METHOD FOR MANAGING AN IMPURITY IN A CYCLIC ADSORPTION PROCESS

Abstract

Title of the invention: METHOD FOR MANAGING AN IMPURITY IN A CYCLIC ADSORPTION PROCESS The invention relates to a method for separating a feed gas (10) by pressure-modulated adsorption in which a treatment unit is fed with the feed gas and produces a first gaseous fraction enriched in the most adsorbable constituents (11) and a second gaseous fraction enriched in the least adsorbable constituents, the treatment unit comprising a plurality of adsorbers (1), each of said adsorbers (1) containing an adsorbent mass (2) composed of a pretreatment adsorbent mass (21) placed upstream of a main adsorbent mass (22), said separation method being characterized by the implementation of an anti-pollution procedure intended to detect an abnormal progression of the impurity (30) in the pretreatment adsorbent mass (21), in particular towards the main adsorbent mass (22),and to avoid contamination of the main adsorbent mass (22) by the impurity (30). Abbreviated figure: Fig. 2,

Description

Title of the invention: METHOD FOR MANAGING AN IMPURITY IN A CYCLIC PROCESS BY ADSORPTION

[0001] The present invention relates to a method for separating a feed gas by PSA type adsorption. More particularly, this invention is intended to keep intact over time the adsorption capacity of the adsorbent mass used to carry out this separation while avoiding the risks of poisoning at least a fraction of this adsorbent mass by an impurity contained in the feed gas.

[0002] Generally speaking, a method implementing the adsorption phenomenon makes it possible to separate one or more molecules from a mixture containing them, by exploiting the difference in affinity of one or more adsorbents with respect to the constituent molecules of the mixture. In this way, from a feed gas, the fraction comprising the least adsorbable constituents is separated from the fraction comprising the most adsorbable compounds. This separation is generally not total and in particular, certain constituents may be found in both of these fractions.

[0003] The processes using adsorption are of several types depending on whether the adsorbent is regenerable in situ or not. We can then distinguish 3 main families of adsorption processes: lost charge processes, TSA processes (Temperature Swing Adsorption) and finally PSA processes (Pressure Swing Adsorption).

[0004] In lost charge processes, a new charge is put in place when the one currently in use is saturated by impurities or more generally when it can no longer play its protective role sufficiently.

[0005] Differently, TSA or PSA type processes are cyclic processes continuously linking adsorption and regeneration phases and generally using a plurality of adsorbers. The cycle time Te is the duration at the end of which an adsorber returns to the same state as previously, for example at the start of an adsorption step.

[0006] In TSA type processes, the adsorbent at the end of use is regenerated in situ, that is to say that the stopped impurities are evacuated so that the said adsorbent recovers the majority of its adsorption capacities and can start a purification cycle again, the essential regeneration effect being due to a rise in temperature. The times of The cycles involved are several tens of minutes and generally several hours.

[0007] Finally, in PSA-type processes, the adsorbent at the end of the production phase is regenerated by the desorption of impurities obtained by means of a reduction in their partial pressure. This pressure reduction can be obtained by a reduction in the total pressure and / or by flushing with a gas free of or containing few impurities. This gas can in certain cases be external to the unit. Unlike TSA, the adsorbent after regeneration generally contains a substantial fraction of constituents still adsorbed.

[0008] It should be noted that a PSA type unit is not exempt from thermal effects. The adsorption and desorption of impurities locally create heating and cooling whose amplitude (beat) can be up to a few tens of °C when the adsorbed quantity is significant. There may also be a temperature profile between the inlet and the outlet with zones that are hotter and / or colder than the average. These thermal effects generally have a negative impact on regeneration. Pressure swing adsorption processes are used both to remove traces of impurities - for example with a content of less than 1% in the feed gas - and to separate mixtures containing tens of % of different gases. In the first case, we generally speak of purification (for example gas drying) and fractionation in the second case (for example production of oxygen or nitrogen from atmospheric air).In the most complex cases, we can of course have purification and fractionation in the same unit. We generally speak of a separation or treatment process or unit in all cases. On the other hand, it is common to give more specific names depending on the pressure levels involved or the time required for an adsorber to return to its initial point (cycle time): .

[0009] VSA processes in which adsorption takes place substantially at atmospheric pressure, preferably between 0.95 and 1.25 bar abs and the desorption pressure is lower than atmospheric pressure, typically from 50 to 400 mbar abs

[0010] MPSA or VPSA processes in which adsorption takes place at a high pressure above atmospheric pressure, typically between 1.3 and 6 bar abs, or even 10 bar abs, and desorption at a low pressure below atmospheric pressure, generally between 200 and 750 mbar abs

[0011] The PSA processes themselves in which the high pressure is substantially higher than atmospheric pressure, typically between 3 and 50 bar abs and the low pressure substantially equal to or higher than atmospheric pressure, generally between 1 and 9 bar abs.

[0012] RPSA (Rapid PSA) processes for which the pressure cycle time is typically less than one minute

[0013] URPSA (Ultra Rapid PSA) processes for which the duration of the pressure cycle is of the order of a few seconds maximum.

[0014] In any case, the cycle time of a PSA is generally significantly lower than that of a TSA.

[0015] It may also be noted that the gaseous fraction recovered in a PSA process may correspond to the fraction produced at high pressure, i.e. to the fraction comprising the least adsorbable constituents, but also to the fraction extracted at low pressure provided that the constituent(s) sought are the most adsorbable in the mixture.

[0016] It should be noted that these various names (PSA, VSA, VPSA, MPSA, RPSA, URPSA) are not standardized and that the limits are subject to variation depending on the authors. Other names exist in the literature highlighting this or that particular point of the process such as a rinsing step or recycling.

[0017] It is recalled that the use in the description of the term PSA here covers all variants of PSA (therefore VS A and VPSA included). Sometimes referring to a specific process, in particular in an example, the terms VS A or VPSA may be used preferentially in the description without thereby limiting the application of the invention to this type of unit.

[0018] Similarly, herein, the use of the term "processing unit" refers to any type of PSA used for either the purification or fractionation of a feed gas.

[0019] Generally speaking, it is known that the cycle of a PSA unit comprises a plurality of stages depending on the process chosen. Examples include: adsorption, a plurality of equilibrations by decompression, supply of elution gas, final depressurization, elution, a plurality of equilibrations by compression, final repressurization, as well as more specific stages such as downtimes, recycling, rinsing stages, etc.

[0020] A step is said to be co-current if the gas flow during this step goes in the same direction as the feed gas during the adsorption step. Otherwise the step is said to be counter-current. For example, when there is pressure balancing between two adsorbers at the production end, the circulation is co-current in the adsorber which is depressurizing and counter-current in the adsorber during repressurization. For simplification, co-current steps are understood to mean all the steps where the gas circulates from the inlet to the outlet of an adsorber and the adsorption step is therefore part of these steps.

[0021] Pressure modulation adsorption treatment units generally include a control-command device which allows them to monitor the selected pressure cycle and to respect a certain number of constraints imposed by their industrial environment or by their own equipment. To this end, the control-command system constantly receives from the various sensors installed on said unit the actual data concerning its operation and sends orders to the control devices, namely essentially the valves (total or partial opening control, closing) as well as possibly the exchangers and machines, compressor or vacuum pump. For example, the aim will be to ensure that the unit operates cycle after cycle while respecting at the end of the different stages the sequence of characteristic pressures defined during dimensioning, flow rates, stage durations as well as for example the purity specified for production. These values to be respected then appear in the form of set points in the control-command system.

[0022] A control panel, also called a human-machine interface, allows the operator to monitor, operate and optimize the unit, in particular by modifying, if necessary, at least one of the pressure cycle set points. However, there are also PSA units operating without personnel nearby. As long as this has been provided for in the design, the control-command system can modify, by itself or via an associated module, certain set points based on information received from the site, for example in order to avoid the drift of a parameter - such as the purity of the production - which could lead to a tripping of the unit.

[0023] It is also known that in many PSA-type processes, the adsorbent mass is made up of several layers of adsorbent, each ensuring the stopping of specific molecules to ultimately produce a gas purified or enriched in certain constituents. This is called multi-bed adsorbents.

[0024] The choice and dimensioning of these successive layers are carried out during the design so that the different constituents are essentially stopped by the adsorbent best suited to this purpose. This is well known to those skilled in the art and numerous publications detail, for example, the nature of the multi-beds of adsorbents intended for H2 PSAs and the reason for their choice.

[0025] The fact that a constituent intended to be stopped on layer i passes partly onto the following layer i+1 has the effect that this layer i+1 no longer has sufficient adsorption capacity to adsorb the constituent intended for it and this phenomenon can be reflected in the following layers to the point of degrading the expected performance of the treatment unit.

[0026] More seriously, the adsorption of a constituent on a layer for which its presence was not intended can have irreversible consequences: destruction of the internal structure of the adsorbent, blocking of the pores, definitive occupation of the active sites... We generally speak of a poison-type impurity or more simply of poison in this case.

[0027] It will also be noted that a similar phenomenon occurs in the case of an adsorbent mass containing a single type of adsorbent, since a constituent is adsorbed very strongly on this material and eliminates any possibility of co-adsorption for the less adsorbable constituents. This effect is then taken into account when designing the unit and a first adsorbent mass is sized to stop the highly adsorbable constituent and a second mass to carry out the desired separation between the constituents of the feed gas thus purified. It is then customary to call the first adsorbent mass "pretreatment adsorbent mass" and the second "main adsorbent mass" and to give the name impurity to the most adsorbable constituent, often present in small quantities in the feed gas.

[0028] Whether in the case of multiple beds of specific adsorbents or a homogeneous layer of a single adsorbent, it is important to ensure during operation of the treatment unit that a constituent does not go beyond what was planned during the design in order to maintain the expected adsorption capacities for the other constituents. In particular, in the case of a poison, too great an advance of this constituent can have such consequences that it will no longer be possible to achieve the required performance.

[0029] To this end, the designer of the treatment unit defines a nominal operation, generally corresponding to 100% of the feed gas flow rate with the most restrictive conditions on the operating characteristics (high pressure, low pressure, temperature, composition) if the latter are provided in the form of ranges or with the operating characteristics as precisely fixed in the study bases. It is at this stage that the main and pretreatment adsorbent masses are determined, among other things. In particular, an impurity, henceforth called the impurity, has led to defining the nature and volume of the pretreatment adsorbent mass necessary for it to remain confined there during operation. Reference will be made frequently in the following text to this nominal operation which has been used for the dimensioning of the treatment unit.

[0030] At least two methods are currently available - one direct, the other indirect - to theoretically verify that the main adsorbent mass has not been affected during operation by the impurity.

[0031] The first is to regularly check that the unit's performance is as expected. Such a finding means, a priori, that the adsorbents are working as expected.

[0032] The second consists simply of verifying by analysis of the gas mixture circulating in the adsorber that the impurity remains in the pretreatment part.

[0033] Unfortunately, it has become apparent that in many cases, and for reasons which will be detailed later, these a priori simple methods are not satisfactory due to a lack of precision due to uncertainties in the measurement of performance parameters (yield, productivity, etc.) or in the analysis of the content, the uncertainties then being linked to the nature of the impurity to be analyzed.

[0034] There is therefore a need in a PSA type adsorption separation unit to be able to detect in time the abnormal advance of an impurity, especially of the poison type, in a pretreatment adsorbent mass intended to stop it and thus to avoid the pollution, often irreversible, by the latter of the main adsorbent mass installed downstream and intended for the production of at least one gas enriched in a recoverable constituent. This risk of pollution generally comes from a change in the operating conditions (feed gas, cycle characteristics) or from the reduction in adsorption performance (capacity, kinetics, etc.) due to the aging of the pretreatment adsorbent.

