DE102009016509A1 - Method for adjusting mass flow in exhaust gas recirculation process in diesel engine in passenger car, involves utilizing model-assisted predictive automatic controller for regulating virtually determined nitrogen oxide value - Google Patents

Abstract

The method involves utilizing a model-assisted predictive automatic controller for regulating a virtually determined nitrogen oxide value, where a nitrogen oxide controlling system determines virtual nitrogen oxide and controls real nitrogen oxide. A controlled variable of the nitrogen oxide controlling system is determined by a virtual nitrogen oxide sensor. A virtual nitrogen oxide dependent controlled variable is compared with the nitrogen oxide desired value determined from a characteristic diagram. An independent claim is also included for a diesel internal combustion engine comprising a diesel particle filter.

Description

The The present invention relates to a method for adjusting a Exhaust gas recirculation of an internal combustion engine taking into account a NOx behavior and a Diesel internal combustion engine with at least one exhaust gas recirculation.

It is known that for a reduction of nitrogen oxide emissions in diesel engines, a portion of an exhaust gas returned becomes. This should be an oxygen concentration at the engine entrance be reduced. This exhaust gas recirculation influences by the height of an exhaust gas recirculation rate including a charge air temperature as well as a boost pressure, in particular for large commercial vehicles by charging is produced. Especially for large commercial vehicles the desire exists that nitrogen oxide emissions as well as particulate emissions can be measured directly and also regulated. While NOx sensors are therefore already in series production, However, particle sensors are still in the developmental stage. A control based on a signal from the NOx sensor is possible; The problem here, however, may be that the sensor signal thus generated under dynamic operating conditions, not the current nitrogen oxide value but only indicate a time-delayed nitrogen oxide value can.

From the WO 2008/131789 A1 For example, a method for adjusting a mass flow of an exhaust gas recirculation of an internal combustion engine taking into account a NOx behavior is known. The scheme underlying the method provides for a coupling of a virtual NOx determination with a real NOx control. The control advantageously takes into account a time delay by a NOx sensor.

task The present invention is intended to improve this process and continue to evolve.

These The object is achieved by a method having the features of the claim 1 and with a diesel internal combustion engine with the features of claim 16 solved. Further advantageous embodiments and further developments are in the respective subclaims specified.

It It is proposed that a method for adjusting an exhaust gas recirculation an internal combustion engine taking into account a NOx behavior is executed, wherein a control is a coupling of a virtual NOx determination with a real NOx control provides. It is proposed that the scheme be modeled predictive controller for regulating a virtually determined NOx value uses.

Of the model-based predictive controller, below Called MPR, uses a dynamic model of the process to be controlled, by one or more controlled variables within that Predict prediction window. The regulatory goals, z. As regards the behavior of the controlled variable or controlled variables, are in a so-called Formulated cost function, which deposited next to the model in the MPR is. As part of an optimization, the future manipulated variables the controlled system within the cost function so determined that the cost function preferably assumes a minimum. To errors in the Model and disturbances affecting the process to wear, is the model deposited and used in the controller the control system continues to be preferred with an observer the real Process tracked.

The Cost function preferably provides a flexible quality function which directly relies on information that from the model of the controlled system carried in the controller be won. The use of the MPR leads additionally to a calming of the manipulated variables. The MPR allows a single or multi-variable control in a state space representation, which opens up a possibility to further rule, Actuating or influencing variables in the same structure integrate, so z. B. a fuel path and / or an air path.

The usual attitude necessary for a good regulatory behavior the controller parameter of conventional controllers, for example a PID controller, Can be quite difficult, time consuming and costly be. Furthermore, it is often not or conventional controllers only possible via maps, other process variables to be included in the scheme. Compliance with limits for individual Sizes are only by one on stationary measurements based map for limiting a manipulated variable to realize.

In contrast to that from the WO 2008/131789 A1 Known method, according to which a PID controller is used to control a virtually determined NOx value, the model-based predictive controller allows a systematic controller setting for the present fast engine process, a high dynamics, based on the exhaust path dead time, measurable interference and Has limitations of the controlled variables and / or manipulated variables.

This method can be carried out particularly advantageously in a special diesel passenger car internal combustion engine, which is also proposed. This has at least one exhaust gas recirculation, a boost pressure tion, a diesel particulate filter, a catalyst, a NOx sensor and a first control with respect to a recirculation rate of an exhaust gas flow for adjusting a NOx value. A first control comprises first control means simulating a virtual NOx sensor, second control means performing an adapted control of the virtual NOx sensor, and third control means implementing a NOx control, the first control being constructed such that the virtual NOx sensor presets a default for the first control. The first control is designed as a control of the type described above.

through of the proposed procedure is intended that a scheme allows an exhaust gas recirculation mass flow becomes. It is preferably an exhaust gas recirculation mass flow used as a controlled variable. The regulation of the exhaust gas recirculation flow has the advantage of faster regulation compared to one Setting a charge pressure. A setting of the supercharging pressure However, it can be integrated as part of an overall scheme. about the adjustment of the exhaust gas recirculation mass flow For example, an oxygen content in the intake manifold of the internal combustion engine determined, preferably measured at an inlet region in the Internal combustion engine. A virtual sensor, preferably a corresponding NOx model, preferably analyzes a virtual Oxygen content, which, for example, by an adapted Value is previously corrected, reducing to a virtual NOx value can be closed in the exhaust.

