DE102017110234A1 - Nitrogen oxide reduction for lean burn engines with SCR storage model - Google Patents

Abstract

The invention relates to an exhaust aftertreatment system (10) comprising an SCR catalyst (16), an upstream reductant injector (15) and a control unit (20). A target value (Qi *) for the reducing agent adding amount is set on the one hand based on a pilot value (Qpre) depending on a feedstock nitrogen oxide content (cPre) and on the other hand on a storage guide value (Qstr). The storage guide value (Qstr) is calculated by a multi-unit simulation model (21). In order to adapt to various disturbing influences that may occur during the lifetime, an efficiency deviation (dEff) of the nitrogen oxide reduction compared to a desired value (rEff *) is detected in a learning process and the correction value adaptation (27) in at least two terms (31, 32) of the simulation model split. The division is preferably carried out according to predetermined division keys (SW). The figure intended for publication with the summary is FIG. 1.

Description

The disclosure relates to an exhaust aftertreatment system for an internal combustion engine having an SCR catalyst and a reducing agent injector and a method for determining the cause of an efficiency deviation of the nitrogen oxide reduction in the exhaust aftertreatment system.

It is known in practice to reduce nitrogen oxides in the exhaust gas of internal combustion engines, in particular lean-burn engines, by adding a reducing agent to nitrogen and other acceptable reaction products. For this purpose, the reducing agent is added to the exhaust gas and stored in an SCR catalyst. The SCR catalyst is a catalyst for selective catalytic reduction in which the added reducing agent attaches to a catalytic surface by adsorption and reacts there with nitrogen oxides in the exhaust gas.

The reducing agent content (stored amount) in the SCR catalyst should be in a predetermined value range for the reduction reaction to take place at a desired efficiency. If the level of reducing agent becomes too low or goes down to zero, sparking oxides can break through the SCR catalyst. If the reductant level becomes too high, the added and eventually toxic reductant may no longer build up in the SCR catalyst or partially desorb it. In either case, reductant may leak out of the SCR catalyst and into the outside atmosphere, or it may be necessary to provide additional measures to remove the excess reductant.

The previously known exhaust aftertreatment systems are not optimally formed. In particular, they are susceptible to certain disturbances, so that the efficiency of the oxide reduction is not ensured over the entire life of the exhaust aftertreatment system and in all operating conditions.

It is a first object of the present disclosure to provide an improved exhaust aftertreatment system. A second object of the disclosure is to provide an improved method for determining the cause of an efficiency deviation of the nitrogen oxide reduction. The claimed invention solves at least one of these objects by the features of the main claim.

The present disclosure encompasses several major aspects that can be utilized on their own or in any combination.

A first major aspect of the present disclosure relates to an exhaust aftertreatment system. The exhaust aftertreatment system is provided for an internal combustion engine and can be fed in particular with the exhaust gas of the internal combustion engine. It comprises at least one SCR catalyst with a flow passage and a wake passage. In the exhaust aftertreatment system, a (current) flow nitrogen oxide content in the flow direction of the exhaust gas upstream of the SCR catalyst, i. in the flow passage, as well as a (current) wake nitrogen oxide content in the flow direction after the SCR catalyst, i. determined in the caster passage. In other words, the flow passage in the flow direction of the exhaust gas is arranged in front of the SCR catalyst and the wake passage is arranged downstream of the SCR catalytic converter in the flow direction of the exhaust gas.

Upstream of the SCR catalyst is disposed a reducing agent injector which preferably adds a reducing agent into the exhaust gas. The reducing agent is supplied (together with the exhaust gas) to the SCR catalyst and stored in the SCR catalyst. Nitrogen oxides in the exhaust gas are reduced by the stored reducing agent. Particularly preferred is a nitrogen oxide sensor for detecting the flow of nitrogen oxide content upstream of the SCR catalyst and further upstream of the reducing agent injector.

The exhaust aftertreatment system comprises a control unit which is connected to the reducing agent injector and controls the instantaneous addition amount of the reducing agent. The control unit has a special design and makes it possible to carry out a learning operation in order to detect a detected deviation of the actual efficiency of the nitrogen oxide reduction from a desired value and to make multiple suitable corrections in order to jointly compensate for a number of disturbing influences. This allows a significantly improved adaptation of the control, which in particular overcomes the disadvantages of previously known, one-dimensional adjustments.

In the exhaust after-treatment system according to the present disclosure, the control unit includes a multi-unit simulation model and a pilot control. The control unit is designed to carry out the steps described below.

A target value for the current reducing agent addition amount (injection quantity) is determined on the basis of a precontrol value and additionally a storage control value. The pilot control value is determined on the basis of the (current) flow nitrogen oxide content. It represents that amount of reducing agent to be added to the exhaust gas to reduce the currently detected amount of nitrogen oxides in the exhaust gas. The feedforward control operates particularly fast and can fulfill the essential part of the control task in stationary states of the internal combustion engine or of the exhaust aftertreatment system in order to ensure the efficiency of the Sickoxid reduction.

The storage guide value is determined by the multi-unit simulation model in response to an estimated reductant level in the SCR catalyst to approximate the estimated reductant level to a desired value. The storage guide value thus indicates which positive or negative amount of reducing agent is to be added to the pilot value in order to keep the reducing agent content in the SCR catalyst in such a range that the nitrogen oxide reduction is carried out at the desired and preferably maximum efficiency. Adjustments can thus be made via the storage guide value which compensate for the various disturbing influences or deficiencies of the precontrol in the event of state changes in the exhaust gas aftertreatment system (transient system states).

Thus, for example, it is possible to determine by the learning method, on the one hand, whether a (determined) efficiency deviation of the nitrogen oxide reduction is based on a reducing agent excess or a reducing agent deficiency, in order to increase or decrease the reducing agent content in the SCR catalyst to approach a target value. In other words, it can be stated that the actual reducing agent content in the SCR catalyst deviates from an estimated reducing agent content. Thereafter, the estimated reducing agent content may be adjusted to the value determined by the learning method, i. a value representing a reductant excess or a reductant deficiency. The estimated reductant level then again represents the actual state in the SCR catalyst. In accordance with the deviation between the (updated) estimated value of the reducing agent content, a storage guide value is generated, which instructs an increased or reduced addition of reducing agent, so that the estimated reducing agent content is further approximated to the desired value.

In the exhaust aftertreatment system, various perturbations may occur intermittently or permanently, reducing the efficiency of nitric oxide reduction or other negative events, such as overfilling of the SCR catalyst (reductant excess) and consequent reductant breakthrough or emptying of the SCR catalyst (reducing agent deficiency) and a resulting nitrogen oxide breakthrough. Through the learning process, different states can be detected and an adaptation in the multi-unit simulation model can be made to compensate for the disturbances. These adaptations are preferably carried out in addition to the above-described approximation of the estimated reducing agent content to the desired value. If, for example, repeatedly determined in several passes of the learning process that there is a lack of reducing agent, this may indicate that the actual loading amount of the reducing agent injector, for example, by drift from the control specification (target Beigabemenge / Injektionskommando) deviates, which an adaptation of the simulation model can be compensated.

If, at the same time, it is found that there is only a small deviation in efficiency despite the lack of reducing agent, this may indicate an error or drift in the sensor signal which is output from a flow nitrogen oxide sensor Results from several rounds of the learning process are determined whether a compensation action (adaptation in the simulation model) was sufficient or expedient. If an insufficient reduction in the efficiency deviation or even an increase in the efficiency deviation is detected, a different adaptation strategy can be selected.

As the reducing agent, any chemical substance capable of reducing nitrogen oxides in the SCR catalyst can be used. Particularly preferred may be an aqueous urea solution (urea solution), which decomposes in the exhaust gas to ammonia. Ammonia (NH3) is the actually reducing substance and a toxic gas, which may enter the outside atmosphere only in certain concentrations according to the statutory provisions.

When an excess of reducing agent is added in the exhaust aftertreatment system, the reductant content in the SCR catalyst may increase beyond a predetermined desired value and even beyond the instantaneous storage capacity, with the non-storable portion of the reductant or one by one desorbed portion of the reducing agent passes through the SCR catalyst. In previously known exhaust aftertreatment systems in which there may be a significant reductant breakthrough, downstream of the SCR catalyst, a further filter or catalytic device must be provided to remove the excess reducing agent from the exhaust gas It may in particular a so-called ammonia slip catalyst act.

• - Abweichung der Ist-Beigabemenge an Reduktionsmittel von der Soll-Beigabemenge, beispielsweise durch Alterung oder Verschleiß des Reduktionsmittel-Injektors oder unzureichende Versorgung des Reduktionsmittel-Injektors;
• - Abweichung der tatsächlich durch chemische Reaktion umgesetzten Menge an Reduktionsmittel in dem SCR-Katalysator von einer geschätzten umgesetzten Menge, beispielsweise aufgrund eines (noch nicht detektierten) Reduktionsmittel-Überschusses oder Reduktionsmittel-Mangels oder bei sich stark ändernden Betriebsbedingungen des Verbrennungsmotors (schwankende Abgastemperatur / schwankender Stickoxid-Gehalt / schwankende Abgas-Strömungsgeschwindigkeit usw.);
• - Desorption von Reduktionsmittel aus dem SCR-Katalysator ohne chemische Umsetzung, beispielsweise durch hohe Strömungsgeschwindigkeit des Abgases, zu niedrige oder zu hohe Abgastemperatur, etc.;
• - Abweichung der Ist-Konzentration des beizugebenden Reduktionsmittels von einer Soll-Konzentration;
• - Abweichung des Versorgungs-Drucks, mit dem Reduktionsmittel zu dem Reduktionsmittel-Injektor gespeist wird, von einem Soll-Druck;
• - Unzureichende Wirksamkeit von Maßnahmen zur Vor-Umsetzung des Reduktionsmittels, beispielsweise unzureichende Ozon-Beigabe zur erhöhten Umsetzung von Harnstoff in Ammoniak;
• - Fehlerhafte Bestimmung von Eingangs-Parametern der Steuerung, insbesondere durch Kreuz-Sensitivität oder Drift an Sensoren, zeitliche Verzögerungen in der Messwert- oder Simulationswerterfassung, etc;
• - Inhomogene Speicherung des Reduktionsmittels im SCR-Katalysator.

Disturbing factors which may lead to insufficient or excessive addition of reducing agent include in particular the following:

• - Deviation of the actual Beigabemenge of reducing agent from the target Beigabemenge, for example by aging or wear of the reducing agent injector or insufficient supply of the reducing agent injector;
• Deviation of the amount of reducing agent actually reacted by chemical reaction in the SCR catalyst from an estimated converted amount, for example due to a (not yet detected) reductant excess or reductant deficiency or severely changing operating conditions of the internal combustion engine (fluctuating exhaust temperature / fluctuating) Nitrogen oxide content / fluctuating exhaust flow rate, etc.);
• Desorption of reducing agent from the SCR catalyst without chemical reaction, for example by high flow rate of the exhaust gas, too low or too high exhaust gas temperature, etc .;
• - Deviation of the actual concentration of the reducing agent to be added from a desired concentration;
• - deviation of the supply pressure, is fed to the reducing agent to the reducing agent injector, from a target pressure;
• - Insufficient effectiveness of measures for the pre-conversion of the reducing agent, for example, insufficient ozone addition to the increased conversion of urea into ammonia;
• - incorrect determination of input parameters of the control, in particular by cross-sensitivity or drift at sensors, time delays in the measured value or simulation value acquisition, etc .;
• Inhomogeneous storage of the reducing agent in the SCR catalyst.

According to the present disclosure, an embodiment of the multi-unit simulation model is proposed in combination with a learning method that is suitable for compensating for a plurality of the aforementioned negative influences.

The simulation model estimates the current reductant level in the SCR catalyst and estimates the various influences that result in an increase or decrease in reductant level in the SCR catalyst. These are in particular the adsorption of reducing agent in the SCR catalyst and the actually added amount of reducing agent. In addition, the desorption of reducing agent (without chemical reaction) and the conversion of reducing agent can be calculated or modeled by chemical reaction.

The simulation model thus comprises at least a first member in the form of a reducing agent adding model and a second member in the form of a reducing agent storage model, on the basis of which the reduction agent content is estimated in the SCR catalytic converter.