[0035] The inventors have demonstrated that it is simpler and safer to no longer try to systematically measure the traces of impurity in the pretreatment adsorbent mass in an attempt to detect their progression towards the main adsorbent but on the contrary to analyze at least periodically the content, in a well-defined place and at a given moment in the cycle, of another constituent a priori less adsorbable than the impurity. Its much simpler and more reliable analysis will make it easier to judge the stability in the adsorbent mass, from one cycle to the next, of the most adsorbable constituents and therefore of the impurity. In one of its variants, the invention also makes it possible to estimate the moment when the risk of pollution by impurities will become too great to continue operating the separation unit under the present conditions.

[0036] Thus, the invention relates more specifically to a process for separating a feed gas by pressure-modulated adsorption in which a treatment unit is supplied with the feed gas and produces a first gaseous fraction enriched in the most adsorbable constituents and a second gaseous fraction enriched in the least adsorbable constituents, the treatment unit comprising a plurality of adsorbers, each of said adsorbers containing an adsorbent mass composed of a pretreatment adsorbent mass placed upstream of a main adsorbent mass, the pretreatment adsorbent mass and the main adsorbent mass being chosen and sized such that: - the pretreatment adsorbent mass stops more than 99%, preferably all of an impurity contained in the feed gas, thus producing a purified feed gas, - the main adsorbent mass fractionates the purified feed gas into the gaseous fraction enriched in the most adsorbable constituents which are adsorbed on the main adsorbent mass and the gaseous fraction enriched in the least adsorbable constituents, said separation process being characterized by the implementation of an anti-pollution procedure intended to detect an abnormal progression of the impurity in the pre-treatment adsorbent mass, in particular towards the main adsorbent mass, and to avoid pollution of the main adsorbent mass by the impurity, said procedure comprising the following steps: 1- Choose one of the Y constituents, other than the impurity, from among the most adsorbable constituents of the purified feed gas, 2- Determine a given instant of the pressure cycle and a point of the adsorbent mass included in a mass transfer zone for this constituent Y at the given instant of the cycle, the mass transfer zone being located between a zone of the adsorbent mass saturated with this constituent Y at the given instant of the pressure cycle and a zone of the adsorbent mass which, at the given instant of the pressure cycle, has not yet been in contact with this constituent Y, 3- Determine, if necessary, for this constituent Y, a forecast value of its content at the point of the adsorbent mass and at the instant of the cycle retained in step 2, 4- Measure, with the treatment unit in operation, at the point of the adsorbent mass and at the instant retained in step 2 and on at least one of the adsorbers of the unit, a content of the constituent Y, 5- Compare this measured content to one or more contents of this constituent Y previously measured at said point of the adsorbent mass and / or compare this content to the predicted value determined in step 3, 6- Determine a parameter linked to the measured content based on the comparison in step 5, 7- Issue an alert concerning a risk of pollution of the main adsorbent mass by the impurity if the parameter linked to the measured content exceeds a tolerance threshold.

[0037] Before mentioning below the different possible implementations of the separation process characterized by this anti-pollution procedure, it is appropriate to mention or recall a certain number of general points. Thus, the articles "a, an, the, the" used in this text do not have the restrictive meaning of uniqueness of the thing associated with the article, such a restriction, if necessary, would then be specified by the adjectives "sole or unique". Stopping "an" impurity contained in the feed gas on the pretreatment adsorbent mass means stopping "at least one impurity" since a feed gas can contain several. A plurality N of elements means any number starting from 1 and without upper limitation except when N is accompanied by a voluntary limitation such as N>2. The expression "at least one" does not necessarily imply that there are several elements, but only that if there are several, at least one will be concerned.

[0038] According to one implementation of the method, the time taken to measure the content of the constituent Y corresponds either to a fixed duration after the start of a step, for example 25 seconds after the start of the adsorption step, or is a function of the duration which may vary during the operation of the chosen step, for example the time taken then corresponds to the end of the adsorption step or to 90% of its duration.

[0039] According to one implementation of the method, the measurement of the content of constituent Y in step 4 is carried out every X cycles, X being able to be a function of the value of the parameter linked to the measured content, X being for example equal to 1 if said parameter indicates a change in the content of (Y) or conversely X being able to correspond to a few hours of operation if said parameter is the indication of stable operation.

[0040] According to one implementation of the method, the successive values of the measured contents of the constituent Y are recorded and processed in the form of a trend curve whose slope, corresponding to a delta of content Y over time, for example expressed in ppm mole per cycle, is representative of the stability or the evolution of this content and constitutes a parameter linked to the content determined in step 6.

[0041] According to one implementation of the method, the tolerance threshold is then a maximum number of successive measurements with a positive slope highlighting a regular progression over time of the constituent Y in the adsorbent mass or a maximum value of this positive slope highlighting a rapid evolution of the content of the constituent (Y) at this given point of the adsorbent mass and at this given instant of the cycle.

[0042] According to one implementation of the method, the predicted value of the content of the constituent (Y) is determined for the nominal operation of the treatment unit, operation for which the pretreatment adsorbent mass installed generally comprises an oversizing constituting a margin with respect to the risk of pollution of the main adsorbent mass by the impurity, the nominal operation corresponding as indicated previously to the conditions used to size the treatment unit.

[0043] According to a complementary implementation of the method, a second forecast value of the content of the constituent Y is determined, corresponding to the extreme operation of the treatment unit, operation for which the impurity reaches the end of the pretreatment adsorbent mass having then used the entire margin provided during the design.

[0044] According to one implementation of the method, a parameter linked to the measured content is the difference between the predicted content and the measured content of the constituent Y.

[0045] According to an implementation of the method, a second parameter linked to the measured content is the difference between the second predicted content and the measured content of the constituent Y.

[0046] According to one implementation of the method, the successive values of the parameter linked to the measured content determined during the cycles, in particular the successive deviations between the forecast content and the measured content of the constituent Y, are recorded and processed in the form of a trend curve whose slope is representative of the stability or the evolution of this content over time, and in the case of a progression, the processing of these successive values makes it possible to estimate the time or the number of cycles remaining before the start of pollution by the impurity of the main adsorbent mass.

[0047] According to one implementation of the method, the adsorbent of the pretreatment adsorbent mass is different from that of the main adsorbent mass, said difference possibly being linked to the nature or chemical composition of the adsorbents (for example activated alumina vs. zeolite or even standard zeolite vs. exchanged zeolite) or only to the physical characteristics (highly activated and acid-washed activated carbon vs. standard activated carbon or even very pure silica gel without traces of alumina vs. standard silica gel).

[0048] According to one implementation of the method, the adsorbent of the pretreatment adsorbent mass is similar to that of the main adsorbent mass, the main adsorbent mass then corresponding to the quantity of adsorbent not polluted by the impurity necessary to fractionate the purified feed gas into the first and second gaseous fractions, and the pretreatment adsorbent mass being intended to stop the impurity.

[0049] According to a characteristic of the process, the presence of the impurity in the main adsorbent mass causes the reduction of the adsorption capacity of said main adsorbent mass and / or of its selectivity and / or of its kinetics, said reductions being able to be caused by a chemical degradation of the internal structure of the adsorbent, by deposits on its surface, by a reduction in the internal porosity, by a preferential adsorption of the impurity occupying the adsorption sites normally allocated to the other constituents of the feed gas.

[0050] According to a characteristic of the process, the preferential adsorption of the impurity in the main adsorbent mass is not completely reversible during regeneration and causes a definitive reduction in the separation performance of the polluted zone.

[0051] According to one implementation of the method, the pretreatment adsorbent mass comprises activated alumina, standard or doped, silica gel, activated carbon or a zeolite of type 3A, NaX or Y, in a single layer, in a mixture or in successive layers.

[0052] According to one implementation of the method, the main adsorbent mass comprises a zeolite, particularly a zeolite of type LSX, X, A, Y, said zeolite being preferentially exchanged with at least one cation such as K, Ca, Li, Ba, Sr, Ag..., or a CMS (Carbon Molecular Sieve) or a synthetic adsorbent specially calibrated for a kinetic type separation, a metal-organic structure (MOF), a resin, a silica gel or an activated carbon, a polymer-based material or more generally a porous structure preferably containing a substance capable of reacting reversibly with certain molecules, a substance such as a physical solvent or a metal complexing agent for example, these adsorbents being able to be in a single layer, in a mixture or in successive layers.

[0053] According to one implementation of the method, the impurity to be stopped in the pretreatment adsorbent mass is part of the group comprising: H2O, NH3, alcohols, ketones, aldehydes, ethers, esters, acids, amines, mercaptans, halogenated hydrocarbons, siloxanes, unsaturated hydrocarbons, C6 and C6+ type hydrocarbons, including BTEX, terpenes, more generally VOCs.

[0054] According to a characteristic of the process, there are a plurality of impurities to be stopped in the pretreatment adsorbent mass.

[0055] According to one implementation of the method, the constituents fractionated by the main adsorbent mass belong to a group of constituents comprising CO2, H2S, hydrocarbons C1 to C7, COS, carbon monoxide, nitrogen, oxygen, argon, hydrogen, neon, krypton, xenon and helium.

[0056] According to one implementation of the method, the constituent Y belongs to a group of constituents comprising CO2, H2S, hydrocarbons C1 to C7.

[0057] According to one implementation of the method, the frontal mass of the constituent Y is located, at the selected instant, at least partly downstream of the frontal mass of the impurity.

[0058] According to one implementation of the method, the time taken is during the adsorption step, preferably at the end of the adsorption step.

[0059] According to one implementation of the method, the selected time is located during one of the co-current steps following the adsorption step, preferably during the last co-current step, even more preferably at the end of this last co-current step.

[0060] According to one implementation of the method, the point of the adsorbent mass and the instant of the cycle retained to determine the content of the constituent Y are chosen such that this content is between 90% and 10% of the maximum content of this constituent at the start of its mass transfer zone, preferably between 80% and 20% of this content, still preferably between 70 and 30%.

[0061] According to one implementation of the method, the point of the adsorbent mass retained is located

[0062] either between 50 and 100% of the pretreatment adsorbent mass, 100% corresponding to the interface with the main adsorbent mass, preferably between 60 and 95%, still preferably between 70 and 90%.

[0063] either in the first 20% of the main adsorbent mass, 0% corresponding to the interface with the pretreatment and 100% to the entire main adsorbent mass, preferably in the first 10% of the main mass.

[0064] According to one implementation of the method, the choice of the constituent Y, of the location of the point retained in the adsorbent mass, of the instant of the cycle retained at which its content is determined was established by means of simulations carried out by dynamic simulation software representative of the operation of the unit.

[0065] According to one implementation of the method, the choice of the constituent Y, of the location of the point retained in the adsorbent mass, of the instant of the cycle retained at which its content is determined was established by means of laboratory tests representative of the operation of the industrial unit or on a pilot unit equipped with the appropriate measuring points and analysis means.

[0066] According to one implementation of the method, the time and the predicted content of the constituent Y at the chosen point were finalized experimentally during the industrial commissioning of the unit.

[0067] According to one implementation of the method, the exceeding of a first tolerance threshold by a pollution risk parameter results in an alarm.

[0068] According to one implementation of the method, exceeding a second tolerance threshold by a pollution risk parameter results in a second alarm and, without acknowledging this alarm, automatically triggers preventive conservation actions, actions allowing the integrity of the main adsorbent mass to be safeguarded.