A Adaptation becomes, for example, in relation to the used NOx model executed. This preferably has adaptive components on. Another adaptation can, for example, in the field of Determining a mass flow done. For example, a Mass loading model are adapted. Preferably, that provides Mass load model on a cylinder mass. By its adaptation results in a more accurate determination of the necessary values in particular compared to an otherwise derived from total measurements individual cylinder filling. For example, an exhaust gas recirculation model used, this may according to one embodiment also have an adaptation. It is also possible also a particle model, for example a particle filter model to integrate. This gives the possibility, for example to set a particle concentration in the exhaust gas. Preferably considered a particle concentration in the exhaust gas in the scheme and thus can adapt to a NOx value to be set to lead. The proposed particle model can according to a Training also take into account a particulate filter loading. This can be a strategy for the regeneration of a particulate filter be won. For example, can be calculated via the model when regeneration is to be performed under NOx aspects. There is also the possibility of smoke development be able to turn off by particles during operation. For a smoke value, a particle load of a particulate filter like also for a particle concentration in the exhaust gas, the model or the regulation has specified limits, taking into account Find.

Advantageous it is that through the proposed combination of a nitric oxide control to a virtually calculated fast NOx signal and an adaptation a virtual nitrogen oxide signal via a NOx sensor transient benefits of a virtual NOx control with an advantage an increase in the accuracy of a direct control is made possible on the NOx signal of a NOx sensor. For this It is further proposed that a controlled variable of NOx control is determined by means of a virtual NOx sensor. Furthermore, it is advantageous if a virtual NOx-dependent Controlled variable with one determined from a map NOx setpoint is compared. As a manipulated variable for a virtually determined NOx control variable For example, an exhaust gas recirculation mass flow used, hereinafter referred to as EGR mass flow. Another acceleration the controller behavior is possible because the control relies on an exhaust gas recirculation model. The exhaust gas recirculation model can be one or several times, in particular depending on whether it a low pressure as well as a high pressure exhaust gas recirculation at the internal combustion engine there. Preferably, the respective Exhaust gas recirculation model of the control in one quasi stationary operating state of the internal combustion engine adjusted by means of a signal of the NOx sensor. This has the Advantage that the exhaust gas recirculation model thus can also be self-learning. On the one hand necessary adjustments are made during the comparison, on the other hand preferably also increases a base of learning values. On The basis of these learning values is the exhaust gas recirculation model interpolate as well as extrapolate. The exhaust gas recirculation model can For example, on simulation techniques of neural network technology, on Fuzzi models as well as in particular on equation systems turn off, based on the internal combustion engine, the existing or connected components and over Balance limits of certain sizes.

It has proved to be advantageous if a virtually determined NOx control variable is adapted within the framework of an adapted control. For this purpose, it is preferably provided that the adapted control uses a real NOx sensor. However, the adaptation can, for example, at fast load be suspended. It has been found that often the dynamics is too high, as that the adapted control would be able to allow a reasonable adaptation of the virtual determined NOx controlled variable. According to one embodiment, however, it is provided that an adaptation is carried out even with fast load change reactions. This can be carried out, for example, in a mirrored system and can be checked for usability of the adaptation following the load change or cycles. Thus, for example, an adaptation to transients for the actual control can be suspended. By comparing the values determined without adaptation and the virtual values obtained in the mirrored system with adaptation, however, it can be provided by appropriate learning algorithms that a quality of the mirrored system is created, so that with a minimum quality, even with fast load changes, the control adaptation by application of the matched model found in the mirrored system. The adaptation as a learning function preferably uses a real NOx sensor, but can also rely on other sensors or data.

A Acceleration of the procedure results for the regulation, if the scheme is an inner and outer cascade operates. The inner cascade preferably acts on one Lambda probe back while the outer Cascade preferably relies on a real NOx sensor. By using the lambda probe in the inner cascade becomes a faster signal flow allows. The lambda probe is less sluggish than those currently on the market real NOx sensors. The lambda probe is used in particular to carry out an adjustment of an air determination. So can For example, the exhaust gas recirculation model a Provide airflow at different locations, respectively. The lambda probe can be used to reconcile these virtually determined values be used. The outer cascade becomes particular used for a comparison of a NOx determination. For this can the values determined by the real NOx sensor with those compared in the exhaust gas recirculation model or be used or determined by the virtual NOx sensor. In particular, can be ensured with the outer cascade be that the modeled values verifiable stay.