On the basis of the flow nitrogen oxide content and the caster nitrogen oxide content, an actual efficiency of the nitrogen oxide reduction is determined. In the learning process, upon determining an efficiency deviation between the actual efficiency and a target nitrogen oxide reduction efficiency, a correction value adjustment is performed in at least two members of the simulation model, for example, by changing correction values with respect to the reduction agent addition estimate and the estimate of reductant storage. A correction value may be, for example, a gain factor or an offset value or a substitute value to which a size of the simulation model is to be changed. By distributing the correction value adjustment to multiple parameters can be adjusted to different Disturbing influences are reacted, so that the efficiency of the nitrogen oxide reduction over the entire life of the exhaust aftertreatment system or its components remains within the allowable limits.

A second main aspect of the present disclosure relates to a method for determining the cause of an efficiency deviation in the nitrogen oxide reduction.

Nitrogen oxide sensors used in practice have a cross-sensitivity for nitrogen oxides and unreacted reducing agent or for reducing agent passing through the SCR catalyst. In other words, they show a positive measurement impact both when contacting with nitrogen oxides (NOx) and, for example, with ammonia (NH3).

Out US 2003/0046928 A1 are a description of the cross-sensitivity of a nitrogen oxide sensor and a method by which a distinction between a nitrogen oxide-based measurement and a reducing agent-based reading in an exhaust aftertreatment system with SCR catalyst and reducing agent injector to be enabled. The exhaust aftertreatment system is operated there in an aggressive loading scheme. As the nominal injection amount of reducing agent, the stoichiometric balance amount is added, and the SCR catalyst is continuously operated at a maximum reducing agent content. By the SCR catalyst breaking through reducing agent in the exhaust gas must therefore be removed if necessary by separate devices.

In a stationary state of the exhaust aftertreatment system according to US 2003/0046928 A1 In accordance with a predetermined scheme, an overdosed reducing agent pulse (injection pulse with test injection quantity greater than the nominal injection quantity) is added and a reaction of the nitrogen oxide measured value detected downstream of the SCR catalytic converter is evaluated. When the nitric oxide reading increases in response to the overdosed injection pulse, it is concluded that the sensor has reacted to reductant (ammonia / urea). If, on the other hand, the nitrogen oxide reading drops in response to the overdosed injection pulse, it is concluded that the sensor has reacted to nitric oxide (NOx). This form of conclusion is based on a simplified view of the processes in the SCR catalyst. The assumption may be summarized as follows: If there is insufficient reductant addition, all of the added reductant would react and nitrogen oxides would break through the SCR so that only nitric oxide reaches the sensor. If in such a case an overdosed reducing agent pulse is added, this should compensate for the previous defect, so that less nitrogen oxide reaches the sensor and the measured value drops. On the other hand, if there is excessive reductant addition, all of the nitric oxide would react so that only unreacted reductant breaks through the SCR catalyst and passes to the sensor. If an overdosed reducing agent pulse is then added, this should increase the excess, so that even more reducing agent reaches the sensor and the measured value increases.

After the revelation in US 2003/0046928 A1 It is furthermore provided that an underdosed reducing agent pulse (injection pulse with test injection quantity smaller than nominal injection quantity) is added, it being determined for each pulse whether the nitrogen oxide measured value downstream of the SCR catalytic converter rises or falls in response to a pulse , If the pulse direction and the direction of change of the sensor value have the same direction, it is concluded that the sensor reacts to reducing agent (ammonia / urea). If the pulse direction and the direction of change of the sensor value are opposite, it is concluded that the sensor has reacted to nitrogen oxide (NOx). For each reducing agent pulse, exactly one change in the sensor value due to the pulse is detected according to a predetermined time rule. In fact, the processes in the SCR catalyst are of a very complex nature and the simplified assumption can lead to erroneous interpretations.

In order to obtain a more reliable result is therefore in US 2003/0046928 A1 proposed to execute several pulses one after the other and to sum up the results of the pulses. However, this method has disadvantages that have proven to be unreliable for exhaust aftertreatment systems that are operated with a conservative supplementation scheme or lead to undesirable side effects.

In a conservative addition scheme, the storage of reducing agent in the SCR catalyst is not maximized, but guided according to a predetermined target value for the reducing agent content. For this reason, storage effects in the SCR catalyst have an increased influence on the change in the caster nitrogen oxide content. A reductant excess in the SCR catalyst can no longer be readily determined. Further, in a pilot control, a value may be a nominal injection quantity can be provided, which corresponds to that amount of reducing agent, which is believed to be responsive to the nitrogen oxide reduction in accordance with the current flow nitrogen oxide content - this amount is hereinafter referred to as the estimated reaction amount.

It should be avoided as possible that unreacted reducing agent breaks through the SCR catalyst, so that no additional purification of the exhaust gas needs to be done by separate devices. Thus, there is some risk involved or undesirable to sequentially carry out a larger number of overdosed reductant pulses. Because it could happen that only through the overdosed pulses a reductant breakthrough would be generated or provoked.

The nitrogen oxide reduction (even with a correctly adjusted reducing agent content in the SCR catalyst) is not 100%, but at a momentary maximum efficiency. In other words, the nitrogen oxide reduction has an efficiency of less than 1. If the stoichiometric equilibrium quantity is defined as the nominal injection quantity, then a certain proportion of the added reducing agent can not be reacted due to the efficiency. In other words, quasi-creeping overfilling of the SCR catalyst is enforced. When the SCR catalyst is already operating at a maximum reductant level, the unreacted amount of stoichiometric equilibrium can not be stored, thus resulting in undesirable reductant breakthrough. The addition of several overdosed reducing agent pulses for testing or learning purposes would further enhance this reductant breakthrough. It is therefore desirable to achieve a reliable result with only one or a few test impulses. This is achieved by the method disclosed herein by means of a multi-response analysis, which will be explained below.

The estimated reaction rate takes into account the efficiency of the SCR catalyst and provides, for example, as the nominal injection quantity, the product of stoichiometric equilibrium quantity and estimated reduction efficiency, that is, a value that is lower than in an aggressive addition scheme. In other words, in the conservative addition scheme according to the present disclosure by the pilot control (only maximum) as much reducing agent is added, as necessary, to compensate for the reduction in a reduction reaction in the SCR catalyst reduction of the reducing agent content.

The main advantage of the conservative addition scheme is thus the avoidance of (creeping) overfilling of the SCR catalyst and a reductant breakthrough generated thereby. Further advantages are the simplification of the exhaust aftertreatment system and a reduced reducing agent consumption.

It has been shown that by a conservative addition scheme some deviations in the measurable behavior of the exhaust aftertreatment system occur, which are not yet fully elucidated. In particular, it was recognized that in US 2003/0046928 A1 described method can lead to significant erroneous detections, because it does not account for memory effects in the SCR catalyst and the different kinetics (conversion rate) of adsorption, desorption and reaction processes under different operating conditions of the exhaust aftertreatment system. There are several operating parameters that can have different and even opposite effects on the kinetics of the adsorption, desorption and reaction processes. Thus, for example, by falling temperatures and decreasing volume flow, the adsorption of reducing agent in the SCR catalyst is enhanced. On the other hand, falling temperatures also inhibit the chemical reaction.

The reaction of a sensor for detecting the caster nitrogen oxide content to a reducing agent pulse takes place in a conservative addition scheme with an unpredictable dynamics, so that a distinction between a positive or negative change in the caster nitric oxide content according to time rules or alone Based on low-pass filtering leads to misinterpretations.

It is assumed that the adsorption and desorption of reducing agent in the SCR catalyst takes place locally differently, ie spatially inhomogeneously, by the conservative addition scheme. On the one hand, in the SCR catalytic converter in the flow direction of the exhaust gas, front surface regions appear to react more quickly and more directly to a change in the reducing agent addition than regions located at the rear. On the other hand, in the cross section of the SCR catalyst, depending on the flow velocity, pressure, temperature and volumetric flow of the exhaust gas, some regions appear to be stronger and some regions to be less involved in the adsorption, desorption and reaction processes. After a transient operation (state with changing operating parameters), it is unpredictable how strongly an inhomogeneous storage of reducing agent in the SCR catalyst is pronounced. It seems that at the same time in some areas the SCR catalyst to a surplus of reducing agent (overflow) and in other areas to a reducing agent deficiency (an idling), with both processes overlap.

By the method according to the second main aspect of the present disclosure, it is possible to determine the cause of an efficiency deviation of the nitrogen oxide reduction (regardless of a temporal prediction). The method is preferably carried out in the exhaust aftertreatment system described above and includes the following steps.

A current flow nitrogen oxide content in the flow passage is determined. The determination can be made on the basis of a measurement by a feedstock nitrogen sensor and / or on the basis of a model-based estimation or calculation of the feedstock nitrogen oxide content. A measurement of the feedstock nitrogen oxide content is preferably performed in the flow direction of the exhaust gas before the reducing agent injector because of the above-mentioned cross-sensitivity, so that the feedstock nitrogen oxide sensor can not come into contact with the added reducing agent. Furthermore, an instantaneous clogging nitrogen oxide content in the catenary passage is determined by a caster nitrogen oxide sensor having a cross-sensitivity to nitrogen oxides (NOx) and the added reducing agent (ammonia / urea).

When a stationary condition of the exhaust aftertreatment system is detected, the control unit causes the reductant injector to execute a positive or negative injection pulse, i. an injection pulse having a test addition amount that is well above or below a nominal allowance determined according to the current flow rate nitric oxide content. The Nominal Beigabemenge can be determined by an aggressive or preferably a conservative addition scheme.

The measurement signal of the wake-up nitrogen oxide sensor is detected and evaluated in a period of time after the injection pulse and before any further injection pulse with a plurality of separate partial responses. The evaluation can be done in any way and will be explained below.

A decision on the cause of the efficiency deviation is made on the basis of an overall consideration of the multiple partial answers of any detected changes in direction of the partial answers.

The disclosed method has several advantages. By evaluating the measurement signal with partial responses, the analysis of the measurement signal can essentially be independent of the dynamics (speed) of the reaction. In particular, the time window for an interpretation of the responses to a pulse may be substantially indefinite. Thus, the same measurement method is applicable to both fast and slow reactions.

Preferably, an efficiency of the nitrogen oxide reduction is calculated on the basis of the flow nitrogen oxide content and the clogged nitrogen oxide content. The calculated efficiency is compared with an estimated efficiency or a target value for the efficiency to determine a deviation. The execution of an injection pulse may be dependent on the condition that the calculated efficiency deviates by an impermissible extent from the desired value, or that there is a minimum deviation. Alternatively, an injection pulse may be performed without a prior determination of an efficiency deviation.

Particularly preferably, the multiple partial answers can be compared with decision patterns. There may be simple decision patterns that allow a quick first interpretation of the cause of a change in efficiency. These decision patterns can be meaningful for cases in which memory effects do not occur or only to a small extent and there is an essentially fast dynamic. In addition, more complex decision patterns can be defined, which can differentiate between early partial answers and opposing later partial answers.

Further advantageous embodiments of the invention are specified in the subclaims, the following description and the drawings.

• 1: eine schematische Darstellung des Abgasnachbehandlungssystems gemäß der vorliegenden Offenbarung;
• 2: Ein erster Satz von Entscheidungsmustern zur Bestimmung der Ursache einer Effizienz-Abweichung der Stickoxid-Reduktion;
• 3: Vorgänge in einem SCR-Katalysator unter Soll-Bedingungen bei konservativer Beigabe von Reduktionsmittel;
• 4A u. 4B: Vorgänge in einem SCR-Katalysator unter Soll-Bedingungen auf einen positiven und einen negativen Injektionspuls;
• 5 u. 6: Mögliche Vorgänge in einem SCR-Katalysator bei Reduktionsmittel-Mangel auf einen negativen Injektionspuls;
• 7 u. 8: Mögliche Vorgänge in einem SCR-Katalysator bei Reduktionsmittel-Mangel auf einen positiven Injektionspuls;
• 9 u. 10: Mögliche Vorgänge in einem SCR-Katalysator bei Reduktionsmittel-Überschuss auf einen negativen Injektionspuls;
• 11 u. 12: Mögliche Vorgänge in einem SCR-Katalysator bei Reduktionsmittel-Mangel auf einen positiven Injektionspuls;
• 13: Mögliche Vorgänge in einem SCR-Katalysator auf eine positiven Injektionspuls, wenn der Reduktionsmittel-Gehalt unterhalb eines Soll-Werts liegt;
• 14: Ein zweiter Satz von Entscheidungsmustern zur Bestimmung der Ursache einer Effizienz-Abweichung der Stickoxid-Reduktion;
• 15: Beispielhafte Analyse des Mess-Signals des Nachlauf-Stickoxid-Sensors nach einem positiven Injektionspuls
• 16: Ein Beispiel für die Ausführung des Lernverfahrens gemäß der vorliegenden Offenbarung in mehreren Durchgängen.