[0069] According to one implementation of the method, the preventive actions implemented to stop the progression of the constituents in the adsorbent mass have been determined beforehand and may include, in a non-exhaustive manner:

[0070] the priority modification of at least one set point normally managed by the unit's control system, for example a reduction in the flow rate of load gas with blocked cycle time, or an increase in the volume of elution gas,

[0071] the establishment of a safety cycle provided during the design of the unit, for example a cycle comprising one less pressure balancing,

[0072] the stopping of the unit.

[0073] According to one implementation of the method, the choice among the preventive conservation actions is made according to the time or the number of cycles estimated before the start of pollution of the main adsorbent mass by the impurity.

[0074] According to one implementation of the method, the predicted value of the content of the constituent Y is a function of the actual operating conditions a, b, c... of the treatment unit according to a previously determined relationship, for example established by means of simulations carried out by dynamic simulation software representative of the operation of the unit for the operating conditions a, b, c... said operating conditions being able to be, in a non-limiting manner, the temperature, the pressure of the feed gas, the content of the constituent Y in the feed gas, the low pressure of the cycle... and the parameter linked to the measured content is then established as a function of this predicted value.

[0075] According to one implementation of the method, the measurement of the content of the constituent Y is carried out during the operation of the unit, at the instant of the cycle selected, by means of a measuring probe located in the adsorber at the selected point of the adsorbent mass or via a sample taken from the circulating gas at the selected point by means of a sampling pipe, the gas thus sampled feeding an analyzer suitable for measuring the content of the constituent Y.

[0076] According to another aspect of the invention, the analysis of the constituent Y at the selected point is carried out continuously - or semi-continuously - and the value of the content of the constituent Y is then extracted from these analyses at the chosen instant of the cycle, the response time of the analysis then being taken into account.

[0077] According to one implementation of the method, the treatment unit comprising a plurality N of adsorbers, with N>2, several adsorbers, preferably all the adsorbers, are provided with a sampling point placed at the same point of the adsorbent mass and each follow the procedure described above.

[0078] According to one implementation of the method, the impurity is H2O and the constituent Y is CO2.

[0079] According to one implementation of the method, the impurity is a constituent whose molecule contains at least 6 carbon atoms and the Y constituent of CO2 or H2S.

[0080] According to one implementation of the method, the treatment unit is a VSA, VPSA or PSA 02 producing an 02-enriched fraction from atmospheric air.

[0081] According to one implementation of the method, the treatment unit treats a feed gas containing hydrogen, in particular hydrogen from an ammonia cracker, an alcohol cracker or reformer, generally methanol, a synthesis gas reactor, a coke oven gas or from a chemical or petrochemical process off-gas preferably containing more than 50 mol% H2.

[0082] According to one implementation of the method, the treatment unit treats a feed gas containing methane, in particular biogas.

[0083] According to one implementation of the method, the treatment unit treats a feed gas containing CO and uses an adsorbent loaded with copper.

[0084] According to one implementation of the method, the treatment unit treats a feed gas generally resulting from combustion or fermentation, said feed gas not having been completely freed from its impurities through upstream treatments (washing, guard bed, etc.).

[0085] The invention will be better understood upon reading the following description and examining the accompanying figures. These figures are given only for illustrative purposes but in no way limit the invention.

[0086] Generally speaking, during an adsorption step, the adsorbable constituents are progressively retained by the adsorbent and occupy, during the step, an increasingly large fraction of the previously regenerated adsorbent mass. There is generally displacement of the less adsorbable constituent by the more adsorbable constituent. It is also said that the most adsorbable constituent drives out in front of it - at least partially - the less adsorbable constituent which previously occupied the place. As a result, at the end of the adsorption step, the adsorbable constituents each occupy a fraction of the adsorbent mass, a fraction in which they are largely in the majority. Thus, in an H2 PSA, we will find, for example, from the feed gas inlet end to the hydrogen outlet end, a zone stopping humidity, a zone stopping CO2, a zone stopping methane, and a zone stopping nitrogen.Due to this segregation between adsorbable constituents, the skilled person quickly introduced the concept of multi-layers (or multi-beds) into PSA processes, each of the layers being selected to stop a given constituent, the whole then making it possible to optimally produce a gas purified or enriched in certain constituents. A rule was then to use "a weak adsorbent to adsorb the most adsorbable constituents placed in front of a strong adsorbent to adsorb the least adsorbable constituents".

[0087] The choice and dimensioning of these successive layers are carried out during the design so that the different constituents of the feed gas are essentially stopped by the adsorbent best suited to this purpose. This nevertheless has the effect of having to dimension N successive layers instead of providing a single homogeneous adsorbent mass, as was the case for example for the first generation of H2 PSA for which it was customary to say that all adsorbents were suitable, in particular activated carbon or zeolite 5A before then applying the rule stated above. This optimization by successive layers nevertheless introduces a risk. The fact that a constituent intended to be stopped on layer i partly passes onto the The following layer i+1 being less suitable for this constituent also has the effect that this layer i+1 no longer has sufficient adsorption capacity to adsorb the constituent intended for it and this phenomenon can be reflected in the following layers to the point of degrading the expected performance of the treatment unit.

[0088] More seriously, the adsorption of a constituent on a layer for which its presence was not intended can have irreversible consequences: destruction of the internal structure of the adsorbent, blocking of the pores, permanent occupation of the active sites, etc. We generally speak of a poison-type impurity in this case. It will also be noted that a similar phenomenon occurs in the case of an adsorbent mass containing a single type of adsorbent, since a constituent is adsorbed very strongly on this material and eliminates any possibility of co-adsorption for the less adsorbable constituents. This effect is then taken into account when designing the unit and, in practice, this amounts to sizing a first adsorbent mass to stop the highly adsorbable constituent and a second mass to carry out the desired separation between the constituents of the feed gas thus purified.It is then customary to call the first adsorbent mass "pretreatment adsorbent mass" and the second "main adsorbent mass" and to give the name of impurity to the most adsorbable constituent, often present in small quantities in the feed gas, of the order of a few percent at most and often at the level of hundreds of ppm.

[0089] Whether in the case of multiple beds of specific adsorbents or a homogeneous layer of a single adsorbent, it is important to ensure during operation of the treatment unit that a constituent does not go beyond what was planned during the design in order to maintain the expected adsorption capacities for the other constituents. In particular, in the case of a poison-type impurity, too great an advance of this constituent can have such consequences that it will no longer be possible to achieve the required performance. In the latter case, it is conventional during the design of the unit to take for the nominal (i.e. sizing) operation a safety on the pretreatment adsorbent mass, generally from 5 to 15% to take into account small variations around the case chosen.It might seem more prudent to take, at least in cases that could become critical, a large margin, for example 30% or even 50% on the sizing. In practice, such a safer approach is rarely adopted because it has a cost not only on the investment but also on the performance (loss in extraction efficiency, additional energy expenditure, etc.). The sizing adopted is therefore a compromise between performance and risks.

[0090] There are methods to try to minimize these risks. The first is simply to ensure that the processing unit always operates as expected at the design and start-up, which constitutes indirect proof that the adsorbents still have the expected adsorption capacity. This method, simple in theory, can actually raise the alarm on an operating incident but as far as we are concerned here, that is to say the pollution of the adsorbent mass by the impurity which has gone beyond the pretreatment mass, it is an alarm a posteriori especially since the comparison between real and theoretical operation is not easy, if for example the unit operates at reduced flow rate. Document FR 3 110 722 describes a solution to make this comparison more effective and detect a performance problem on the treatment unit more quickly, however this solution has its limits.The accuracy that can be obtained from a unit performance indicator in operation, such as an extraction efficiency, is generally of the order of 2 percent minimum when this indicator is based on flow rate measurement and gas analysis. Over time, this can highlight a drift but without information on the cause (charge gas, cycle, instrumentation, valves, etc.) which will have to be sought elsewhere. This procedure also requires having adsorption process simulation software available, which is still very rare on site, or having a data transmission network and a central unit where the calculations would be carried out by specialists.

[0091] The second method is conversely the most direct because it consists of measuring the content of the impurity in the circulating gas, for example at 90 or 95% of the pretreatment adsorbent mass, where, due to the margin taken into account at the design stage, the impurity should not be detected if the operation of the unit is in accordance with what is expected.

[0092] It will be noted in this regard that the person skilled in the art hesitates to say that the entirety of a constituent has been stopped at a point in the adsorbent mass because it is likely that some molecules are still in the gas beyond this theoretically calculated limit point, due to dispersion or diffusion. It is then common to write, as has been done here, that the pretreatment adsorbent mass stops more than 99% (and preferably all) of an impurity contained in the feed gas. In practice, this means here that at any time during the adsorption step, less than one percent of the quantity of impurity entering the feed gas passes into the main adsorbent mass. Given the phenomenon of progression of the constituents in an adsorbent, over the duration of the step the quantity actually passed would be much lower.

[0093] This analysis of a constituent of the feed gas at a point located towards the end of the adsorbent layer specifically dedicated to it is generally carried out when the installed margin is relatively low and the presence of said constituent in the following layer poses a problem. For example, it is verified in this way that H2S is stopped on a very pure activated carbon in terms of its chemical composition and does not pass onto the next layer composed of a standard activated carbon, carbon on which it would decompose, giving sulfur deposits blocking the porosity. However, in many cases, these analyses are not satisfactory due to a lack of precision because of uncertainties in the analysis of the content, these uncertainties being then linked to the nature of the impurity to be analyzed, to the low contents that we seek to detect as well as to the operating conditions of the treatment unit.

[0094] Systems for measuring pollutants on adsorption columns have already been described in the literature. In the majority of cases, the measurement is carried out with an online analyzer supplied with a small fraction of the gas flow rate to be analyzed. The sample is generally taken as already indicated in the guard bed, for example at 85% of the installed volume, so that there remains a margin downstream allowing for action. These measurement means are easily applicable in the case of processes in a classic TSA cycle where production times are of the order of a few hours, long enough to sample gas under stabilized conditions in order to accurately detect the passage of the end of an impurity front. The evolution of the front itself is slow, leaving several minutes or even tens of minutes for reaction.Conversely, such a system applied to a PSA in which a plurality of steps of a few seconds or tens of seconds are linked together leads to imprecise analyses due to the rapid variation in the composition of the gas flow during the progress of the steps. In addition, along the sampling circuit, the content gradient resulting from these variations in the adsorber leads to a mixing effect by diffusion and dispersion. Finally, the response time of the analyzer, the adsorption / desorption of impurities in the connections between the sample intake and the analyzer tend to average the content peaks distorting the analysis... Manual sampling over short periods in flasks or other containers for subsequent analysis in the laboratory is not a very precise solution because, from experience, the results obtained vary from one time to another.This non-reproducibility is mainly due to the fact that it is impossible on site to start and finish each sample at exactly the same time during a step and to open and close the sampling valve in the same way. Small air inlets at the time of connections can also distort the analysis result, especially in the case of sampling from a container initially under vacuum.

[0095] These considerations on the difficulty of accurately analyzing constituents, in particular if they are in trace form, that is to say with contents of a few ppm or tens of ppm, are unfortunately particularly true for the majority of impurities which can be considered as poisons or at least be very troublesome if they migrate out of the adsorbent mass reserved for them, such as H2O, NH3, unsaturated hydrocarbons, C6 and C6+ hydrocarbons, including BTEX, more generally VOCs or tars (sometimes called tar vapor). Humidity in particular has an excessively unfavorable effect on the adsorption capacity and selectivity of a majority of zeolites, especially on specific zeolites such as those exchanged with lithium intended for O2 / N2 separation.

