaces, more preferably tubular reactors