The invention is illustrated by way of example and schematically in the drawings. These show:

• 1 1 is a schematic illustration of the exhaust aftertreatment system according to the present disclosure;
• 2 : A first set of decision patterns for determining the cause of an efficiency deviation of the nitric oxide reduction;
• 3 : Processes in an SCR catalyst under nominal conditions with conservative addition of reducing agent;
• 4A and , FIG. 4B: operations in an SCR catalytic converter under nominal conditions on a positive and a negative injection pulse; FIG.
• 5 and , 6: Possible processes in an SCR catalyst with reducing agent deficiency on a negative injection pulse;
• 7 and , 8: Possible processes in an SCR catalyst with reducing agent deficiency on a positive injection pulse;
• 9 u , 10: Possible processes in an SCR catalyst with reducing agent excess to a negative injection pulse;
• 11 u , 12: Possible processes in an SCR catalyst with reducing agent deficiency on a positive injection pulse;
• 13 : Possible events in an SCR catalyst to a positive injection pulse, if the reducing agent content is below a target value;
• 14 : A second set of decision patterns for determining the cause of an efficiency deviation of nitric oxide reduction;
• 15 : Exemplary analysis of the measuring signal of the wake-up nitric oxide sensor after a positive injection pulse
• 16 : An example of the execution of the learning method according to the present disclosure in multiple passes.

The exhaust aftertreatment system ( 10 ) according to the present disclosure is in 1 outlined. At the bottom is a catalysis section ( 12 ) shown with a flow passage ( 11 ) and a trailing passage ( 13 ) connected is. Through the flow passage ( 11 ), an exhaust gas is added to the catalysis section ( 12 ). The exhaust gas contains nitrogen oxides (NOx). That from the catalysis section ( 12 ) and largely purified by nitrogen oxides (NOx) exhaust gas is passed through the wake passage ( 13 ) dissipated.

In the catalysis section ( 12 ) is at least one SCR catalyst ( 16 ), which may have a known structure. With the SCR catalyst ( 16 ), if necessary, a filter ( 17 ), in particular a soot particle filter. Alternatively or additionally, a separate filter in the exhaust aftertreatment system ( 10 ), in particular downstream of the SCR catalyst ( 16 ).

In the example of 1 is upstream of the SCR catalyst ( 16 ) a diesel oxidation catalyst ( 14 ) is provided as an optional element. Further, upstream of the SCR catalyst ( 16 ) a reducing agent injector ( 15 ), which may also have a known structure. The reducing agent injector ( 15 ) is in the example of 1 in the flow direction of the exhaust gas directly in front of the SCR catalyst ( 16 ) arranged. Alternatively, it may be arranged at any other location, for example upstream of a turbine or / or an exhaust gas mixer (not shown), through which a mixing of reducing agent and exhaust gas is improved.

The reducing agent injector ( 15 ) is preferably supplied with a liquid reducing agent, which is fed with a delivery pressure. In the context of the present disclosure, for reasons of simplified illustration, it is not distinguished whether the reducing agent is injected as precursor (urea / urea), which first converts into the reducing substance (ammonia), or whether it is a ready-made preparation. The amount of reductant added means that amount of precursor or reducing substance required according to the underlying chemical reaction to reduce a certain amount of nitrogen oxides (NOx). In the case of ammonia, the following chemical equations may be used to determine a required amount of admixture: 4NO + 4NH ₃ + O ₂ → 4N ₂ + 6H ₂ O, 2NO ₂ + 4NH ₃ + O ₂ → 3N ₂ + 6H ₂ O.

If a distinction is made between the different types of oxides (NO, NO2, NO3 ...) when determining the nitrogen oxide content (cPre, cPost), the simplified empirical formula can be used: NO + NO ₂ + 2 NH ₃ → 2N ₂ + 3H ₂ O.

As a result, a simplified stoichiometric ratio of 1 to 1 between nitrogen oxides (NOx) and ammonia (NH3) can be assumed. The presumed ratio of nitrogen monoxide (NO), nitrogen dioxide (NO 2) and other nitrogen oxide compounds (NO x) may possibly be determined by model-based calculation or on the basis of maps, etc. depending on the operating state of the internal combustion engine, to even more accurate to determine the stoichiometric equilibrium amount, ie what amount of reducing agent according to the chemical equations to convert the detected amount of nitrogen oxides is provided. When using another reducing agent, other formulas must be used accordingly.

The reducing agent injector ( 15 ) is connected to the control unit ( 20 ) of the exhaust aftertreatment system ( 10 ) and is powered by an injector driver ( 24 ) energized. The injector driver ( 24 ) can be arranged at any position, for example as part of the control unit ( 20 ) (please refer 1 ) or as part of the reducing agent injector ( 15 ).

Through the injector driver ( 24 ) is set according to a target value (Qi * ) for the reduction agent addition an injection command ( CQ ), with which the reducing agent injector ( 15 ) is applied. The injection command ( CQ ) can in particular an activation current for an electric actuator of the reducing agent injector ( 15 ) be.

By the reducing agent injector ( 15 ) (actually) added quantity ( Qi ) of reducing agent is the SCR catalyst ( 16 ). In the enlarged section of 1 some operations are outlined, by which the (actual) reducing agent content ( cRed ) in the SCR catalyst increases or decreases. The proportion of the added amount (Qi) of reducing agent which is adsorbed on the catalytic wall is indicated by (Qad). By (currently) adsorbed amount (Qad) of reducing agent increases the reducing agent content or the stored amount of reducing agent ( cRed ).

The proportion of the stored amount (cRed) of the reducing agent, which is reacted by chemical reaction with nitrogen oxides (NOx), is the reaction quantity ( qr ). The (currently) reacted amount (Qr) of reducing agent lowers the reducing agent content ( cRed ) in the SCR catalyst. It can be assumed in a simplified manner that reducing agent adsorbed exclusively on the catalytic wall contributes to a chemical conversion of nitrogen oxides (NOx), ie it is first necessary for a reducing agent constituent to be adsorbed on the wall so that it can subsequently be bound to the reduction Reaction can participate. A direct conversion of nitrogen oxides (NOx) by the reducing agent present in the exhaust gas does not occur or at negligible conversion rates.

The desorbed amount ( qdc ) of reducing agent indicates how much reducing agent is removed from the catalytic wall again without chemical reaction. It may also indicate what excess of the quantity actually contributed ( Qi ) leaves the SCR catalyst to reducing agent without chemical reaction, for example. When an overfilling of the SCR catalyst occurs and for this proportion no adsorption on the catalytic wall is possible.

The control of the reducing agent addition takes place in such a manner that, on the one hand, the reducing agent content ( cRed ) is kept so high in the SCR catalyst that the desired efficiency of the nitrogen oxide reduction is achieved, and on the other hand as possible no overfilling and no empty state of the SCR catalyst ( 16 ) enter. The target value (cRed *) for the reducing agent content in the SCR catalyst ( 16 ) represents such an amount of stored reducing agent that the above-mentioned objects are achieved. The target value ( cRed * ) is adapted to the current operating state of the exhaust aftertreatment system, for example, to take into account the changing storage capacity at increasing and decreasing temperatures as well as increasing or decreasing volume flows.

The target value for the reducing agent addition amount ( Qi * ) can be defined in any desired manner, for example as an injection rate (amount added per time interval). The reductant additives ( Qi ) can occur continuously or intermittently, for example at a constant or variable injection cycle, wherein for each injection or groupwise different individual injection quantities are predetermined in order to obtain the predetermined injection rate or the desired value (in the mean value over one or more cycles). Qi * ) to reach. With an increased specification of the desired value ( Qi * ) Injections can be carried out with increased amount of added and vice versa. Alternatively, in each case uniform single injection quantities can be predetermined, wherein the amount of admixture per time interval can be changed by specifying an injection frequency an adjustment to the target value ( Qi * ) to reach. Alternatively, mixtures of a change in the single injection quantity and a change in the injection frequency are possible. For the sake of simplification, it is assumed below that a desired value ( Qi * ) (per time interval) for the reducing agent addition is changed, for example. Depending on the state parameters of the exhaust aftertreatment system ( 10 ) into concrete injection commands ( CQ ) is implemented for single injections.

While the reducing agent addition ( Qi ) occurs rather intermittently, the chemical reaction takes place rather continuously. The storage of the reducing agent results in a buffering effect which allows a balance between the intermittent reductant addition and the continuous reductant conversion.

The setpoint ( Qi * ) for the reducing agent addition is determined by a setpoint calculation ( 23 ) and at least based on the pre-tax value ( Qpre ) and the memory management value ( SCNTR ) for the reducing agent addition. The underlying calculation can be of any kind. In a simple implementation, the instantaneous precontrol value ( Qpre ) and the current memory management value ( SCNTR ) are added. The storage management value ( SCNTR ) can have a positive or negative value or zero. As an alternative or in addition, more complex calculations may be used, which, for example, enable a temporary suppression of the addition of reducing agent with subsequent compensation, ie a temporal shift of add-on quantities between (adjacent) time intervals. A temporary suppression of reducing agent additives may, for example, be necessary in such time intervals in which the operating state of the internal combustion engine or of the exhaust gas aftertreatment system changes very greatly.

The pre-tax value ( Qpre ) is controlled by the feedforward control ( 22 ), ie in the open loop, and based on the instantaneous flow nitrogen oxide content ( CpRe ) certainly. The feedstock nitrogen oxide content ( CpRe ) can be determined in any way. In the example of 1 is a flow nitrogen oxide sensor ( 40 ) in the flow passage ( 11 ), which determines the flow nitrogen oxide content ( CpRe ) recorded directly. Alternatively or additionally, a flow nitrogen oxide content ( CpRe ) may be provided as an estimated value from another control component, in particular from an engine controller having a combustion model that determines the instantaneous nitrogen oxide output from a combustion chamber of the internal combustion engine. Both a measured value and an estimated or calculated value of the precursor nitrogen oxide content ( CpRe ), whereby these values can be checked for consistency and, if necessary, adapted. Again alternatively or additionally, a total amount of nitrogen oxides (NOx) can be determined via the flow nitrogen oxide sensor, and the nitrogen oxides (NO, NO2, NO3, etc.) contained can be determined via an estimation or calculation. In the following, it is simplified to assume that the precursor nitrogen oxide content ( CpRe ) is sensory detected.

The multi-unit simulation model ( 21 ) explained in more detail. In the drawings, estimated values representing an actual size in the control are indicated by a prime and set values by an asterisk, respectively. ( Qi ) denotes the actual amount of reducing agent added, which is usually not directly determinable, while ( Qi * ) the target value and ( Qi ' ) specify an estimated value for the quantity to be added. ( cRed ) gives the actual reducing agent content in the SCR catalyst ( 16 ), which is also not directly determinable while ( cRed * ) a target value and ( cRed ' ) indicate an estimated value of the reducing agent content, etc.

The storage management value ( SCNTR ) is preferred as a deviation of the estimated reducing agent content ( cRed ' ) from the setpoint ( cRed * ). The setpoint ( cRed * ) can be specified in any way. According to a simple embodiment, it may be defined as a static value and, for example, a maximum storage capacity of the SCR catalytic converter ( 16 ) under normal operating conditions. Alternatively and preferably, the desired value ( cRed * ) dynamically determined, for example, depending on the state parameters of the exhaust aftertreatment system ( 10 ) and / or the internal combustion engine. The determination can be carried out in particular as a function of the temperature of the SCR catalyst ( 16 ) and / or the temperature of the exhaust gas and / or an age of the SCR catalyst ( 16 ) respectively. Furthermore, the flow velocity or the volume flow of the exhaust gas can cause an adjustment of the desired value (cRed *).