[0096] Quite logically, the detection of water towards the pretreatment / main adsorbent interface has been the subject of numerous tests. The various devices for determining the humidity of a gas, devices which, like probes, could be introduced locally into the adsorber, do not meet the needs well for various reasons: need for calibration, response time, precision, result depending on possible contacts with the adsorbent particles... The safest way would consist of using a system for sampling the circulating gas at a suitable position and analyzing its water content. However, it is well known that water analysis is very delicate even for specialists, in particular if low contents are sought, which moreover vary during the cycle. The mere installation of heated lines and a thermostatically controlled enclosure does not then guarantee the reliability of the results.There are, at least residually and despite an adequate choice of materials, adsorption / desorption effects of water molecules on the wall which can prevent, for example, the detection of a pollution peak. Cells measuring water content also experience drift phenomena generally indicating lower contents than reality. Finally, the intrinsic response time of these cells may be insufficient in the case of PSA with very short steps, in which case the measured value corresponds more to an average than to an instantaneous value. It would be advisable to place the already well-calibrated analyzer in the immediate vicinity of the sampling point to improve the accuracy of the measurements, but this can a priori only be a temporary procedure given all the requirements linked to the use of these analyzers.

[0097] It should be noted that the measurement of the water content at the desired level can be a one-off analysis of the type described above carried out by specialists at the time of start-up to check the satisfactory operation of the pretreatment (sizing chosen, state of the adsorbent, etc.) and / or periodic checks spaced out, for example every 6 or 12 months, but very difficult to envisage in a continuous or semi-continuous manner such as once a day or a week. It is then hardly credible to want to use such analyses to operate the unit, or even automatically adapt the operating characteristics of the unit according to the measured values obtained. Such adaptations, on the contrary, presuppose continuous or quasi-continuous monitoring, i.e. at least once a day for a stability monitoring and every few cycles during periods of changing operating conditions or detected pollution risk. In addition, the analysis must be reliable and accurate.

[0098] Faced with this observation, the question arose as to whether there could be another indicator of the risk of pollution, which constitutes an increase in the impurity in the adsorbent mass.

[0099] A first observation is that in many cases, the impurity is one of the most adsorbable constituents (water, heavy hydrocarbons, VOCs, etc.) or the most reactive (acids, polymerizable constituents, etc.), which clearly justifies stopping it on the first layer of adsorbent, which constitutes the pretreatment mass.

[0100] A second observation is that the feed gases that we seek to treat in PSAs comprise a plurality of constituents, one part constituting the most adsorbable fraction, the other part the least adsorbable fraction. For example, in the case of a PSA 02, we will have H2O, CO2, N2, 02, Ar without counting the secondary compounds of atmospheric air.

[0101] We have seen previously that depending on the affinity between constituents and adsorbents, said constituents end up occupying distinct zones of the adsorbent mass during the adsorption stage, zones in which they are in the majority. This distribution of constituents is governed by the physical laws relating to dynamic adsorption. This point will be addressed in a little more detail when discussing simulation software for this type of process. Without going into detail, we note that during the adsorption stage, the transfer of gas molecules to the adsorption sites takes place on only a portion of the adsorbent bed.We then differentiate 3 zones: the upstream zone called the saturated zone where the adsorbent is completely loaded with constituent, the downstream zone which has not yet been in contact with the constituent in question and which is called the virgin zone and between the two, the zone where the transfer of the constituent contained in the gas takes place towards the adsorption sites still partially free, zone to which we give indifferently the name of adsorption front, material front, frontal zone, mass transfer zone or which we also call by the acronyms ZTM or MTZ (Mass Transfer Zone). This zone moves in the direction of the gas flow. During its co-current movement, it generally has an inverted S shape both when we trace the composition of the constituent in the gas phase or in the adsorbed phase (adsorbate). We generally speak of the end of the front to define the interface between the front and the virgin zone.The presence of several constituents that can co-adsorb simultaneously complicates the shapes of the fronts but does not fundamentally modify the phenomenon described.

[0102] There is generally displacement of the less adsorbable constituent by the more adsorbable constituent. It is also said that the more adsorbable constituent drives out in front of it - at least partially - the less adsorbable constituent which previously occupied the place.

[0103] During regeneration, the adsorbed constituents are pushed back against the current towards the inlet of the adsorber and, in a somewhat similar manner, for a given constituent, an end of the front is obtained at the end of regeneration, the end downstream of which the constituent in question has been completely desorbed.

[0104] Stable operation of the PSA corresponds to an operation for which the extreme positions of the adsorption and regeneration fronts no longer change from one cycle to the next. There is then neither accumulation nor loss of molecules in the adsorbent mass over a complete cycle and what enters with the feed gas comes out in the production or the waste of the treatment unit.

[0105] [Fig.l] [Fig.l] schematically represents the envelope of an adsorber 1 containing an adsorbent mass 2 of which the pretreatment mass 21 and the first part of the main adsorbent mass 22 have been shown. The feed gas 10 enters at one end, on the pretreatment side, and the gas separated from its most adsorbable constituents 11 exits at the other end. The ordinate represents the content of the constituents in the gas circulating along the adsorber. The impurity 30 is distributed as described above into a first saturated zone 31, a mass transfer zone 32. Upstream of the end of the mass transfer zone is the virgin zone 33. The end of the mass transfer zone of the impurity has not reached the main adsorbent mass 22. Also shown in this [Fig.l], the constituent 40, less adsorbable than the impurity, with its own mass transfer zone 42.

[0106] If we are at the end of the co-current stages, the mass transfer zones 32 and 42 are then at the limit of their extension towards the outlet of the adsorber. Assuming that we are at the nominal operating point, the point which was used to size the unit, the volume of adsorbent between the start of the virgin zone and the mass 22 constitutes the margin taken into account in the design of the treatment unit to avoid the risks of pollution. Curve 34 represents, still for the impurity, the mass transfer zone at the end of regeneration. As already indicated, the desorption of the constituents during regeneration is only partial.

[0107] It is understood that with unchanged operation of the unit, in particular with constant cycle time, if the feed flow rate decreases, a smaller quantity of the different constituents will enter the adsorbent mass and the extension limit of the mass transfer zones will move back towards the inlet end. The margin with respect to pollution will be increased. With a cycle time adapted to the new reduced flow rate, i.e. extended in inverse proportion to the flow rates, the equations governing adsorption show that the mass transfer zones would return to their initial position.

[0108] The question that then arises is to determine the causes that can lead to pollution of the main adsorbent mass when the treatment unit has been sized according to the rules of the art for the nominal case. In practice, these causes appear relatively multiple with variable occurrences. We can try to classify them into 3 categories, each of which can correspond to several scenarios: - General reduction in the adsorption capacity of the adsorbent mass, - Reduction in the adsorption capacity of the impurity on the pretreatment adsorbent mass, - Unforeseen increase in the content of the impurity in the feed gas.

[0109] The general decrease in the adsorption capacity of the adsorbent mass affects a priori all the constituents even if this can be to different degrees. This can be the result of a modification of the operating parameters such as the pressure or temperature of the feed gas but more probably of a decrease in the regeneration power of the treatment unit which is significantly more difficult to notice. Concerning the feed gas, it is understood that if its pressure is lower than the design pressure, the adsorption capacities are automatically reduced. This is all the more true when the partial pressure of the constituent is low and its adsorption therefore takes place in the linear zone of the isotherm. Since the impurities are often in the form of traces, this is generally the case. Their progression will then be greater than that of the main constituents.Normally the treatment unit pressure is regulated and a drop in operating pressure is signaled by an alarm, however a pressure sensor may drift or a partial blockage downstream of the sensor may cause it to operate at a lower pressure than expected without attracting the operator's attention. The impurity will then progress into the main adsorbent mass before the unit's performance is eventually affected enough to highlight the problem.

[0110] It should be noted that the start of pollution of the main adsorbent mass is rarely detected because the accuracy of the measurements, in particular the flow rates, does not allow for judging the performances until there is sufficient degradation of these. The control system will correct automatically in order, for example, to maintain the purity of the product by reducing the cycle time by a few seconds. It is generally when the demand for gas produced increases and the treatment unit is to be operated at its maximum speed that the problem appears. This remains true whatever the cause of the start of pollution of the main adsorbent mass by the impurity.

[0111] Another cause creating a decrease in the adsorption capacity of the adsorbent mass is a decrease in the regeneration power of the treatment unit. In this case, at the end of the regeneration, the adsorber contains more adsorbable constituents than expected during the design of the unit and consequently at the end of adsorption, the stopping of the feed gas constituents requires an additional mass of adsorbent. The mass transfer zones all progress towards the outlet end. The regeneration efficiency is characterized by what is called the regeneration power (PF) of the PSA which is defined by the ratio of the gas volumes passing countercurrently through the adsorbent mass during a cycle compared to the gas volumes passing cocurrently through said mass, the respective volumes being counted in real m3.By regeneration power (PF), we mean here its most complete definition, corresponding to the theoretical "Purge Factor", that is to say that it is determined locally (here for example in the pretreatment adsorbent mass) and takes into account all the gas flows circulating at this level during all the stages of the cycle. For information, there is another global, simpler definition which only takes into account the feed gas during the adsorption stage (for the co-current) and the elution gas during the elution stage (for the counter-current). This is not the definition retained here, but depending on the cycles envisaged for the PSAs, this simplification can indeed be a good approximation of the more rigorous formulation. The regeneration power (PF) must be greater than 1 and is often between 1.1 and 1.3 for a majority of PSAs.For the most complex PSA cycles with a plurality of co- and counter-current stages, a modification to one of these stages can modify the regeneration power and have an effect on the position of the adsorbates along the adsorbent mass. A relaxation of the constraint on the production purity, for example for a VS A 02, a demand going from 94% to 91% mol will allow the specific energy of the oxygen produced to be reduced but will have the hidden consequence of advancing the mass transfer zone of atmospheric water. If this operation lasts too long, the unit will no longer regain its initial performance, the exchanged zeolite fraction allowing the O2 / N2 separation once polluted loses most of its properties.

[0112] The second category of incident leading to the advance of the impurity front is much more direct: the pretreatment adsorbent mass becomes partially ineffective in adsorbing the impurity. At the near-ambient temperature at which PSAs operate, this may involve the destruction of the adsorbent structure, deposition on its surface or in its internal porosity, partial clogging of the adsorbent layer leading to distribution defects, or more rarely polymerization of the adsorbed constituent which then permanently fixes itself to the active sites. These problems may come from the impurity itself or from secondary impurities which may not have been mentioned in the study bases for the sizing of the unit. The adsorbent chosen is then not adapted to its function, such as certain standard silica gels in the presence of entrainment of water drops, activated alumina in the presence of acid such as HCl or H2SO4, activated carbon rich in metallic impurities in the presence of reactive constituents, etc. The impurity ends up automatically passing into the main adsorbent mass.

[0113] The last scenario corresponding to pollution due to an impurity present in excess quantity in the feed gas appears to be easily avoidable. Nevertheless, if it is common to analyze production continuously or at least periodically and frequently, it is much less common to do the same with the feed gas whose composition or content ranges for its constituents are part of the study bases provided to the designer. The feed gas generally comes from an upstream unit and its composition very often comes from a thermodynamic equilibrium (washing, distillation column top, condensation, reaction, etc.). The upstream unit has its own regulations and alarms and the vast majority of the time the composition of the feed gas is consistent with what was planned.

[0114] However, it may happen that a very secondary constituent such as the impurity which interests us here has a content which drifts while the main constituents remain at their nominal content: final condensation less efficient in hot periods, catalyst at the end of its life, guard bed beginning to be saturated... It is generally considered that the risk incurred does not justify the installation of an expensive analysis system.

[0115] From these various observations, it appeared that pollution by the impurity of the main adsorbent mass is always accompanied by an advance of the mass transfer zones of the other constituents and that in very many cases, the feed gas included at least one other constituent, among the most adsorbable constituents, which is significantly easier to analyze and much more precisely than the impurity. This is the case, for example, of CO2 with respect to water, or of H2S with respect to heavy hydrocarbons.

[0116] The complementary idea is not to seek to detect the end of the mass transfer zone of this second constituent, which we will call constituent Y from now on, but a point towards the middle of its transfer zone then benefiting from a much higher content and much easier to analyze precisely.

[0117] [Fig.2] [Fig.2] illustrates these 2 points at the basis of the invention. It uses the references already used in [Fig.l]. [Fig.2] represents the position of the frontal masses at the end of the co-current stage for impurity 32 and constituent Y - reference 42 - for nominal operation, with a margin 51 to protect the main adsorbent mass from pollution by the impurity. An analysis of the content of Y at point 50 of the adsorbent mass gives the predicted value y* of this content, let's say 20% of the maximum content standardized to 100% of the constituent Y at the start of its mass transfer zone.

[0118] [Fig.3] [Fig.3] corresponds to the case where, following an incident, the impurity has progressed in the mass of the pretreatment 21 and has reached the main adsorbent mass 22. Still at point 50, the analysis of Y will now give a content y** of 75%.

[0119] If we take the case of atmospheric air separation where, to monitor the progress of humidity towards the main adsorbent mass, the choice of retaining CO2 as constituent Y would lead, for a composition of approximately 400 ppm of CO2 in the air, to respective values of y* and y** of 80 and 300 ppm, which is significantly easier and more precise than trying to measure traces of water. Similar results are obtained for a PSA H2 type treatment unit by measuring H2S downstream of an adsorbent zone intended to stop traces of very heavy hydrocarbons (C9+).

[0120] The interest of proceeding in the manner described above being manifest in a certain number of cases, it is then a question of having a means of choosing the constituent (Y) whose evolution one wishes to follow, as well as the point and the moment of the measurement of its content in the charge gas, of organizing this analysis, of automatically processing the results and of planning preventive actions to avoid pollution of the unit by the impurity.

[0121] The means which logically imposes itself for the choice of the constituent Y, of the position in the adsorbent mass retained to observe the evolution of its content, of the instant t retained to determine y* corresponding to the nominal operation of the unit, possibly also to determine y** corresponding to the critical operation at the pollution limit and to measure the effective content y of the constituent Y during operation - is the use of the calculation means which made it possible to initially size the treatment unit.

[0122] To carry out this sizing, the person skilled in the art currently uses dynamic simulation software for cyclic adsorption processes, which is either a "proprietary" simulator developed internally by the designer of the unit, or one of the commercial software packages. In all cases, the simulation software is based on a division in the space of the adsorber (into slices called meshes) and in the time of the cycle and performs successive sequential calculations until a stable state is found verifying the material and thermal balances as well as all the imposed constraints.

[0123] The latest generation integrates in practice all the parameters that can have an effect on the operation of a PSA unit. The physical properties of the gases (density, viscosity, heat capacity, conductivity, etc.), the basic data of adsorption such as isotherms taking into account the effect of competition between different species, the kinetics taking into account the various constituent phenomena, the thermal effects in the gas, in the adsorbent, at the adsorber wall, the permeability of porous media, the dispersion linked to the circulation of the gas, are, if necessary, determined beforehand in laboratories using suitable equipment. The detailed geometric description of the adsorbers, including dead volumes (volumes accessible to the gas not containing adsorbent), the description of the valves, any machines, regulations and all the stages of the cycle (durations, pressure levels, gas exchanges, etc.) are also integrated into the software so that the results obtained do not require post-processing to approximate reality. Additional input data such as those concerning the feed gas mixture (composition, flow rate, pressure, temperature) then set all the parameters of the software and allow it to carry out a complete calculation.In practice, a parameter is generally left free to allow the simulation to converge on the main constraint (for example, modification of the phase time or the feed gas flow rate to meet the purity required for production). This lengthens the calculation time but avoids restarting the software several times to achieve a specification required for the unit. The detailed description of such software, in particular the increasingly sophisticated convergence methods used, is now found in the literature, often accompanied by the series of equations that the software must solve at each mesh and each time step and will not be repeated here.

[0124] The advantage of such software is that it stores all the parameters at every point of the adsorber (at each mesh) and at each time step. It is therefore possible to plot the respective contents of the impurity and those of other constituents along the adsorbent masses at different times of the cycle without any problem and to select a priori the most interesting points for the claimed process.

[0125] We will return later to the whole point of using such software to be able to extend the monitoring procedure recommended here to the case of variation in the operating conditions of the treatment unit.

[0126] In the absence of suitable simulation software, it is possible to rely on a specially instrumented pilot to determine with more or less precision the profile of the impurity and of the Y constituent in the adsorbent mass at the end of the co-current stage. In particular for sufficiently low operating pressures, for example less than 10 bar abs. and constituents not presenting any particular danger, the use made possible of a gas sampling rod, sliding through the adsorbent mass, makes it possible to obtain sufficient data to have a good assessment of the content profiles.

[0127] It is understood that the important thing is to correctly fix the point of analysis of the constituent Y in order to place it in its mass transfer zone. However, this point being fixed geometrically, there remains a free parameter which is the time chosen to carry out the analysis. It is recommended at the time of unit start-up to validate by measurements on the treatment unit stabilized under nominal conditions, that the respective positions of the impurity and of the Y constituent are those expected. Indeed, an adsorbent used in the pretreatment with an adsorption capacity a little better than its specification or a filling rate of this same adsorbent a little lower than that taken into account can lead to a margin greater or less important than expected with regard to pollution risks and therefore to adapt by a few seconds the time chosen to determine the forecast value of the Y content.As mentioned, the analysis of the impurity content will require specific equipment and specialists, but this is an exceptional measurement which justifies the efforts required on this occasion. Similarly, if it is a new process or the use of an adsorbent which has not yet been tested industrially, it may be decided at the time of design to provide several locations for carrying out the analysis in order to choose the one which will subsequently provide the most precision.

[0128] The maximum advance of the end of the mass transfer zone corresponds to the end of the co-current stages, each of these stages having contributed more or less to advancing the material front of the constituents. However, it is not always desirable to measure the content of the constituent (Y) at this moment in the cycle. Indeed, while the adsorption stage itself is relatively stable in pressure and generally lasts a few tens of seconds, the following co-current stages take place at variable pressure and are generally shorter, for example with two successive equilibrations over the same phase time. This makes the analysis of the gas circulating in the adsorber much more imprecise. For VSA type treatment units, the stages following adsorption are under vacuum, making the withdrawal of a gas sample to the analyzer very complex.For these reasons, the chosen time will generally be towards the end of the adsorption stage unless the pressure cycle of the treatment unit lends itself to carrying out the analysis later in the cycle. This may be the case if there is an elution gas supply stage taking place at a pressure higher than atmospheric pressure, over several tens of seconds and with a small pressure variation, for example of the order of magnitude of one to two bars.

[0129] The point chosen to measure the content of constituent Y must obviously take into account the time chosen to carry out the analysis. Depending on the type of cycle implemented on the treatment unit, the time chosen will correspond either to a fixed duration after the start of a step, for example 25 s after the start of the adsorption step, or will be a function of the duration of a step, this being able to vary during operation, for example the time selected will correspond to the end of the adsorption step or to 90% of its total duration. This essentially depends on the regulations implemented on the treatment unit and the dynamic simulation which can take into account this type of regulation is again a very useful tool to make the best choice of the pair point of the adsorbent mass and time selected to determine the content of the constituent Y. In theory several pairs may be suitable but the person skilled in the art will ensure that for the pair selected the content of Y is between 90% and 10% of the maximum content of this constituent at the start of its mass transfer zone, preferably between 80% and 20% of this content, still preferably between 70 and 30%, thus the position of the front can be known with good precision.Similarly, the point of the adsorbent mass chosen to carry out the analysis is preferably located: - either between 50 and 100% of the pretreatment adsorbent mass, 100% corresponding to the interface with the main adsorbent mass, preferably between 60 and 95%, still preferably between 70 and 90%, - or in the first 20% of the main adsorbent mass, 0% corresponding to the interface with the pretreatment and 100% to the entire main adsorbent mass, preferably in the first 10% of the main mass.

[0130] This means that the measurement of the content of (Y) is carried out in the area of interest, namely near the interface between the pretreatment adsorbent mass and the main adsorbent zone.

[0131] This having been fixed, it is then theoretically possible to analyze at each cycle i, i+1, i+2, the content of (Y) is yi, yi+1, yi+2. Depending on the analysis procedure and the processing of the results of successive analyses, it may be more advantageous to determine the content of (Y) only periodically, for example every few hours, if the operation of the unit is stable and on the contrary to carry out the maximum number of analyses in the event of changes in the latter. For simplicity, it is assumed that an analysis is available at every cycle.

[0132] The very principle of the invention then consists in comparing the value of the content yi+1 obtained in cycle i+1 with those obtained previously yi, yi-1, yi-2, etc... and, if as is very desirable, the corresponding forecast value of (Y) has been determined beforehand by one of the means described (dynamic process simulation, analysis on a pilot, on the processing unit after start-up), in comparing yi+1 with y*, y* being this forecast content.

[0133] To take full advantage of the data thus collected, it is advisable to record and process them automatically. The series of measurements yi makes it possible to establish the corresponding trend curve. With several measurements per day or per hour, small random variations are eliminated and the slope of the curve will give an accurate picture of the operating status of the unit with respect to the evolution of the constituent (Y): stability, regression or progression. This slope, for example translated into ppm / cycle, can constitute the parameter retained to monitor pollution risks with a tolerance threshold which can be for example a percentage increase in the content y compared to the previous stabilized content, the duration since the start of the progression or a maximum speed of evolution. Obviously, several thresholds can have been planned in parallel corresponding to different types of evolution which can also lead to pollution (sudden and rapid or progressive but continuous).

[0134] The preliminary determination of the forecast value y* of the content of (Y) in nominal operation will allow a more precise judgment of the risk of pollution of the main adsorbent mass by comparison of yi+1 with y*. The fact of having also determined a second forecast value y** corresponding to the actual start of pollution then allows a precise view of the position of the impurity with respect to the interface with the main adsorbent mass and the provision of automatic precautionary measures to safeguard the latter.

[0135] To illustrate this, we take the example of monitoring at a fixed point and time the CO2 content of a feed gas, this constituent having then been retained as constituent (Y) to protect against pollution by humidity of an exchanged zeolite very sensitive to water. In nominal operation, the treatment unit being stabilized, the forecast value is 140 ppm of CO2 (y*= 140) to have the margin deemed necessary with regard to pollution. The second forecast value y** corresponding to the start of pollution is 290 ppm of CO2. Compared to the maximum CO2 content at the start of its mass transfer zone (360 ppm), the characteristic contents are located respectively at approximately 40% and 80% of the material front, which allows for good precision because it is in a substantially linear zone and is sufficiently far from the ends of the transfer zone (0 and 100%) which no longer provide useful information on the position of the front.From successive measurements of the CO2 content y, all information concerning the current evolution of the risk can be obtained based on the deviations (y*-yi) and (y**-yi) and the forecast evolution of the pollution risk by introducing the rate of variation of the measured CO2 content, Delta y / Tc expressed for example in ppm per minute, with Te the duration of a cycle. It is then possible to define thresholds to generate alarms and, if necessary, to take precautionary measures by modifying the current cycle of the processing unit. This could, for example, be a first alarm if the system carrying out the data processing determines that the content y will reach 190 ppm in the next quarter of an hour, then a second alarm if the forecast becomes 240 ppm. If this second alarm is not acknowledged and / or if the CO2 front continues to advance, the control system of the unit. treatment can switch to a backup program which consists of modifying the current pressure cycle in order to push back the material fronts of the constituents, in particular that of the impurity.