The setpoint value ( cRed * ) are determined for the reducing agent content according to a predetermined addition scheme.

In the context of the present disclosure, a distinction is made between at least one aggressive and one conservative addition scheme, whereby the specification of the desired value ( cRed * ) for the reducing agent content is preferably carried out according to the conservative addition scheme.

A conservative addition scheme stipulates that the target value ( cRed * ) for the reducing agent content in the SCR catalyst ( 16 ) is significantly lower than the (instantaneous) maximum storage capacity of the SCR catalyst ( 16 ) for reducing agents. For example, a conservative addition scheme sees the setpoint ( cRed * ) in a range of 30% to 80% of the (current) maximum storage capacity. This is an undesirable overfilling of the SCR catalyst ( 16 ) and a resulting reductant breakthrough prevented.

In the case of an aggressive add-on scheme, the target value ( cRed * ) is set in a range greater than 80% or greater than 90% of the (current) maximum storage capacity, which is also possible, but may require that downstream of the SCR catalyst is an ammonia slip catalyst (not shown) or other element Removal of excess reducing agent is provided in the exiting exhaust gas.

The abovementioned value ranges are given purely by way of example and may vary depending on the size and effectiveness of the SCR catalyst ( 16 ) to change.

In the multi-unit simulation model ( 21 ) according to 1 are contained various members which, depending on the state parameters of the exhaust aftertreatment system ( 10 ) and in particular the SCR catalyst ( 16 ) estimate or calculate by what influences the current reducing agent content (cRed) increases or decreases. The number and design of the links may vary depending on the design of the exhaust aftertreatment system ( 10 ) and in particular the catalysis section ( 12 ) be modified.

In the example of 1 includes the multi-unit simulation model ( 21 ) a first member ( 31 ) in the form of a reducing agent addition model. This determines depending on an actuation ( Qi * / CQ ) of the reducing agent injector ( 15 ) an estimated amount ( Qi ' ) on added reducing agent. In other words, the reducing agent adding model ( 31 ) an estimated value ( Qi ' ) for the quantity actually added ( Qi ) of reducing agent. The actuation of the reducing agent injector ( 15 ) can be characterized by any system size, preferably by the desired value ( Qi * ) for the reducing agent addition or by the injection command ( CQ ).

• - Konzentration der reduzierenden Substanz oder eines Vorprodukts im Speise-Fluid des Reduktionsmittel-Injektors;
• - Speise-Druck des Reduktionsmittel-Injektors;

The estimated amount ( Qi ' ) of the added reducing agent may furthermore depend on the following parameters:

• Concentration of the reducing substance or a precursor in the feed fluid of the reducing agent injector;
• - Feed pressure of the reducing agent injector;

15 shows an example of the course of the maximum storage capacity ( Cmax ), the target value ( cRed * ) and the estimated value ( cRed ' ) for the reducing agent content compared to the actual reducing agent content ( cRed ). The actual reducing agent content ( cRed ) is not directly measurable and therefore not known to the controller. In the time range shown, three stationary states ( S1 . S2 . S3 ), between each of which there are transient or transient states. In each of the three stationary states ( S1 . S2 . S3 ) is another operating state of the exhaust aftertreatment system ( 10 ), so that different maximum storage capacities ( Cmax ) are present. Accordingly, for each of the stationary states ( S1 . S2 . S3 ) another setpoint value ( cRed * ) for the reducing agent content.

In the first steady state, the reductant addition and storage management are done correctly. The estimated value ( cRed ' ) of the reducing agent content corresponds to the actual reducing agent content (cRed) and there is no undue deviation between the estimated reducing agent content ( cRed ' ) and the target value ( cRed * ) in front. Accordingly, it is expected that the nitrogen oxide reduction will occur at the correct efficiency.

In the second stationary state ( S2 ) is another operating state of the exhaust aftertreatment system ( 10 ), which is why a higher storage capacity ( Cmax ) of the SCR catalyst ( 16 ) and a correspondingly higher setpoint value ( cRed * ) are present for the reducing agent content. The reducing agent addition and the storage management are also correct.

During the transition to the third stationary state, ie in the transient transition, a deviation occurs between the estimated reducing agent content ( cRed ' ) and the actual reducing agent content ( cRed ) on. The estimated reducing agent content ( cRed ) is erroneously at the level of the target value ( cRed * ). In fact, however, there is an excess of reducing agent which, according to the measured values of the clogged nitrogen oxide content ( cPost ) leads to a reduction in the efficiency of nitric oxide reduction. So there is a reductant breakthrough instead.

In a first implementation of the learning process below, the excess of reducing agent is detected. As a reason for the formation of the reducing agent excess, various phenomena come into consideration. On the one hand, the reducing agent injector ( 15 ) an actual amount ( Qi ) have been added to reducing agent, which was greater than the target amount ( Qi * ). On the other hand, in the SCR catalyst, in fact, more reductant could have been adsorbed or less reductant reacted than estimated in the simulation model. That is, the estimated adsorption amount ( Qad ) could have been smaller than the actual adsorption amount in the pre-learning period ( Qad ' ) and the estimated reaction amount ( qr ' ) could have been smaller than the actual reaction rate ( qr ).

As a countermeasure on the one hand, the estimated reducing agent content ( cRed ' ) to the value ( Cmax ) customized. In addition, further correction value adjustments may be made in the simulation model, which will be explained below.

If the estimated reductant content ( cRed ' ) to the value ( Cmax ) is adapted, the impermissible deviation from the desired value ( cRed * ) for the reducing agent content detected by the controller. Accordingly, a negative storage guide value ( SCNTR ), which leads to lower target injection quantities ( Qi * ) and thus causes the estimated reductant level ( cRed ' ) again the target value ( cRed * ) is approximated.

In the example of 16 does the simulation model go after the first execution ( L1 ) of the learning process assumes that the estimated reducing agent content ( cRed ' ) at the target value ( cRed * ) remains. In fact, in the example shown, however, a deviation occurs again, which leads to a reducing agent excess, ie the actual reducing agent content ( cRed ) is again above the target value ( cRed * ) and drifts against the value of ( CMAS ). The reason for this could be that the actual quantity of Qi ) is greater than the target value ( Qi * ), which is due to a measurement error of the flow nitrogen oxide sensor ( 40 ) or an error of the injector ( 15 ) could be traceable.

In a further implementation ( L2 ) of the learning process, the renewed excess of reducing agent can be recognized, wherein the estimated reducing agent content ( cRed ' ) is changed back to a suitable value, and further adaptations can be made in the simulation model to determine the detected drift of the reducing agent content ( cRed ) to compensate. Suitable measures are explained below.

At or in the reducing agent adding model ( 31 ), any correction value ( W1 ) that is adaptable when performing a learning operation. In the example of 1 For simplification, it is assumed that the correction value W1 is a multiplicative factor representing the estimated value ( Qi ' ) of the amount added to a base calculation increased or decreased. Alternatively or additionally, other correction values may be provided which, for example, provide one or more concrete adjustments with respect to the aforementioned influencing parameters.

• - Momentaner (geschätzter) Reduktionsmittel-Gehalt (cRed') im SCR-Katalysator (16) - entspricht Besetzungs- bzw. Füllungsgrad der Katalysatorwandung;
• - Momentane (geschätzte) Beigabemenge (Qi') an Reduktionsmittel - bezogen vom Reduktionsmittel-Beigabemodell (31);
• - Temperatur des SCR-Katalysators (16);
• - Temperatur des zugeführten Abgases;
• - Konzentration der reduzierenden Substanz im zugeführten Abgas;
• - Durchmischungsgrad von Reduktionsmittel und Abgas stromaufwärts zu SCR-Katalysator, beispielsweise abhängig von Anordnung des Reduktionsmittel-Injektors und Betrieb oder Last einer zwischen dem Reduktionsmittel-Injektor (15) und dem SCR-Katalysator (16) etwaig angeordneten Turbine;
• - Strömungsgeschwindigkeit bzw. Volumenstrom des Abgases, bspw. bezogen von einem Strömungssensor (nicht dargestellt) oder von einer Motorsteuerung.

The second link ( 32 ) of the simulation model ( 21 ) is a reducing agent storage model ( 32 ). This estimates as a function of state parameters of the exhaust aftertreatment system ( 10 ), whitch amount ( Qad ' ) reducing agent currently in the SCR catalyst ( 16 ) is stored by adsorption. The estimated amount (Qad ') may depend on the following state parameters:

• - Current (estimated) reducing agent content ( cRed ' ) in the SCR catalyst ( 16 ) - corresponds to occupation or degree of filling of the catalyst wall;
• - Current (estimated) amount of add-on ( Qi ' ) of reducing agent - based on the reducing agent addition model ( 31 );
• Temperature of the SCR catalyst ( 16 );
• - temperature of the supplied exhaust gas;
• - Concentration of the reducing substance in the supplied exhaust gas;
• - Degree of mixing of reducing agent and exhaust gas upstream to SCR catalyst, for example, depending on the arrangement of the reducing agent injector and operation or load of a between the reducing agent injector ( 15 ) and the SCR catalyst ( 16 ) possibly arranged turbine;
• Flow velocity or volume flow of the exhaust gas, for example, based on a flow sensor (not shown) or of a motor control.

In the example of 1 For the sake of simplification, it is assumed that a correction value ( W2 ) on or in the reducing agent storage model ( 32 ) is a multiplicative factor that determines the value ( Qad ' ) is increased or decreased compared to a base calculation. Alternatively or additionally, other correction values may be provided which, for example, provide one or more concrete adjustments with respect to the aforementioned influencing parameters.

This in 1 separately shown summation element ( 25 ), which calculates the estimated reducing agent content (cRed '), can also be part of the reducing agent storage model ( 32 ) be. Accordingly, the correction value ( W2 ) - alternatively or in addition to the abovementioned possibilities - a multiplicative factor, an offset value or a substitute value for the estimated reducing agent content ( cRed ' ) be. Particularly preferably, it is an offset value.

• - Momentaner (geschätzter) Reduktionsmittel-Gehalt (cRed') im SCR-Katalysator (16) - entspricht Besetzungs- bzw. Füllungsgrad der Katalysatorwandung;
• - Temperatur des SCR-Katalysators (16);
• - Temperatur des zugeführten Abgases;
• - Konzentration der Stickoxide (NOx) im zugeführten Abgas;
• - Strömungsgeschwindigkeit des Abgases.

The third link ( 33 ) of the simulation model ( 21 ) is in the example of 1 a reaction model ( 33 ), which depends on the state parameters of the exhaust aftertreatment system ( 10 ) estimates what proportion ( qr ' ) of the reducing agent stored in the SCR catalyst is removed or converted by chemical reaction. The estimated amount ( qr ' ) of the reducing agent converted by reaction may depend on the following state parameters:

• - Current (estimated) reducing agent content ( cRed ' ) in the SCR catalyst ( 16 ) - corresponds to occupation or degree of filling of the catalyst wall;
• Temperature of the SCR catalyst ( 16 );
• - temperature of the supplied exhaust gas;
• - concentration of nitrogen oxides (NOx) in the supplied exhaust gas;
• - Flow rate of the exhaust gas.

In the example of 1 For the sake of simplification, it is assumed that a correction value ( W3 ) on or in the reaction model ( 33 ) is a multiplicative factor representing the estimated value ( qr ' ) is increased or decreased compared to a base calculation. Alternatively or additionally, other correction values may be provided which, for example, provide one or more concrete adjustments with respect to the aforementioned influencing parameters. In particular, a correction value ( W3 ) may be a multiplicative factor and / or an offset value that determines the feedstock nitrogen oxide content ( CpRe ) compared to the measured value of the flow nitrogen oxide sensor ( 40 ) increases or decreases. A correction value adjustment with respect to the measured value of the forward spark-type oxide sensor ( 40 ) can simultaneously affect the feedforward ( 22 ) and thus the pre-tax value ( Qpre ) for the reducing agent addition. Thus, a correction value adjustment in addition to the changes in the multi-unit simulation model ( 21 ) a correction value ( W3 ) for or in the feedforward control ( 22 ) to change.

• - Momentaner (geschätzter) Reduktionsmittel-Gehalt (cRed') im SCR-Katalysator (16) - entspricht Besetzungs- bzw. Füllungsgrad der Katalysatorwandung;
• - Momentane (geschätzte) Beigabemenge (Qi') an Reduktionsmittel - bezogen vom Reduktionsmittel-Beigabemodell (31);
• - Temperatur des SCR-Katalysators (16);
• - Temperatur des zugeführten Abgases;
• - Strömungsgeschwindigkeit des Abgases.

The fourth term in the simulation model ( 21 ) is a desorption model ( 34 ), which depends on the state parameters of the exhaust aftertreatment system ( 10 ), which proportion ( qdc ' ) of the reducing agent stored in the SCR catalyst is desorbed without chemical reaction. The estimated amount ( qdc ' ) of the desorbed reducing agent may further comprise a proportion of the estimated added amount ( Qi ' ) which can not be adsorbed in the SCR catalyst. The estimated amount ( qdc ' ) may depend on the following state parameters:

• - Current (estimated) reducing agent content ( cRed ' ) in the SCR catalyst ( 16 ) - corresponds to occupation or degree of filling of the catalyst wall;
• - Current (estimated) loading amount (Qi ') of reducing agent - related to the reducing agent addition model ( 31 );
• Temperature of the SCR catalyst ( 16 );
• - temperature of the supplied exhaust gas;
• - Flow rate of the exhaust gas.

In the example of 1 For the sake of simplification, it is assumed that a correction value ( W4 ) on or in the desorption model ( 33 ) is a multiplicative factor that determines the value ( qdc ' ) is increased or decreased compared to a base calculation. Alternatively or additionally, other correction values may be provided which, for example, provide one or more specific adjustments with respect to the aforementioned influencing parameters.

The control concept according to the present disclosure provides that by the precontrol value ( Qpre ) a significant proportion of the target quantity ( Qi * ) of the reducing agent to be added as a function of the instantaneous flow nitrogen oxide content ( CpRe ). This can be done with a very small time delay. In other words, always (almost) so much reducing agent according to the precontrol value ( Qpre ) that the loss of reducing agent in the SCR catalyst ( 16 ) is compensated as a result of the chemical transformation. Thus, by the pilot control ( 22 ) a basic adjustment of the reducing agent addition to the current need.

By specifying the storage guide value ( SCNTR ), which depends on a simulation on the processes in the SCR catalyst ( 16 ), which leads to an increase or decrease in the instantaneous reducing agent content ( cRed ), especially in unsteady states of the internal combustion engine or the exhaust aftertreatment system ( 10 ) take place a compensating or even finer control of the reducing agent addition. The underlying calculations and estimates may be due to the temporal buffering effect of the SCR catalyst ( 16 ) with a greater time delay.

The following is an example of the implementation of a learning process explained.

A comparator ( 26 ) compares the instantaneous flow nitrogen oxide content ( CpRe ) (estimated and / or measured) and the instantaneous clogged nitric oxide content ( cPost ) (measured) and calculates therefrom an instantaneous nitrogen oxide reduction efficiency. The comparator ( 26 ) compares the current efficiency with a setpoint ( Reff * ) and calculates an efficiency deviation ( deff ).

An efficiency deviation occurs in particular when the SCR catalyst ( 16 ) has a reduced efficiency, is overcrowded or runs idle. A reduced effectiveness can, for example, on an aging of the SCR catalyst ( 16 ), by which the efficiency is reduced. Overfilling or idling of the SCR catalytic converter can, for example, be based on a temporary or permanent deviation of the actual amount added ( Qi ) from the target value ( Qi * ), or due to insufficient adsorption (Qad).

The efficiency deviation ( deff ) is replaced by a correction value adjustment ( 27 ) are used in at least two of the members ( 31 . 32 . 33 . 34 ) of the simulation model ( 21 ) make a correction, in particular by changing the correction values ( W1 . W2 . W3 . W4 ). In addition, a correction of the estimated reductant level ( cRed ' ) respectively.

Particularly preferred is the efficiency deviation ( deff ) according to a weighting key ( SW ) to two or more correction values ( W1 - W4 ) in the at least two members ( 31 - 34 ) of the simulation model ( 21 ), in particular by dividing the deviation value (dEff) into two or more percentage shares.

In or for the correction value adjustment ( 27 ) two or more alternative weighting keys (SW) are preferably predefined, from which a suitable weighting key (SW) SW ) is selected. The selection can be made on the basis of any criteria, for example based on the success of a previous learning process. Alternatively or additionally, the selection based on further state parameters of the exhaust aftertreatment system ( 10 ), for example a detected inconsistency between a model-based determined value and a measured value for the lead nitrogen oxide content ( CpRe ). A correction value ( W1 - w4 ) may alternatively or in addition to the above-mentioned possibilities provide for a gain for a measured value or a calculated value which corresponds to the simulation model ( 21 ) is supplied. These include in particular a measured value of a flow nitrogen oxide sensor, a lagging nitrogen oxide sensor, an exhaust gas temperature sensor and a catalyst temperature sensor.

Particular preference is given by a weighting key ( SW ) simultaneously an adaptation of at least one correction value ( W1 ) in or on the reducing agent addition model ( 31 ) and additionally a correction value ( W2 ) in or on the reducing agent storage model ( 32 ) performed.

In the example of 16 will be used at the first implementation ( L1 ) of the learning process the correction value ( W2 ), for example by an offset value the estimated reducing agent content ( cRed ' ) updated. In other words, a weighting key is applied, the 100% adjustment for the correction value ( W2 ).

The representation in 16 is selected purely as an example in order to determine possible courses of the calculated quantities ( cRed * . cRed ' ) in the simulation model compared to the actual changes of the physical quantities and on this basis to explain an example of the multiple execution of a learning process. The local fluctuations of the actual reducing agent content ( cRed ) are purely illustrative and do not reflect the timing or number of single injections.

In the second round ( L2 ) of the learning process, it is determined that the pure adjustment of the estimated reducing agent content ( cRed ' ) was insufficient. With an unexpectedly short time delay, a new efficiency deviation is determined. The efficiency deviation can now on the one hand to a new adjustment of the correction value ( W2 ) can be used to calculate the estimated reducing agent content ( cRed ' ) by an offset value. For this purpose, a weighting of, for example, 50% of the efficiency deviation can be provided. The other 50% of the efficiency deviation can be used for a change of the correction value ( W1 ) can be used to calculate the estimated quantity ( Qi ' ) against the underlying. This compensates for the fact that the injector ( 15 ) maybe opposite to the target value ( Qi * ) too high an actual amount ( Qi ).

After adjusting the estimated reduction center content ( cRed ' ) again a deviation from the target value ( cRed * ) detected. So a negative memory management value ( SCNTR ), by which a reduction of the reducing agent content ( CpRe . CpRe ' ) is achieved.

After the second implementation ( L2 ) it is initially assumed that the estimated reductant content ( cRed ' ) the target value ( cRed * ) corresponds. In fact, however, a lack of reducing agents is beginning to appear, which - with a little more time lag of L2 to L3 - leads to a renewed efficiency deviation of the nitrogen oxide reduction.

In a new learning process ( L3 ) the reductant deficiency is detected and new adjustments are made in the simulation model. In this example, the correction values ( W1 . W2 ) changed. This time, however, with another weighting key of, for example, 75% weighting for the adjustment of the correction value ( W2 ) or the estimated reducing agent content ( cRed ' ) and 25% for the correction value ( W1 ) bszw. the estimated amount of added (Qi ').

After the third round ( L3 ) of the learning process, the system returns to the desired state, so that no inadmissible efficiency deviations of the nitrogen oxide reduction are found.

In the case of the multiple execution of the learning process, results from a previous execution (adaptation was quickly successful / slowly successful / unsuccessful) can be used to determine the adjustments in the current execution. In other words, the selection of a weighting key (SW) in dependence on a time delay between two bushings ( L1 . L2 ) of the learning process or between two determined efficiency deviations of the nitrogen oxide reduction are made. It may also depend on whether between two bushings ( L1 . L2 . L3 ) of the learning process the same or different causes of the efficiency deviation (reducing agent excess or reducing agent deficiency) were found.

• - Sensorisch erfasster Vorlauf-Stickoxid-Gehalt (cPre);
• - Temperatur des Abgases, insbesondere in der Vorlauf-Passage;
• - Drehzahl des Verbrennungsmotors;
• - Strömungsgeschwindigkeit des Abgases;
• - Massenstrom des Abgases;
• - Effizienz der Stickoxid-Reduktion;
• - Geschätzte Beigabemenge (Qi') an Reduktionsmittel;
• - Druck im Ansaugkrümmer;

The execution of a learning process can take place at any time. According to a preferred variant, a learning process is only carried out if a specific precondition for carrying out a correction value adaptation is present. Such a precondition may in particular be that a stationary state of the exhaust aftertreatment system ( 10 ) and / or the internal combustion engine is present. The parameters for determining the presence of a steady state may depend on the type and design of the internal combustion engine or the exhaust aftertreatment system ( 10 ) depend. Preferably, predetermined absolute value ranges and time intervals can be defined as a scale for the parameters mentioned below, which can each be used individually or in any combination to define a stationary state:

• - Sensory flow rate of nitric oxide (cPre);
• - Temperature of the exhaust gas, especially in the flow passage;
• - speed of the internal combustion engine;
• - Flow rate of the exhaust gas;
• - mass flow of the exhaust gas;
• - Efficiency of nitric oxide reduction;
• - estimated addition amount (Qi ') of reducing agent;
• - pressure in the intake manifold;

The correction value adjustment ( 27 ) or the selection of a weighting key (SW) can further alternatively or additionally depending on the cause of a determined efficiency deviation ( deff ) respectively. The cause of the efficiency deviation ( deff ) can be determined in any way. It can at least determine whether there is a reductant deficiency or reductant surplus, ie whether the SCR catalyst is overfilled or run dry. A reducing agent deficiency can be determined, for example, by detecting a breakthrough of nitrogen oxide. A reductant excess can be determined when a reductant breakthrough is detected. This will be explained below with reference to various examples. In addition, however, other causes of a detected efficiency deviation ( deff ), which are detectable in other ways. For example. a reduction in the efficiency of the SCR catalyst ( 16 ) are detected when, over several passes of the learning process, an adjustment of the correction values ( W1 . W2 ) or the estimated reducing agent content ( cRed ' ) and the estimated quantity ( Qi ' ) Although high time delays between the detection of a reductant excess or a reducing agent deficiency are detected, but still large efficiency deviation of the nitrogen oxide reduction ( deff ) compared to the desired value ( Reff * ) are present. Another indication of a reduction in the efficiency of the SCR catalyst ( 16 ), it may be found that, in a method for determining the cause of an efficiency deviation, it is found that despite the specification of a conservative addition scheme, a reaction of the measurement signal of the wake-up nitrogen oxide sensor ( 41 ), which would fit with an aggressive addition scheme, for example because of a trend towards creeping overfilling of the SCR catalyst ( 16 ) is determined.

Hereinafter, embodiments of the method for determining the cause of an efficiency deviation ( deff ) of the nitrogen oxide reduction according to the second main aspect of the present disclosure with reference to FIG 2 to 15 explained. First, with reference to 3 . 4A and 4B Operations in an SCR catalyst under target conditions explained.

3 shows a simplified phase representation of the processes in the SCR catalyst when the nitrogen oxide reduction and the reducing agent addition operate under nominal conditions. Starting from this state, with reference to the 4 to 13 deviating states and their effects on the measured value (cPost) of the wake-up nitrogen oxide sensor ( 41 ). The figures use a uniform representation that includes phase diagrams and reaction diagrams. A phase diagram is always given by the respective number of phases ( P0 . P1 . P2 . P3 ), which determines the order of the phases (but not their duration). The phase diagrams and reaction diagrams include summation representations for operations that take place in an unknown time reference one by one. In the figures, the processes are presented as a progressive sequence and distributed among phases and subsequent reactions to facilitate understanding. Even if the representation of the diagrams in the 3 to 13 arranged one behind the other, this does not mean that the processes also take place in this order in reality.