[0136] In practice, each type of PSA has several means at its disposal to push back the impurity front, even if it means losing a little in terms of unit performance, in terms of yield and / or production. Thus, it is for example possible to limit the advance of the mass transfer zones by reducing the gas volumes passing through the adsorbent mass in co-current and / or to increase the regeneration power by increasing the gas volumes in counter-current. Among these means, we can cite, in a non-exhaustive manner: - limit the number of pressure balancings between adsorbers, for example by implementing partial balancing instead of complete balancing, - reduce the cycle time while maintaining the charge gas flow rate (and the value of the internal pressures in the cycle), - reduce the flow rate of the feed gas while maintaining the cycle time, - increase the volume of the elution gas by increasing its flow rate, by lengthening the corresponding step or by lowering the pressure (for example in the case of using a vacuum pump during this step).

[0137] If the speed of evolution of the material fronts allows it, these modifications can be made in successive steps until the desired result is obtained. In the case where these changes of set points are not sufficiently effective to stop the progression of the constituent (Y) and therefore a priori of the impurity, it could have been programmed for example as a possible fallback a backup cycle allowing to continue to produce a fraction of the required production at the specified purity. As a last resort, the management system of the unit can trigger the shutdown of the latter.

[0138] Until now, it has been assumed that the unit had been sized under the most restrictive conditions with regard to obtaining the required performance and also the risks of pollution; this is the so-called nominal operation for which the content of the constituent (Y) at a point of the adsorbent mass and at a chosen time was linked to the admissible advance of the impurity and possibly to a second content corresponding to the start of pollution. The forecast values of (Y), y* and y** were respectively called and actions were planned by comparing the actual content y with these two values.

[0139] If the unit is likely to operate over time with different operating conditions a, b, c..., a corresponding for example to the nominal dimensioning step mentioned up to now, and b and c to frequent steps linked to different performance specifications, or even to changes in parameters external to the unit such as the recovery pressure of the residual, the content of a constituent of the feed gas or the inlet temperature in the unit, it is possible to adapt the forecast values y* and y** to these new conditions, as well as the tolerance thresholds associated with them. This again assumes the use of the simulation software for dynamic adsorption processes already widely mentioned. In practice, this amounts to using forecast values depending on the operating conditions a, b, c..., namely y*(a, b, c...) and y**(a, b, c...). The module for processing the actual measured yi contents then stores in memory the characteristics a, b, c... corresponding to the measurements. A PLC or computer, integrated or not in the control-command system, can easily manage several parameters, or even make various interpolations from the moment it has a catalog of forecast values and receives information on the characteristics (a, b, c...) of the current operation.

[0140] It should however be noted that the claimed method is only applicable if the characteristics of the feed gas do not change constantly or too significantly. This is very generally the case, because a PSA type unit needs to operate in a stable manner in order to be efficient or even just to be able to comply with production constraints such as, for example, product purity. The case where there are 2 - or more - different steps originally planned, corresponding to different production demands or to changes in the steps of downstream units, can, however, fall within the scope of this document, each of the new conditions being able to last a few weeks a priori.

[0141] If we actually have a simulator that has made it possible to scan different operating points, to derive sufficiently precise forecast values to diagnose the position of the impurity in the adsorbent mass as a function of the measurement of the content y of the constituent Y and this using an automaton, the last point that should be ensured is that we have a reliable and precise system for measuring the successive contents yi over time. The constituent Y was chosen during the design of the unit so that this is the case a priori, but the fact that the treatment unit is a PSA can nevertheless complicate things because of short-term stages, the rapid evolution of the mass transfer zones in the adsorbent, rapid pressure variations and the possibility that certain stages are under vacuum, thus complicating the sampling of the gas circulating in the adsorber.

[0142] There are different options for carrying out these measurements depending on the characteristics of the pressure cycle of the processing unit. The purpose of what follows is not to deal exhaustively with the different possibilities of the analysis chains but to highlight some problems and the solutions in principle. The person skilled in the art will be best placed to make the best choice depending on the different constraints.

[0143] A first method consists of using an analysis probe located in situ in the adsorbent mass. This technology theoretically allows continuous measurement of the concentration of a gas (H2O, NH3, CO2, H2S, CO, CH4, 02...) in a gas mixture. Before adopting this method, it is advisable to ensure the scale of measurable contents, the possible interactions between constituents of the gas mixture on the response of the probe and any possible drifts.

[0144] The second method consists of taking a gas sample at the selected point of the adsorbent mass and analyzing the sample using a conventional analyzer adapted to the constituent Y whose content is to be known.

[0145] The sampling itself can be one-off, semi-continuous or continuous.

[0146] A particular system which is the subject of an application currently under examination has been developed to reproducibly take a gas sample in a very short time, for example of the order of a second, while continuously supplying an analyzer with this gas. The measured content then corresponds directly to the desired value y.

[0147] More generally, the sample is taken over a fraction of the cycle, for example during the adsorption step or part of this step. A make-up gas, for example nitrogen or argon, can then be introduced into the circuit supplying the analyzer to ensure a continuous gas flow towards it. The content then taken into account must then correspond to the selected time. It will be noted that in this case, there will be at least part of the passage of the mass transfer zone of the constituent Y and that the contents at times preceding or following the selected time can be extracted from it if desired, which can confirm the stability or the evolution of said constituent during the cycles.

[0148] In the case where a gas sample is taken and then sent to an analyzer, the results of which are transmitted to an automaton which uses them as described above, it may be very interesting, for a relatively low investment, to equip several adsorbers and preferably all the adsorbers with a sampling point placed at the same point of the adsorbent mass and to have each follow the same procedure. In addition to the fact that a plurality of measurements makes it possible to confirm the stability or the evolution of the operation of the treatment unit, such a device makes it possible to highlight any asymmetry between adsorbers and also to judge whether or not this asymmetry is stable over time. As long as a gas flow rate can be taken for each of the adsorbers over the duration of a phase, the analyzer can then be continuously supplied without having to use a make-up gas as mentioned above.The extraction of the different contents y of the constituent Y at the chosen time, each corresponding to a defined adsorber, does not pose any particular problem.

[0149] We have seen that the purpose of the procedure proposed in this document was to ensure that a constituent called an impurity would not advance during the cycles in the adsorbent mass of a PSA type treatment unit beyond a certain limit so as not to interfere negatively with the adsorption of the other constituents on the downstream part of this adsorbent mass. There may be two or more impurities that are totally undesirable in the main adsorbent mass. Generally, one of them presents the highest pollution risk for the main adsorbent and will become the dimensioning impurity. A person skilled in the art will be able to determine, depending on the operating conditions, the contents of the various constituents of the feed gas, and the adsorbents retained, this constituent which becomes, for the anti-pollution procedure described here, "the impurity".

[0150] The impurity to be stopped in the pretreatment adsorbent mass will generally be part of the group comprising: H2O, NH3, alcohols, ketones, aldehydes, ethers, esters, acids and constituents giving acid groups in the presence of moisture, amines, mercaptans, halogenated hydrocarbons, siloxanes, unsaturated hydrocarbons, C6 and C6+ hydrocarbons, including BTEX, terpenes, more generally VOCs. In practice, to determine the impurity to be monitored as the most problematic for the separation unit, it is also necessary to take into account the main constituents to be fractionated because it is on these constituents that the choice of adsorbents of the main adsorbent mass will depend and ultimately the compatibility or not of these adsorbents with the impurity.These main constituents generally belong to a group of constituents including CO2, H2S, hydrocarbons Cl to C7, COS, carbon monoxide, nitrogen, oxygen, argon, hydrogen, neon, krypton, xenon and helium. It is difficult to be exhaustive in these lists of constituents given the number of applications using a PSA, and this in particular for impurities which, at the ppm level, can be found by the dozen in combustion, chemical reaction or fermentation gases.

[0151] The pretreatment adsorbent mass may comprise one or more adsorbents in successive layers or as a mixture. At least one of these adsorbents will generally be different from the adsorbents of the main adsorbent mass. By different, we mean both of a distinct nature (activated alumina, activated carbon, silica gel, zeolite) and of different physical or chemical characteristics (standard or doped alumina, silica gel of more or less high porosity, more or less pure carbon, zeolites A, X, Y, etc.). The adsorbent mass may also be of the same nature as the first layer, or even as the single layer of the main adsorbent mass. This is the case where a fraction of the adsorbent mass is allocated to the impurity, knowing that the presence of the latter eliminates virtually any possibility of co-adsorption to the impurities. other constituents of the feed gas. It is then necessary to ensure that during operation, the impurity will remain in the zone allocated to it in the design of the unit. This situation essentially arises in two cases. In the first, it turns out that the adsorbent chosen to effectively stop the impurity is also the best for stopping a main constituent, generally the most adsorbable of them. The second case is not directly linked to the nature of the constituents but to the sizing of the adsorber, in particular to its geometry. For a cylindrical adsorber with a vertical axis of large diameter, for example 5 m, too thin thicknesses of adsorbent layers pose problems of horizontality of the interfaces between layers and distribution and a single layer will generally be preferred to two adjacent layers even if the use of a second adsorbent could be a little more effective.The same choice arises when the adsorber is of the radial type: each additional layer requires an additional intermediate grid to separate and hold the adsorbents in place. The single layer generally proves to be the best solution.

[0152] Whatever the case, different adsorbents or the same adsorbent, the abnormal presence of the impurity in the main adsorbent mass causes the reduction of the adsorption capacity of said main adsorbent mass and / or its selectivity and / or its kinetics, said reductions being able to be caused by a chemical degradation of the internal structure of the adsorbent, by deposits on its surface, by a reduction in the internal porosity, by a preferential adsorption of the impurity occupying the adsorption sites normally allocated to the other constituents of the feed gas.

[0153] In particular, the preferential adsorption of the impurity in the main adsorbent mass is not always completely reversible during regeneration and then causes a definitive reduction in the separation performance of the polluted zone, even if the defect which caused the incursion of the impurity has been corrected.

[0154] It is then recalled that the main adsorbent mass has been defined as the adsorbent whose full separation capacity is to be preserved by protecting it from pollution, particularly irreversible pollution, by an impurity and the pretreatment adsorbent mass as the adsorbent installed for this protection purpose, said pretreatment adsorbent mass being able to include a safety (10 or 15% of its mass for example), this safety mass normally only being in contact with the impurity exceptionally and being able, depending on its nature, to begin to fractionate the main constituents, possibly in a relatively inefficient manner. It cannot be excluded either that some of the most adsorbable compounds are slightly co-adsorbed at the same time as the impurity on the pretreatment adsorbent mass. This is for example the case of CO2 on wet alumina.

[0155] Generally, the pretreatment adsorbent mass comprises, depending on the nature of the impurity, activated alumina, standard or doped, silica gel, activated carbon or a zeolite of type 3A, NaX or Y, in a single layer, in a mixture or in successive layers. In order to be very selective with respect to molecules of smaller dimensions (H20, NH3), a zeolite 3A strongly exchanged with potassium (deeply K exchanged 3A) may be used which in practice only adsorbs these 2 constituents.

[0156] As for it, the main adsorbent mass may comprise, in a non-limiting manner, a zeolite, particularly a zeolite of type LSX, X, A, Y, said zeolite being preferentially exchanged with at least one cation such as K, Ca, Li, Ba, Sr, Ag..., or a CMS (Carbon Molecular Sieve) or a synthetic adsorbent specially calibrated for a kinetic type separation, a metal-organic structure (MOF), a resin, a polymer-based material, or more generally a porous structure preferably containing a substance capable of reacting reversibly with certain molecules, a substance such as a physical solvent or a metal complexing agent for example, a silica gel or an activated carbon, these adsorbents being able to be in a single layer, in a mixture or in successive layers.