On the left side of 3 is an initial phase ( P0 ), in which by a flow nitrogen oxide sensor ( 40 ) a flow nitrogen oxide content ( CpRe ) is detected. In the SCR catalyst ( 16 ) is a certain amount of reducing agent as the instantaneous reducing agent content ( cRed ) contain. Further, by the injector ( 15 ) in each phase ( P0 . P1 ) an actual amount of reducing agent ( Qi ) and by the wake nitrogen oxide sensor ( 41 ) is in each phase ( P0 . P1 ) determines a current caster nitrogen oxide content (cPost) as a measured value. The wake nitrogen oxide sensor has a cross-sensitivity for unreacted nitrogen oxides and unreacted reducing agent. For reasons of simplified illustration, it will not be possible between the actual caster nitrogen oxide content and that from the caster nitrogen oxide sensor ( 41 ) distinguished value. From the measured value ( cPost ) Alone is for the controller anyway not detectable if the wake-up nitric oxide sensor ( 41 ) reacted only to nitrogen oxides, only to unreacted reducing agent or to both.

In the 3 to 13 Boxes and circles are shown in different numbers to illustrate the various units of nitrogen oxides and reducing agents present in the SCR catalyst in each phase at a particular location, as well as the change in these quantities due to the particular measures. A unit of nitrogen oxides is represented by a square box. A unit of quantity of reducing agent is represented by a circle. The number of units of measure is used to illustrate the approximate changes in quantity that may occur. In practice, different amounts and proportions may occur.

In the middle diagram of 3 are reactions that are due to the measures in phase ( P0 ) enter. In reality, the reactions take place gradually and at different speeds or dynamics. The reaction diagram shows a summary of the reactions as summation.

In the right diagram is a state in a following phase ( P1 ). The state shown follows on the one hand from the reaction in the SCR catalyst to the measures in phase ( P0 ). On the other hand, in the phase, a renewed instantaneous measurement of the precursor nitrogen oxide content ( CpRe ) as well as a new addition amount ( Qi ) of reducing agents that are essentially independent of the processes in phase ( P0 ) are. The processes of adsorption, reaction (chemical reaction) and desorption shown in the reaction diagram and the consequent change in the reducing agent content ( cRed ) in the SCR catalyst take place with unpredictable dynamics. Thus, the reactions to the measures in the initial phase ( P0 ) already at the beginning of the phase ( P1 ) to be finished. However, it is also possible that some of the operations shown are only after the phase ( P1 ) will take place (not shown for simplicity). The illustrated processes thus show only qualitative processes, without specifying a specific temporal reference. Nevertheless, they help to understand the processes in the SCR catalyst, which can be used to determine the cause of an efficiency deviation.

The (actual) reducing agent content ( cRed ) is below target conditions according to phase ( P0 ) in 3 at the target value ( cRed * ). The actual amount added (Qi) corresponds to the target value ( Qi * ), ie the nominal quantity added ( Qn ). It is assumed in the present case that the nominal quantity ( Qn ) according to the conservative scheme of the estimated reaction quantity ( qr ' ) corresponds. Furthermore, it is assumed that as a result of a correctly working simulation model, the estimated reaction quantity ( qr ' ) corresponds to the actual reaction quantity (Qr).

The reaction diagram of 3 comprises in the upper area a representation of the amount of nitrogen oxide supplied, ie the amount which in the preceding phase ( P0 ) has been detected by the feedstock nitrogen sensor. Below this is the amount of reducing agent ( Qi ) is represented as a block, ie the quantity which was used in the previous phase ( P0 ) through the injector ( 15 ) has been added. The lower part of the reaction diagram shows the stored amount of reducing agent in the SCR catalyst. Black filled boxes in the reaction diagram indicate reacted nitrogen oxides (NR), ie the amount of nitrogen oxides that is reduced by chemical reaction with the reducing agent. Black filled circles indicate the amount of reducing agent actually converted by reaction ( qr ). Unfilled boxes in the reaction diagram indicate the unreacted nitrogen oxides which are correspondingly in the following phase ( P1 ) lead to a nitric oxide breakthrough.

In the example of 3 is the amount (Qi) of reducing agent that is in phase ( P0 ) is completely adsorbed on the catalytic wall in the middle reaction diagram. Thus, the actual adsorption amount ( Qad ) equal to the actual quantity added ( Qi ). Furthermore, in the transition from phase ( P0 ) to the phase ( P1 ) just as much reducing agent ( Qad ) adsorbed, as in the sum consideration by reaction ( qr ) is taken from the SCR catalyst. For this reason, the reducing agent content ( cRed ) in the phases ( P0 ) and ( P1 ) equal.

The efficiency of the reduction of nitrogen oxides in the simplified example is 80%, that is to say of those in phase ( P0 ) units of nitrogen oxide are reacted eight units (NR) and two units (NS) are in the phase ( P1 ) as caster nitrogen oxide content ( cPost ) detected.

In the following, it is assumed purely by way of example that the efficiency of 80% is the desired efficiency of the SCR catalytic converter and, in the illustrations shown, respectively indicates the desired value (rEff *) for the nitrogen oxide reduction. In practice, other values may occur.

As from the illustration in 3 As can be seen, the same stable result can be achieved again and again when a stationary state continues with substantially continuous feeding of ten units of nitrogen oxides and addition of eight units of reducing agent.

4A and 4B show possible reactions to a positive and a negative injection pulse, when it is connected to an SCR catalyst ( 16 ) under nominal conditions. It is further assumed that the reducing agent content ( cRed ) in phase ( P0 ) the target value ( cRed * ) corresponds. Furthermore, it is assumed that according to the conservative addition scheme, the nominal injection quantity ( Qn ) basically the estimated reaction quantity ( qr ' ), which in turn assuming the presence of target conditions with the actual reaction amount ( qr ) matches.

In the example of 4A a positive injection pulse (I +) is performed for test purposes. In the case of the positive injection pulse (I +), the injection quantity ( Qi ) in phase ( P0 ) by a considerable amount ( dQ ) greater than the nominal injection quantity ( Qn ). In the present case, for the sake of simplicity, no distinction is made between estimated and actual values of the addition quantity.

As shown in the following reaction diagram, the supplied 10 units of nitrogen oxides are reacted in the same proportion (NR) as in the illustration 3 although there is more reducing agent in the SCR catalyst than in the state after 3 , The reason for this is that the reduction reaction in the SCR catalyst is primarily dependent on the stored amount of reducing agent ( cRed ) and only indirectly by the amount ( Qi ) being affected. Accordingly, in the example of 4A in the phase ( P1 ) also two unreacted units (NS) of nitric oxide as caster nitrogen oxide content ( cPost ) left over. Since the reducing agent content ( cRed ) the target value ( cRed * ) and thus is less than the maximum storage capacity (cMax), the excess ( dQ ) of added reducing agent (in this case six units of measure) completely in the SCR catalyst ( 16 ) are adsorbed (Qad = Qi). As a result, the positive injection pulse (I +) in phase ( P1 ) not to a noticeable change in the efficiency of nitric oxide reduction. The measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) remains essentially unchanged. This illustrates the buffering effect caused by the storage effect in the SCR catalyst.

4B analogously shows a negative injection pulse (I-), which is carried out under the same starting conditions. So here is in phase ( P0 ) the amount added ( Qi ) of reducing agent by a certain amount (dQ) less than the nominal amount of admixture ( Qn ). As can be seen from the reaction diagram, here too the added quantity ( Qi ) are completely adsorbed to reducing agent (Qad = Qi) and the reduction reaction takes place at the expected (maximum) efficiency, so that the measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) does not change.

From the representations of 3 . 4A and 4B can be derived from a first decision rule, which can be summarized as follows: If a positive or negative injection pulse (I +, I-) in a stationary state, no significant change in the trailing nitrogen oxide content ( cPost ) indicates that the actual reductant content ( cRed ) is not inappropriately deviated from the target value (cRed *). If nevertheless an efficiency deviation ( deff ) of the nitrogen oxide reduction is detected, this could cause other causes than a reducing agent excess or a reducing agent deficiency in question, in particular a sensor drift of the flow nitrogen oxide sensor or the wake-up nitric oxide sensor or a general reduction in the efficiency of the SCR catalyst. This insight can be used alone or in combination with the results of other test procedures to make a suitable change to the simulation model.

As already explained above, the kinetics of the processes in the SCR catalyst and the consequent dynamics of the change in the measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) not sufficiently constant or predictable. For this reason, it is usually not possible to have a clear separation between the phases ( P0 . P1 ) and the intervening reactions. So it is possible, for example, that the adsorption ( Qad ) of the added reducing agent takes place very quickly. In this case, a prompt and easily detectable change of the measured value occurs ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ). In such a case, the interpretations known from the prior art can be used be in 2 as simple decision rules ( 42 ) are shown. These can be summarized as follows:

If a positive (I +) injection pulse causes a (definite) negative change in the measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ), can indicate a reducing agent deficiency ( cRed-- ) getting closed. If, on the other hand, there is a (clear) positive change in the measured value ( cPost ), can indicate a reducing agent excess ( cRed ++ ) getting closed.

With a negative injection pulse (I-) the interpretation is exactly the opposite. If a (unambiguous) negative change in measured value is detected, an excess of reducing agent ( cRed ++ ), while an (unambiguous) positive change in measured value to a reducing agent deficiency ( cRed-- ).

The following are with reference to the 5 to 12 however, other processes in the SCR catalyst that may occur with a reductant deficiency or reductant surplus do not explain the simple decision rules 2 or the state of the art.

The 5 to 8th go each in phase ( P0 ) of a reducing agent deficiency ( cRed-- ) and show possible reactions to a negative reaction pulse (I-) and a positive reaction pulse (I +) with subsequent addition of reagent ( Qi ) according to a conservative addition scheme (Qn = Qr) ( 5 and 7 ) and according to an aggressive addition scheme (Qn = Qs) ( 6 and 8th ).

Shown are each in analogy to the representations of 3 . 4A and 4B an initial situation in the initial phase ( P0 ) with subsequent reaction diagram and subsequent state in phase ( P1 ). In addition, follow-up changes in each case a further reaction diagram and a further phase ( P2 ) P3 ) with the respective effects on the measured value (cPost) of the wake-up nitrogen oxide sensor ( 41 ).

At the bottom left in the figures, in each case for a better overview, it is shown whether there is a positive or negative injection pulse (I +, I-). Furthermore, in the lower bar graph, the absolute measured values ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) in the phases ( P0 ) to ( P3 ). In the bar graph above, the relative changes ( dC ) of the measured value ( cPost ) compared to the pre-pulse value ( CAVG ). The pre-pulse value is the measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) in phase ( P0 ).

5 shows an initial situation in the phase ( P0 ) with reducing agent deficiency ( cRed-- ). A negative injection pulse (I-) is executed. In the reaction diagram between the phases ( P0 ) and ( P1 ), only a small amount of nitrogen oxides (NR) can be reacted because of the insufficient amount of stored reducing agent. A significant proportion (NS) of the nitrogen oxides is not reacted and exits the SCR catalyst. It is correspondingly a significantly higher measured value (cPost) in phase ( P1 ) detectable.

In the phase ( P1 ), the nominal quantity added ( Qn ) of reducing agent through the injector ( 15 ) injected. The reducing agent content ( cRed ) is due to the small addition amount (Qi = Qn-dQ) in the phase ( P0 ) has fallen even further, ie the lack of reducing agent has intensified. Accordingly, in the subsequent phase ( P2 ) an even higher reading ( cPost ) of the wake nitrogen oxide sensor. However, in the phase ( P1 ) again increased addition amount (Qi = Qn = Qr) of reducing agent also to a relative recovery of the reducing agent content ( cRed ) in the SCR catalyst. Accordingly, until the subsequent phase ( P3 ) already reacts with a higher amount of nitrogen oxides (NR), and the measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) falls in phase ( P3 ), but to a value that is higher than under target conditions according to 2 is to be expected. The reducing agent content (cRed) may also differ from the phase ( P2 ) to the phase ( P3 ) have recovered a bit.

6 shows 5 analogous processes assuming an aggressive addition scheme, ie that the nominal addition amount here corresponds to the stoichiometric equilibrium quantity (Qn = Qs).

A comparison of 5 and 6 shows that the results are essentially identical to 5 are. It can only be a slightly faster recovery of the reducing agent ( cRed ) take place in the SCR catalyst, which, however, in the example shown, does not result in a change in the measured values ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ).

7 shows events in response to a positive injection pulse (I +) assuming a conservative loading scheme (Qn = Qr). The starting situation in the phase ( P0 ) corresponds to the illustrations in 5 and 6 ,

In the phase ( P0 ), although an increased amount of reducing agent (Qi = Qn + dQ) is added. However, this does not yet lead to an improvement of the nitrogen oxide reduction because of the time-delayed adsorption. Accordingly, in phase ( P1 ) first an increased measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ). However, the increased loading quantity (Qi = Qn + dQ) leads to a much faster recovery of the reducing agent content ( cRed ) in the SCR catalyst. Accordingly, in the transition to phase ( P2 ), a significantly larger proportion (NR) of nitrogen oxides has already been converted so that in phase ( P2 ) the measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) decreases considerably. However, there is still a reductant deficiency so that in phase ( P3 ) still the substantially same measured value ( cPost ) is present.

8th shows that too 7 analogous scenario assuming an aggressive reductant addition (Qn = Qs). Here is an even faster and somewhat stronger recovery of the reducing agent content ( cRed ), so that in phase ( P3 ) compared to 7 an even lower reading ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) is achieved. Otherwise, the processes are essentially identical.

9 shows in phase ( P0 ) an initial situation with reducing agent excess (cRed ++). A negative injection pulse (I-, Qi = Qn-dQ) is performed, followed by a nominal amount (Qn = Qr ') according to the conservative addition scheme.

Since the SCR catalyst is in phase ( P0 ) is well filled, the reduction of nitrogen oxides in the first reaction diagram can be done at maximum efficiency. Thus, in analogy to the processes in 2 eight units of nitrogen oxide (NR) reacts, whereby the reaction amount ( qr ) of the stored reducing agent is eight units. However, only the two units of reduction agent added in accordance with the negative induction pulse are ready for adsorption. Thus, more reducing agent in the transition from phase ( P0 ) to the phase ( P1 ) degraded in the SCR catalyst than is resumed by adsorption. As a result, the reducing agent excess is reduced. In the phase ( P1 ) the measured value (cPost) of the wake-up nitrogen oxide sensor ( 41 ) two units of measure. In the previous phase ( P0 ), he had carried four units of mass, because there has been detected as a result of the excess reducing agent and a reductant breakthrough.

In the transition to the further phases ( P2 . P3 ) is due to the conservative reducing agent addition essentially the reducing agent content ( cRed ) is kept at a constant level which is lower than in the initial phase ( P0 ). The measured value ( cPost ) thus remains at the low level.

In the example of 10 takes from a surplus of reducing agent ( cRed ++ ), a negative injection pulse (I-) takes place, although an aggressive addition scheme (Qn = Qs) is used.

Due to the negative injection pulse (I-), first in phase ( P1 ) is reduced in analogy to the previous example, the reducing agent excess because the reaction amount (Qr) is greater than the added and thus absorbable amount (Qad = Qi). The reducing agent content ( cRed ) remains at an elevated level, which is why further aggressive addition to phase ( P3 ) again such a reducing agent excess arises that a unit of quantity ( qdc ) in phase ( P2 ) amount added ( Qi ) can no longer be adsorbed and in phase ( P3 ) as reductant breakthrough increases the measured value (cPost) again. Thus, there is a tendency for a higher measured value (cPost) of the wake-up nitrogen oxide sensor ( 41 ), than in the example of 9 ,

11 shows again an example starting from a reducing agent excess in the phase ( P0 ). Here, a positive injection pulse (I +) is carried out using a conservative addition scheme (Qn = Qr ').

Due to the overfilling of the SCR catalyst, a significant part (Qde> 0 ) in the phase ( P0 ) amount added ( Qi ) are not adsorbed (Qad <Qi) and leaves the SCR catalyst in a reductant breakthrough. Accordingly, it comes in phase ( P1 ) to an increase in the measured value (cPost). The following conservative addition (Qi = Qn) is also used in subsequent phases ( P2 . 3 ) a reductant breakthrough occurs due either to incomplete adsorption (of the added amount (Qad <Qi) or to a redissolving or desorbing of an already stored one Quantity unit of the reducing agent is based. Accordingly, even in the phases ( P2 . P3 ) an increased reading (cPost).

12 also shows a reductant excess ( cRed ++ ) in the initial state ( P0 ) and a positive injection pulse (I +). In the following phases ( P1 - P3 ), however, an aggressive addition scheme (Qn = Qs) is used. The sequence is qualitatively similar to the previous example of 11 , However, in the example of 12 in the phases ( P2 . P3 ) an even higher measured value (cPost), which is based on the relatively higher amount added (Qs> Qr). Because in the phase ( P1 ) already clearly overfilled SCR catalyst can no longer absorb the slightly oversized addition quantity (Qi = Qs), so that the desorption amount (Qde) increases even further.

13 shows a special case of the possible processes in the SCR catalytic converter, which was determined using a conservative loading scheme. In the phase ( P0 ) is an initial state in which the reducing agent content ( cRed ) in the SCR catalyst, although clearly above the setpoint shown as a dashed line ( cRed * ) lies. However, the SCR catalyst is not overcrowded yet. This condition can be considered as a moderate excess of reducing agent.

When using a conservative schema for setting the target value ( cRed * ) for the reducing agent content ( cRed ) in the SCR catalyst may enter a state in which the target value for the efficiency of the nitrogen oxide reduction ( Reff * ) is below the maximum efficiency. In the phase ( P0 ), a positive injection pulse (I +) is performed. Surprisingly, according to the reaction diagram, in the transition from phase ( P0 ) to the phase ( P1 ) an increased reduction of nitrogen oxides takes place very early on, the cause of which is still unclear (possibly local reductant deficiency and rapid adsorption, which increases conversion efficiency). Compared to the example of 11 and 12 Thus, the number (NR) of the reacted units of nitrogen oxide increases. Accordingly, by the wake nitrogen oxide sensor ( 41 ) in phase ( P1 ) less nitrogen oxides than in the phase ( P0 ) so that the reading (cPost) decreases. Furthermore, any part of the phase ( P0 ) amount added ( Qi ) does not adsorbed (Qad <Qi), but causes a reductant breakthrough (Qde) (local excess of reducing agent) which leads to an increase in the measured value (Qd). cPost ) leads. In the phases ( P1 . P2 ), the nominal addition amount (Qn = Qr ') is injected according to the conservative addition scheme, which can also lead to a reduction agent breakthrough due to the continuing excess of reducing agent, which in the phases ( P2 . P3 ) to an increase in the measured value ( cPost ) leads.

The illustrated change (dC) of the measured value (cPost) in 13 can be even more pronounced. Due to the unpredictable kinetics of the processes, it can happen that, starting from the state in phase ( P0 ) in the execution of the positive injection pulse (I +) initially a very significant drop in the measurement result (cPost) takes place, while subsequently takes place an even higher reductant breakthrough. This could possibly be due to a temporary storage of data in the phase ( P0 ) be attributed to much injected amounts of reducing agent, initially excessive adsorption takes place, which is subsequently compensated by a creeping desorption.

A comparison of the possible changes (dC) of the measured value (cPost) in the 5 to 13 , which may actually occur with a reductant deficiency or a reductant excess or moderate reductant surplus, with the decision rules 2 clarifies the above-mentioned risk of misinterpretations. A first reaction of the measurement signal (cPost) may (for example due to memory effects in the SCR catalyst) lead to an interpretation that is opposite to the actual state.

14 shows the relative changes of the measured value ( cPost ) for in the 5 to 13 Cases shown as possible responses to a negative injection pulse (I-) and a positive injection pulse (I +) using a conservative supplementation scheme (Qn = Qr ') or an aggressive addiction scheme (Qn = Qs).

In addition, partial answers are considered as successive increases in the measured value ( cPost ), which will be explained in detail below.

15 shows exemplary courses of a feed nitrogen oxide content ( CpRe ), a loading quantity (Qi), a measured value ( cPost ) of the wake-up nitrogen oxide sensor ( 41 ) as well as a response characteristic (R).

In the example of 15 is a stationary state. The flow nitrogen oxide content (cPre) is within a tolerance band (T) during the stationary state. If the lead nitrogen oxide (cPre) content left the tolerance band (T), this could be detected as the steady state end.

As can be seen from the course of the addition amount (Qi), a positive injection pulse (I +) is used as the first injection pulse ( I1 ). The positive injection pulse (I +) may be performed by a single boost or by a contiguous set of injections at an increased injection rate. A pre-pulse value ( CAVG ) of caster nitrogen oxide content ( cPost ) is detected, which is a measured value of the wake-up nitrogen oxide sensor ( 41 ) before or at the beginning of the execution of the first injection pulse ( I1 ). Preferably, the pre-pulse value (cAvg) may be an average value that is determined in a time window before the execution of the injection pulse (FIG. I1 ) is detected. This time window is in 15 marked as averaging zone (Zavg).

The measuring signal (cPost) of the caster nitrogen oxide sensor ( 41 ) is in a (indeterminate duration) period after the injection pulse or from the beginning of the injection pulse (I +) and before any further injection pulse ( I2 ) with several separate partial answers ( R1 - R13 ) rated. The partial answers can be set according to any evaluation scheme to record a successive relative change of the measured value (cPost). According to a preferred embodiment, a first partial response ( R1 ) when the measurement signal ( cPost ) increases or decreases by a certain amount from the pre-pulse value (cAvg). Each further partial answer ( R2 - R13 ) is set when the measuring signal ( cPost ) compared to the value of the measuring signal ( cPost ) at the end of the previous partial answer ( R1 ) increases or decreases by a certain amount, in particular the same amount.

In the example of 15 are at each successive falling change of the measurement signal ( cPost ) a negative partial answer ( R1 - R5 ) and with each successive positive change of the measuring signal ( cPost ) a positive partial response ( R6 - R13 ) set.

According to the same evaluation scheme are in the 5 to 13 Representative profiles of the partial answers for the operations shown in the SCR catalyst shown. The representative progressions of the partial answers are also in the decision patterns in 2 and 14 shown.

From the evaluation of the measuring signal ( cPost ) with partial answers, an easily analyzable qualitative change of the measuring signal ( cPost ), which is largely independent of the dynamics of the change. From the sequence, the distribution and possibly the ratio of negative and positive partial answers, a pattern can be identified, which can be evaluated to determine the cause of the efficiency deviation.

Preference is given to the multiple partial answers ( R1 - R13 ) with decision patterns ( 42 ) (cf. 2 and 14 ), the representative courses of partial responses and / or changes (dC) of the measurement signal (cPost) of a cause (reducing agent excess ( cRed ++ ), Reducing agent deficiency ( cRed-- ) assign. From the pattern comparison, even when only a single injection pulse ( I1 ) or less injection pulses ( I1 . I2 ) Interpretations on the cause of an efficiency deviation over the presence of a reductant excess or a reductant deficiency. The results of the pattern comparison are usually more accurate and reliable than those obtained with previously known analysis methods. For example. The captured pattern may conform to an aggressive add-on scheme, although the controller provided a conservative addition scheme. This may indicate that the determination of the nominal quantity ( Qn ) is erroneous, ie at a value above the reaction rate ( qr ) lies. For compensation, a correction value adjustment in the precontrol can be carried out, if necessary, paired with a change in the estimated reducing agent content ( cRed ' ) by an offset value.

The analysis period for the evaluation of the measurement signal ( cPost ) after or from the beginning of an injection pulse ( I1 ) need not be fixed and in particular not to be divided according to time rules in phases. Rather, it is possible to carry out a valuation until either the stationary state ends, a pattern comparison becomes permanently impossible due to excessive deviations, or a renewed injection pulse (FIG. I2 ) is performed. The evaluation period can therefore be variable and, for example, up to 15s (Seconds) or 30s or even longer. Thus, even particularly slow adsorption and desorption and thus particularly strong memory effects of the SCR catalyst can be evaluated with.