[0157] The adsorbent masses can be made up of particles or monoliths. The shape of the particles (balls, rods, pellets, etc.) and their main dimension, generally from 0.5 to 8 mm, do not modify the idea of the invention.

[0158] One of the most delicate points that arises for the person skilled in the art is to choose a constituent (Y) whose content evolution at a selected point of the adsorbent mass will provide information on the stability or drift of the mass transfer zone of the impurity. This choice will be relatively easy for complex feed gases, i.e. having more than 5 or 6 constituents and more random in the opposite case, especially since the main constituents will be very different at the adsorption level of the impurity. For example, for a water-saturated H2 / N2 feed gas, measuring the N2 content will not provide any usable information. In such a case, a solution will be, for example, to oversize the pretreatment adsorbent mass and to install near the interface a continuous sampling of a small flow of purified feed gas to a humidity trap which will be periodically checked.This provides much less information than the procedure presented here but can confirm by periodic examination of the trap that the gas circulating in the adsorber was a priori dry until then. On the other hand, the detection of water in the trap will mean that the beginning of pollution has started for some time, for example a few weeks. This is retrospective information. We can protect ourselves better by providing an additional margin between the sampling point and the interface with the main adsorbent mass.

[0159] Fortunately, there are impurity / component (Y) pairs which work particularly well, the impurities being essentially those belonging to the group defined above and the constituent (Y) a C1 to C7 hydrocarbon, CO2 or H2S. These constituents are classic in many feed gases treated by PSA, they are often present in small quantities, from a few hundred ppm to a few mol% maximum, relatively well adsorbable but less so than the impurities, which means that their mass transfer zone is located downstream of that of the impurity without being far from it.

[0160] It will thus be possible to easily apply the principle of the invention to PSA 02 (PSA, MPSA, VSA), the impurity being atmospheric humidity and the constituent (Y) CO2, to PSA H2 or CH4, the impurity being water, heavy hydrocarbons, NH3, various molecules comprising more than 6 atoms, polar or with bonds favoring their adsorption... the constituent Y chosen being H2S, CO2 or intermediate hydrocarbons (C3 / C7).

[0161] Two examples of application of the claimed method are detailed below: the first corresponds to its most direct application, the second uses the result of the analysis of the progression of the constituent Y and therefore of the impurity in a more specific manner.

[0162] Finally, a potential application of the method is described in a third and final case.

[0163] The production of oxygen from atmospheric air is a typical case of possible application of the invention. A pilot unit equipped with H2O and CO2 analysis means particularly suited to use on short-term cycles was used in order to validate the claimed process.

[0164] Generally speaking, the PSA 02, VSA 02 (Vacuum Swing Adsorption) and VPSA (Vacuum-Pres sure Swing Adsorption) 02 units are air gas separation units differentiated by the pressure levels of their respective cycle as indicated in the introduction.

[0165] The production of gaseous oxygen reaches a purity of the order of 90% to 93%, or even more rarely 94% mole and the usual production range of this type of device varies from approximately It / d to 400t / d with a very large number of units with a capacity of between 30t / d and 150t / d. These processes find applications in fields such as water purification, glass manufacturing, paper pulp processing, steelmaking, fish farming, medical, etc. The most numerous units today are of the VPSA type with a high pressure of approximately 1.5 bar abs (abbreviation for absolute bar) and a low pressure around 0.5 bar abs.

[0166] The adsorbers of the VSA 02 units are composed of several layers of adsorbent. The so-called pretreatment layer allows the stopping of at least part of the pollutants from air and prevents pollution of the main adsorbent mass dedicated to N2 / O2 separation. Currently, this first layer can be composed of activated alumina, doped activated alumina, zeolite 13X, or even silica gel in the case of acid pollutants or a superposition or mixture of these.

[0167] On an industrial scale, it is appropriate to distinguish the effects of an ingress of water or CO2 on the main adsorbent mass which comprises one or two zeolites at least partially exchanged with calcium and / or lithium. Pollution by humidity has very significant long-term consequences on the production and operation of VS A units. In fact, water is very strongly adsorbed on the zeolite and considerably degrades its performance, without it being possible to desorb it significantly under the conditions of an industrial VS A 02 cycle. In the long term, this pollution which occurs at each cycle causes an accumulation of water in the zeolite, and it becomes necessary to change this adsorbent to regain the initial performance of the unit.The CO2 that can reach the main adsorbent during a cycle is then completely evacuated, which means that on the one hand, there is no accumulation over time and on the other hand, the impact on performance, due to the co-adsorption phenomenon, is limited. The presence of CO2 therefore does not represent a particular risk. Each VS A or VPSA 02 designer has his own pretreatment sizing procedure which differs essentially by the nature of the adsorbent, the margin taken on the pretreatment adsorbent mass necessary to stop all the water and on the conditions chosen to carry out this sizing: extreme conditions of relative humidity and atmospheric temperature, average annual conditions, or for example intermediate conditions such as the average over a week of extreme conditions.. .Depending on the assumptions and the sizing adopted, the CO2 penetrates more or less into the main adsorbent mass, but is mostly adsorbed on the pretreatment, in particular partly in the water mass transfer zone and partly in the safety water-free zone.

[0168] Since the pretreatment adsorbent is not or only slightly effective for O2 / N2 separation, a significant margin taken on the sizing means that the air occupying this zone, essentially the dead volumes that constitute the intergranular spaces and a fraction of the internal porosity of the adsorbent, will be compressed then pumped under vacuum for nothing. This has a cost in investment and energy consumed. In order to be competitive, it is necessary to make the right compromise between risk and performance. It is understood that eliminating the risk or at least limiting it was one of the main objectives of the designers. Quite logically, efforts were focused on detecting water towards the pretreatment / main adsorbent interface, preferably a little upstream to be able to act in time. As already discussed, the various devices for determining air humidity, devices that could be introduced locally in the adsorber, do not meet the needs well for various reasons: need for calibration, response time, precision, result depending on possible contacts with the adsorbent particles. Theoretically, the safest way would be to use a system for sampling the circulating gas at a suitable position, preferably in the pretreatment layer or at the interface between the pretreatment layer and the main adsorbent and to analyze its water content. However, it is well known that water analysis is very delicate even for specialists, particularly if we are looking for low contents, which also vary during the cycle. The mere installation of heated lines and a thermostatically controlled enclosure does not then guarantee the reliability of the results.Despite an appropriate choice of materials, there are, at least residually, adsorption / desorption effects of water molecules at the wall which can prevent, for example, the detection of a pollution peak. Cells measuring water content also experience drift phenomena generally indicating lower contents than reality. Finally, the intrinsic response time of these cells may be insufficient in the case of PSAs with very short stages, in which case the measured value corresponds more to an average than to an instantaneous value. It should be noted again that the measurement of water content can be a one-off analysis at the time of start-up to check the satisfactory operation of the pretreatment (sizing chosen, state of the adsorbent, etc.)) and / or periodic checks at intervals, for example every 6 or 12 months, but very difficult to envisage continuously or semi-continuously, such as once a day or a week. It is therefore unlikely to use such analyses to automatically or even manually adapt the operating characteristics of the unit based on the measured values obtained. Such adaptations, on the contrary, require continuous or quasi-continuous monitoring, i.e. at least once a day for stability monitoring and every few cycles during periods of changing operating conditions or when a risk of pollution has been detected.The process recommended here for impurity management was therefore tested on a pilot plant implementing a VPSA O2 cycle, comprising 2 cylindrical adsorbers with a vertical axis operating with a phase shift, each of the adsorbers comprising a pretreatment layer composed solely of zeolite 13X followed by a lithium-exchanged LSX zeolite conventionally used for O2 / N2 separation. Water is then the impurity that we want to prevent from reaching the lithium-exchanged zeolite and CO2 is chosen as the reference constituent Y. A water injection system was installed in order to be able to scan a wide range of water content in the feed air. The CO2 content is that found naturally in atmospheric air, in the order of 360 to 400 ppm. A gas sampling and water and CO2 analysis system was . developed especially for these tests in order to obtain the best possible results. The position of the sample can also be modified on this pilot, thus allowing analyses to be carried out throughout the pretreatment. Consisting of a mobile sampling rod in the adsorbent mass associated with very rapid opening and closing valves and a temporary storage volume for the sampled gas, this system, which will not be described in more detail in this document, makes it possible to obtain curves highlighting the parallel evolution of the water and CO2 contents along the adsorber at selected times in the cycle, curves which cannot in practice be obtained on an industrial unit. It is understood that if we want to repeat these measurements for several experimental conditions of pressure, temperature and relative humidity, this corresponds to a test campaign lasting several weeks.

[0169] These tests first confirmed that the simulation software gave a good representation of the stable state of the VPSA. It was verified that the CO2 front tends to establish itself more quickly than the water front but that it only reaches its final form when the water front is stabilized. A phenomenon of co-adsorption of CO2 is observed in the pretreatment zeolite X, essentially in the water front zone. The initial part of the CO2 front (spatially speaking, i.e. on the air supply side) is therefore in competition with the water front. The end of the CO2 front (still spatially speaking) is established upstream of the water front, in the dry zone, partly in the pretreatment zeolite X and partly in the initial zone of the LiLSX type zeolite.

[0170] The fact that the adsorption of water displaces the CO2 and therefore advances the CO2 front makes it possible to establish a link between the two mass transfer zones. For given operating conditions, in particular temperature and relative humidity of the air that will be chosen as a sizing basis depending on the site where the unit will be installed, we will determine the mass of zeolite X necessary to stop all the water, the possible margin to be taken for example 15% of this necessary mass, the CO2 content at the point of the 13X and at the moment of the cycle retained, for example 180 ppm, as well as the CO2 content corresponding to the start of the pollution of the LiLSX, for example 230 ppm. The operating conditions retained do not generally correspond to the wettest day but for example to the average of the wettest week. There are Atlases giving all the local information on the climate in almost all countries.

[0171] Concerning the industrial unit carried out according to this principle, in the absence of an incident, and as long as the water content at the adsorber inlet is lower than the maximum content taken as reference, which is the case the vast majority of the time, the measured content y of CO2 will be lower than y* (here 180 ppm). The industrial unit will thus operate in a verifiable manner without any particular risk of pollution. Conversely, in the presence of an event leading to an advance of the water front in the pretreatment (higher quantity of water entering the adsorber following a long hot and humid period, drop in regeneration power following a desired or imposed modification of the cycle... the CO2 front will advance jointly with that of the water. The measured CO2 content at the sampling point will then increase to possibly reach the value y* of 180 ppm and trigger the alert phase then without effect on the evolution, automatically, a series of safeguard actions which may end with the shutdown of the unit if the measured content reaches 230 ppm. One of the first actions which can be taken automatically is to lower the low pressure, of the order of 0.43 bar abs in nominal operation, in steps of 15 mbar every n cycles, for example every 100 cycles, down to a minimum pressure agreed upon during the design.This modification of the cycle is normally effective in ensuring better regeneration of the adsorbent masses and only has an effect on energy consumption, which will increase slightly. It should be noted that this type of unit is often monitored remotely, sometimes only on partial standby. The fact of implementing this anti-pollution system, which can take relatively simple precautionary measures on its own, in the absence of an operator nearby, such as modifying a set point (low cycle pressure), or even triggering the unit shutdown procedure, is a safety feature that avoids much more cumbersome interventions such as changing the adsorbent or a significant economic penalty if the nominal O2 flow rate can no longer be reached by the unit.

[0172] The second example concerns a PSA H2 type treatment unit supplied by a feed gas from a coke oven, this gas having undergone a minimum of treatment between the outlet of the oven and the PSA; As a result, besides H2, CH4, CO, CO2, N2 and O2, the gas contains traces of numerous constituents such as acids, ammonia, HCN, H2S, benzol, naphthalene and heavier hydrocarbons (tars or tar vapor). The adsorbent mass comprises around 80% by volume of carbon and 20% of zeolite, generally 5A or 13X. This process has a particularity: the heavier elements of the feed gas are not completely desorbed and accumulate slowly in the first fraction of the activated carbon. It is possible for an equilibrium to be reached but only after a very long time and at the cost of a very substantial volume of carbon.It is therefore planned to use special adsorbers for which it is possible, by means of a set of sloping grids, to quickly replace the initial fraction of activated carbon, said replacement requiring only a short stop. The activated carbon must be replaced before the tars pass into the upper layer and reduce its efficiency in adsorbing the main constituents. The duration of an activated carbon charge is unpredictable because the coking plant processes coals of various origins and their tars content can vary by a factor of 3 or 4. For safety, the first layer of . The activated carbon was initially changed every 3 months and the carbon was more or less loaded with tars depending on the period. Premature changes represented a fairly significant financial loss. The coke oven gas was treated before being sent to the PSA in a desulfurization unit designed to remove only a fraction of the sulfur compounds and as a result the PSA charge gas permanently contained an H2S content of around a hundred ppm. It appeared that with a new charge of activated carbon, the residual H2S content at the interface with the main mass of coal was low, around ten ppm, and that it gradually increased over time. After a series of charge changes and verification of the condition of the activated carbon with respect to tars, it appeared that a change was necessary when the H2S content was around 60 / 70 ppm.Monitoring the H2S content, the Y compound of the proposed process, at the end of the adsorption stage at the interface between the pretreatment activated carbon and the main activated carbon thus made it possible to space out the emptying and refilling of the adsorbers by an average factor of 2 to 3. The frequency of H2S analyses varies depending on the operating time after the installation of a new load and the last measured content. Each adsorber is equipped with a gas sampling port and data processing does not require any special equipment. The decision to change the lower part of the activated carbon was taken when the average H2S content over a week was around 65 ppm. The load change is not automatic but carried out by the operator.

[0173] Although very different from the previous example, in both cases it is an application of the principle claimed in this document. In the case of VPSA 02, even in the absence of a forecast value for CO2, cycle-by-cycle analyses nevertheless make it possible to judge the stability of the unit. In the case of PSA H2, it is the knowledge of this forecast value, 60 / 70 ppm of H2S, which makes it possible to avoid pollution while spacing out the renewal of adsorbent.

[0174] It will be noted that the invention can also find an application in cases where one wishes to monitor the advance of the mass front of one or more secondary constituents that are difficult to analyze directly because they are highly adsorbable. As in the previous examples, the content at a point of the adsorbent of a less adsorbable constituent is then measured since the position of its mass transfer zone is impacted by the presence of the secondary constituent(s) that one wishes to monitor. The potential application described here is based on the different effects that the regeneration power, more particularly its PF ratio, has on the constituent / adsorbent pairs. It is recalled that the regeneration power is what causes the mass transfer zone of the constituents to retreat during the desorption steps. The more this power is locally greater than 1.0, the better the adsorbent is regenerated during the cycle and the smaller the quantity that is required to carry out the separation. Due to the shape of the isotherm, the adsorption capacity, the kinetics, the regeneration of a constituent / adsorbent pair can be difficult and lead either to installing relatively high adsorbent masses as pretreatment, which impacts the overall performance of the unit in particular because of the resulting dead volumes, or to adopting a high regeneration power which allows the adsorbent mass to be reduced but to the detriment of efficiency or energy consumption. It turns out that for constituent / adsorbent pairs that are difficult to regenerate by simple pressure effect, in particular when the constituent content is low, of the order of a percent for example, the state of equilibrium is long to reach and this in particular when the regeneration power is close to its limit value of 1.0, for example 1.1.For higher regenerative power, the steady state is reached much faster. It is recalled that the state of equilibrium is called the moment when the material fronts at the end of co-current and counter-current stages no longer move from one cycle to the next and where, consequently, there is neither accumulation nor loss of adsorbate. It is then possible to size the unit by taking into account for the cycle only the optimization of the main separation corresponding to a regeneration power of the order of 1.10 or 1.15 and by installing for the pretreatment adsorbent mass only a fraction, for example 50 or 60% of that theoretically necessary for equilibrium for such a regeneration power ratio. This assumes that periodically, the performance of the unit is degraded by temporarily adopting a ratio of 1.3 for example to push back the impurity mass transfer zone.The economic balance sheet can actually be favorable to this mode of operation, and this is very likely if the treatment unit is regularly in reduced operation, at night or at the weekend. We are then in the presence of unstabilized operation for the impurity whose front advances for example for 18 hours and retreats for 6 hours. It is understood that being able to verify that we are then operating without risk of pollution by following the frontal mass of a Y constituent that is easy to analyze becomes almost necessary.

[0175] In summary, [Fig.4] [Fig.4] then schematically represents the succession of steps characterizing the invention (in solid line rectangles) as well as any additional steps (in dotted lines) providing additional information or automatically generating conservative actions with the aim of preserving the integrity of the adsorbent mass.

[0176] Step E0 is that of sizing the treatment unit. The designer defines a nominal operation, generally corresponding to 100% of the flow rate of the feed gas with the most restrictive conditions on the operating characteristics (high pressure, low pressure, temperature, composition) if the latter are provided in the form of ranges or with the characteristics set out in the study bases. It is at this stage that the main and pretreatment adsorbent masses are determined, among other things. In particular, an impurity, called the impurity, led to defining the nature and volume of the pretreatment adsorbent mass necessary for it to remain confined during operation.

[0177] This step E0 is common to all PSA type separation units and is not strictly speaking part of the anti-pollution procedure.

[0178] Step E1 consists of choosing a reference constituent called constituent Y, easily analyzable given its nature and its content in the feed gas, which is among the most adsorbable compounds, excluding impurities, so that its adsorption zone is located downstream but close to the adsorption zone of the impurity so that there is a direct effect of the latter on the adsorption of constituent Y.

[0179] Step E2 consists of determining a point of the adsorbent mass and an instant of the cycle corresponding to the passage of the mass transfer zone of the constituent Y which is representative of the operation of the cycle in progress, the instant retained being for example beyond half the duration of the adsorption step, and the point of the adsorbent mass beyond half the pretreatment adsorbent mass and this for the nominal operation defined in step E0. It will be noted that the use of simulation software representative of the treatment unit to define the best position in the adsorber / observation instant pairs constitutes a very valuable aid if not essential.

[0180] Step E3 is a generally optional step, nevertheless strongly recommended when the software mentioned above or other means of calculation or testing are available. It consists of determining the forecast value y* of the content of the constituent Y in the case of the nominal operation at the point of the adsorbent mass and at the instant retained. The nominal operation corresponding by definition to the dimensioning operation for the unit, one should therefore not, except in the event of an incident, measure a content y greater than y*. In the usual case, where during the design a margin has been provided on the pretreatment adsorbent mass, it is possible, in particular by using the simulation software again, to define a second forecast value y** corresponding to an abnormal operation for which one begins to pollute the main adsorbent mass.

[0181] It will be noted that in the case of the second example, steps E1, E2 and E3 took place empirically, several months after the PSA unit was put into service.

[0182] Step E4 consists, with the treatment unit in operation, of measuring the content y of the constituent Y at the point of the adsorbent mass and at the instant of the cycle retained, and this with an agreed frequency, possibly as a function of the measured content.

[0183] The next step E5 consists of comparing the last value obtained with the previously measured values, possibly with the forecast value y* or with the forecast value y**.

[0184] Step E6 is the formatting of an easily exploitable parameter, the result of the comparison carried out in step E5. This step can be carried out manually but will generally be done by an automaton. It consists, for example, of establishing the trend curve of the evolution of the successive contents y and of deriving from it as a parameter the slope of evolution in ppm per hour or by comparison with the forecast values, the evolution of the safety margin y*-y and y**-y with a forecast for future cycles.

[0185] Step E7 corresponds to sending an alarm message informing of a risk of pollution, possibly with an estimate of the time remaining before this event, if one of the parameters determined in the previous step exceeds an agreed threshold.

[0186] Step E8 is also optional. It will depend on the type of monitoring of the treatment unit and will generally be provided at least for units operated remotely with only personnel on call part of the time. It consists of automatically triggering preventive actions intended to preserve the performance of the main adsorbent mass from pollution by the impurity. This will initially involve modifying one or more set points in the unit's control system, for example when the value of the content y becomes higher than the forecast value y* or, if the most critical tolerance threshold (y=y**) is reached, initiating the unit shutdown procedure.

[0187] As indicated, the measurement step E4 continues during operation of the unit. Depending on the analysis system implemented, the measurements may be continuous, i.e. cycle after cycle, even if they are only processed with an agreed frequency or only effective every X cycles, X then depending on the values y of Y measured.

Claims

[Claim 1] Claims A method for managing an impurity in a process for separating a feed gas (10) by pressure swing adsorption in which a treatment unit is supplied with the feed gas (10) and produces a first gaseous fraction enriched in the most adsorbable constituents (11) and a second gaseous fraction enriched in the least adsorbable constituents, the treatment unit comprising a plurality of adsorbers (1), each of said adsorbers (1) containing an adsorbent mass (2) composed of a pretreatment adsorbent mass (21) placed upstream of a main adsorbent mass (22), the pretreatment adsorbent mass (21) and the main adsorbent mass (22) being chosen and sized such that: - the pretreatment adsorbent mass (21) stops more than 99%, preferably all of an impurity (30) contained in the feed gas (10), thus producing a purified feed gas, - the main adsorbent mass (22) fractionates the purified feed gas into the gaseous fraction enriched in the most adsorbable constituents (11) which adsorb on the main adsorbent mass (22) and the gaseous fraction enriched in the least adsorbable constituents, said separation method being characterized by the implementation of an anti-pollution procedure intended to detect an abnormal progression of the impurity (30) in the pretreatment adsorbent mass (21), in particular towards the main adsorbent mass (22), and to avoid pollution of the main adsorbent mass (22) by the impurity (30), said procedure comprising the following steps: 1- Choose one of the constituents (Y), other than the impurity (30), from among the most adsorbable constituents (11) of the purified feed gas, 2- Determine a given instant of the pressure cycle and a point (50) of the adsorbent mass (2) included in a mass transfer zone (32, 34, 42) for this constituent (Y) at the given instant of the cycle, the mass transfer zone (32, 34, 42) being located between a zone of the saturated adsorbent mass (31) in this constituent (Y) at the given instant of the pressure cycle and a zone of the adsorbent mass which, at the given instant of the pressure cycle, has not yet been in contact with this constituent (Y), 3- Determine, if necessary, for this constituent (Y), a forecast value of its content at point (50) of the adsorbent mass and at the instant of the cycle retained in step 2, 4- Measure, with the treatment unit in operation, at point (50) of the adsorbent mass (2) and at the time retained in step 2 and on at least one of the adsorbers (1) of the unit, a content of the constituent (Y), 5- Compare this measured content to one or more contents of this constituent (Y) previously measured at said point (50) of the adsorbent mass (2) and / or compare this content to the forecast value determined in step 3, 6- Determine a parameter linked to the measured content based on the comparison in step 5, 7- Issue an alert concerning a risk of pollution of the main adsorbent mass (22) by the impurity (30) if the parameter linked to the measured content exceeds a tolerance threshold.