On the other hand, it is possible, already on the basis of one or more earlier partial answers ( R1 - R5 ) a preliminary ruling ( D1 ) about the cause. This can be done in particular if initially one of the simple decision rules according to 2 a high agreement with the response characteristic (R). If the in-patient condition ends early, the preliminary ruling can be taken as a final decision, which may be a correct decision for a significant number of cases.

If the stationary state persists and further partial answers ( R6 - R13 ) are no longer consistent with the decision making D1 ), a different final decision ( D2 ) if, in addition, a match to one of the other decision patterns ( 42 ) is detected. This is exemplary in 15 shown. In the example, the first five partial answers ( R1 - R5 ) matches a decision pattern 2 on. From the partial answer ( R6 ) the match is lost. From about the partial response (R9 / R10), a new match with a decision pattern is made 14 that determines to the in 13 heard operations described. Accordingly, the final decision ( D2 ) that the identified cause in the present example is a moderate excess of reducing agent (cRed> cRed *).

A decision pattern ( 42 ) can therefore provide that one or more early partial answers ( R1 - R5 ) in a first direction and one or more late partial answers ( R6 - R13 ) in the opposite direction. If appropriate, ratios for the respective number of partial answers in the first direction and in the opposite direction can be provided.

In addition, a decision pattern may provide that the measurement signal ( cPost ) in the range of at least one late partial response ( R6 - R13 ) in the opposite direction to the one or more partial answers ( R1 - R5 is present to a limit (K) of the pre-pulse value ( CAVG ), where the deviation must be in the direction of the late partial response. In other words, in some cases, if the measurement value falls early ( cPost ), which leads to a correspondence with a first decision pattern, a change to another decision pattern only takes place when, in the course of such a strong increase of the measured value ( cPost ), that the measured value ( cPost ) by a definable limit (K) above the level of the pre-pulse value ( CAVG ) - and vice versa.

Optionally, one or more subsequent injection pulses ( I2 . I3 ), preferably within the same stationary state, wherein the test loading amount (Qi) or the direction of the following injection pulse (I2 = I or I2 = I +) and optionally the duration of the following injection pulse ( I2 ) depending on the result of the previous injection pulse ( I1 ) be determined. Thus, for example, after a first positive detection of a cause, a targeted counter-sample or a specific confirmation sample can be carried out. As a result of the targeted specification, the number of injection pulses required overall can be significantly reduced, so that the risk of generating a reducing agent breakthrough by the test method is minimized.

Modifications of the invention are possible in various ways. In particular, all the features described, shown, claimed or disclosed in any other way may be combined with one another in any desired manner, replaced or omitted from one another.

The internal combustion engine may preferably be a direct injection diesel engine.

The determination of the pre-tax value ( Qpre ) can be done in any way. It may preferably be adapted as a function of an addition scheme, in particular according to a conservative addition scheme. Then the precontrol value ( Qpre ) is set to be lower than the stoichiometric balance amount. In other words, in a conservative addition scheme, a little less reducing agent is added on the basis of the pilot value (Qpre) than would be required for the complete reduction of the amount of nitrogen oxide which according to the feedstock nitrogen oxide content ( CpRe ) currently the SCR catalyst ( 15 ) is supplied. In a conservative addition scheme, the pilot control value (Qpre) can still be below the estimated reaction quantity (Qpre). qr ' ).

Alternatively, the precontrol value ( Qpre ) according to a balanced addition scheme correspond to 95% to 100% of the stoichiometric equilibrium amount, which according to the current flow-nitric oxide content ( CpRe ) is determined. In the case of an aggressive adding scheme, the pre-tax value ( Qpre ) are at or above the stoichiometric equilibrium amount, which may also require that an ammonia slip catalyst or similar component downstream of the SCR catalyst ( 16 ) is arranged.

The addition scheme may be adapted depending on an operating condition of the internal combustion engine and / or the exhaust aftertreatment system. For example, a conservative addition scheme can be selected when a steady state condition is established. On the other hand, if a transient condition is detected in the direction of an increasing load or in the direction of increased combustion temperatures, a balanced or aggressive delivery schedule may be temporarily provided.

LIST OF REFERENCE NUMBERS

10 aftertreatment system Exhaust gas after treatment system 11 Forward-Passage Upstream passage 12 Catalysis section Catalytic section 13 Trailing Passage Downstream passage 14 Diesel oxidation catalyst Diesel oxidation catalyst 15 Reducing agent injector Reductant injector 16 SCR catalyst SCR Catalyst 17 filter filter 18 Forward-NOx sensor Upstream NOx sensor 19 Trailing NOx sensor Downstream NOx sensor 20 control unit Control Unit 21 Multi-unit simulation model Multi-segment simulation model 22 feedforward Pre control 23 Setpoint calculation Target value calculation 24 Injector driver Injector Driver 25 Summation member Summation segment 26 comparator Comparator 27 Weighting element / correction value adjustment Weighting segment / correction value adaptation 28 condition check Condition check 31 Reducing agent adding model Reductant addition model 32 Reductant storage model Reductant storage model 33 Response model Reaction model 34 Desorption model Desorption model 40 Forward-nitrogen oxide sensor Upstream nitrogen oxide sensor 41 Tracking nitrogen oxide sensor Downstream nitrogen oxide sensor 42 decision-making patterns Decision pattern CAVG Pre-pulse value / pre-pulse mean value Pre-pulse value / pre-pulse average value CpRe Forward-nitrogen oxide content Upstream nitrogen oxide content cPost Tracking nitrogen oxide content Downstream nitrogen oxide content cRed Actual reducing agent content Actual reductant content cRed ' Estimated Reductant Content Estimated reductant content cRed * Target value for reducing agent content Target value for reductant content deff Efficiency deviation of the nitrogen oxide reduction Efficiency deviation of nitrogen oxide reduction Reff * Target value for nitrogen oxide reduction Target value for nitrogen oxide reduction Cond Prerequisite for learning / correction value adjustment Precondition for learning cycle / correction value adaptation CQ Injection command Injection command D1 preliminary ruling Preliminary decision D2 final decision Final decision I + Positive injection pulse Positive injection pulse I- Negative injection pulse Negative injection pulse K Limit value (compared to cAvg) Threshold (wrt cAvg) NOx Nitrogen oxides Nitrogen oxides NO Reacted nitrogen oxides Reacted nitrogen oxides NS Unreacted nitrogen oxides / nitric oxide breakthrough Non-leached oxides / Nitrogen oxides break through P0-P4 phases Phases Qad Actual adsorption amount Actual adsorbed quantity Qad ' Estimate of adsorption amount Estimated value for adsorbed quantity qdc Actual desorption amount Actual desorbed quantity qdc ' Estimated desorption amount Estimated desorbed quantity Qi Actual reductant addition amount Actual Reductant addition quantity Qi * Setpoint for reducing agent addition amount Target value for reductant addition quantity Qi ' Estimated amount of reductant added Estimation value for reductant addition quantity Qpre Precontrol value for reducing agent addition pre control value for reductant addition qr Actual converted by reaction amount of reducing agent / reaction amount Actual quantity of reductant converted by chemical reaction / reaction quantity qr ' Estimated reaction quantity Estimated value forquality qs Stoichiometric equilibrium amount Stoichiometric balance quantity SCNTR Storage management value for reducing agent addition Storage guidance value for reductant addition R Response characteristic Response characteristic R1 separated partial answers Separated section R13 on injection pulse responses to injection pulse S1 First stationary condition First steady state S2 Second stationary state Second steady state S3 Third stationary state Third steady state SW weighting key Weighting ratio T tolerance band Tolerance band W1 First correction value First correction value W2 Second correction value Second correction value W3 Third correction value Third correction value W4 Fourth correction value Forth correction value Z avg Averaging zone Averaging zone

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• US 2003/0046928 A1 [0025, 0026, 0027, 0028, 0034]

Claims (11)

An exhaust aftertreatment system (10) for an internal combustion engine comprising an SCR catalyst (16) having a flow passage (11) and a wake passage (13), wherein a current flow rate nitric oxide content (cPre) in the flow passage (11 ) and an instant caster nitrogen oxide content (cPost) in the wake passage (13), and a reductant injector (15) is disposed upstream of the SCR catalyst (16) for injecting a reductant contained in the tail gas SCR catalyst (16) is stored and nitrogen oxides (NOx) reduced, and wherein the exhaust aftertreatment system (10) comprises a control unit (20) which is connected to the reducing agent injector (15) and controls the current Beigabemenge (Qi) of the reducing agent , characterized in that the control unit (20) comprises a multi-unit simulation model (21) and a feedforward control (22) and is designed to carry out the following steps: - In the control unit (20), a setpoint value (Q i *) for the current reducing agent addition amount based on a pilot value (Qpre) and a storage guide value (Qstr); - The pilot control value (Qpre) is determined by the feedforward control (22) on the basis of the feedstock nitrogen oxide content (cPre); the storage management value (Qstr) is determined by the multi-unit simulation model (21) as a function of an estimated reducing agent content (cRed ') in the SCR catalytic converter (16) such that the estimated reducing agent content (cRed') is equal to a target Value (cRed *) is approximated; the simulation model (21) comprises at least a first member in the form of a reducing agent adding model (31) and a second member in the form of a reducing agent storage model (32) on the basis of which the reduction agent content (cRed ') is estimated; an actual nitrogen reduction efficiency is determined on the basis of the lead nitrogen oxide content (cPre) and the clogged nitrogen oxide content (cPost); in a learning process, when an efficiency deviation (dEff) is determined in relation to a desired efficiency (rEff *) of the nitrogen oxide reduction, a correction value adaptation (27) is made in at least two members (31, 32, 33, 34) of the simulation model ( 21). Exhaust after-treatment system according to Claim 1 wherein the efficiency deviation (dEff) is distributed according to a weighting key to the correction value adjustment (27) in the at least two members (31, 32, 33, 34) of the simulation model, in particular by dividing the deviation value (dEff) into two or more shares. Exhaust after-treatment system according to Claim 2 wherein two or more alternative weighting keys (SW) are predefined, between which, in particular, a change is made depending on a success of a previous learning operation. An exhaust after-treatment system according to any one of the preceding claims, wherein a correction value adjustment (27) alters at least one correction value (W1) on or in the reducing agent adding model (31) and additionally a correction value (W2) on or in the reducing agent storage model (32). An exhaust aftertreatment system according to any one of the preceding claims, wherein the reducing agent adding model (31) calculates an estimated addition amount (Qi ') of reducing agent in response to an operation (Qi * / CQ) of the reducing agent injector (15). The exhaust aftertreatment system of any one of the preceding claims, wherein the reductant storage model (32) estimates which amount (Qad ') of reductant is currently stored in the SCR catalyst (16) by adsorption depending on state parameters of the exhaust aftertreatment system (10). Exhaust after-treatment system according to one of the preceding claims, wherein the target value (cRed *) for the reducing agent content in the SCR catalytic converter (16) and / or the pilot control value (Qpre) is determined according to a conservative addition scheme. Exhaust after-treatment system according to one of the preceding claims, wherein it is checked as a precondition (Cond) for carrying out a correction value adaptation (27) that there is a stationary state of the exhaust aftertreatment system (10) and / or of the internal combustion engine. Exhaust after-treatment system according to one of the preceding claims, wherein the cause of an efficiency deviation (dEff) is determined, namely whether the deviation is based on a reduction agent deficiency ( Nitrogen oxide breakthrough) or reductant excess (reductant breakthrough), and the correction value adjustment (27) is made depending on the cause of the efficiency deviation, in particular by the selection of a weighting key. Exhaust after-treatment system according to one of the preceding claims, wherein a reaction model (33) and / or a desorption model (34) are provided as further members of the simulation model (21), the reaction model (33) being dependent on state parameters of the exhaust aftertreatment system (10 ) estimates what proportion (Qr ') of the reducing agent stored in the SCR catalyst is converted by chemical reaction, and the desorption model (34) in dependence on state parameters of the exhaust aftertreatment system (10) estimates what proportion (Qde') of the in the SCR catalyst stored reducing agent is desorbed without chemical reaction. Exhaust after-treatment system according to one of the preceding claims, wherein a correction value adjustment additionally changes a correction value (W3) for or in the feedforward control (22).

Images