Next This embodiment has also proven a cascade, where the outer cascade reacts faster as the inner cascade. In this case, for example, a real NOx sensor to enter in the inner cascade while virtual detected Values, preferably oxygen values, in the outer Enter cascade.

Next a cascade control is still possible that in addition to or instead of the cascade control a pre-regulation is provided. Here, for example, the modeled virtual signal a first adjustment of the NOx value then submit the relevant provision under The values of the real NOx sensor are further treated. Furthermore, a manipulated variable connection, for example, an auxiliary control variable switched as well as an auxiliary control variable can be switched on would. There is also the possibility of a follow-up regulation to provide with feedforward control, in particular with a connection the derivation of the respective reference variables.

A further embodiment provides that a lambda probe is arranged in the air path, for example in the intake manifold. Thus, an oxygen content can be determined by measurement, which results before the internal combustion engine. This can still be calculated via a model, but not necessarily. If a model is used for determining oxygen, for example as a virtual oxygen sensor, this can also be adapted with the values of the lambda probe. Furthermore, a special lambda probe in the air path or in the exhaust line can be used, which is particularly suitable for use in cold start. These may preferably be broadband lambda probes, in particular improved broadband lambda probes as they are currently under development. The lambda probe may, for example, have a heating element. It can for example be constructed and / or operated as it is from the DE 10 2004 057 929 A1 which is referred to in the context of the disclosure.

In a preferred use of the method, in particular on a diesel passenger car internal combustion engine, the first control is constructed, for example, as a higher-level outer control cascade having a second, inner control cascade with a faster control time than the outer control cascade. With respect to this passenger car internal combustion engine, it is further preferably provided that the determination means for determining an oxygen concentration at an engine inlet of the diesel internal combustion engine and for determining an oxygen content of a recirculated exhaust gas are provided. In this way, there is the possibility that the model can be compared or by direct determination of correlations, a correlation of a nitrogen oxide concentration in the exhaust gas can be calculated. Such a correlation is evident, for example, from the dissertation by OE Hermann at RWTH Aachen University. This dissertation is titled "Emission Control in Commercial Vehicle Engines via the Air and Exhaust Gas Path". On this dissertation is referenced for correlation within the scope of this disclosure. The same applies to the basic structure of a control with respect to a signal of a real NOx sensor, as can also be seen from this dissertation. In particular, reference is also made to an EGR controller, which is also described there.

Preferably It is envisaged that the adaptation to adapt one or more Models of the scheme for comparing virtually determined values the models in the regulation of the diesel passenger car internal combustion engine is provided for this purpose with a signal flow from the lambda probe as well as being connected by a NOx sensor. This allows in particular a constant comparison and by utilization the learning function improved behavior of the diesel commercial vehicle internal combustion engine. A further development provides that in particular by the model and the learning function determined data can also be read. This is the case with a majority of the same diesel passenger car internal combustion engines executed, this data can be merged and by a corresponding preparation, in particular in each case against each other weighted weighting merged into a single record become. This record can then be used as a default in new diesel passenger car internal combustion engines be deposited.

advantages and further features of the invention are described below with reference to the Drawings explained in more detail. There described However, features are not limited to the particular embodiment shown. Nor are the figures to be construed restrictively. Much more can use the features shown there with other features in other embodiments as well as those of those described above Characteristics of unspecified further training be linked. Show it:

1 a schematic view of a diesel passenger car internal combustion engine with actuators and sensors;

2 : a schematic overview of an adaptation of an EGR mass flow via a lambda probe,

3 FIG. 2: a schematic view of an adaptation of a NOx model via a NOx sensor, FIG.

4 FIG. 2: a schematic view of a mass flow determination by means of a mass-loading model, and FIG

5 : A schematic representation of a model-based predictive controller.

1 shows a schematic representation of an internal combustion engine, in particular a diesel passenger car internal combustion engine 1 with respective associated aggregates, sensors and actuators. The diesel passenger car internal combustion engine has a high pressure exhaust gas recirculation 2 and a low pressure exhaust gas recirculation 3 on. In an air supply 4 to the diesel passenger car internal combustion engine 1 Various sensors or devices are used. Sensor locations or sensors are in the 1 with outlined numbers again clarified. If an air through the diesel passenger car internal combustion engine 1 sucked in, this can be directly on entry through an air mass sensor 5 , in particular a Heißfilmuftmassensensor be measured. Is, as shown, the low-pressure exhaust gas recirculation 3 existing, additional exhaust gas is supplied from there. This may require that another air mass meter 5 is provided. In the low-pressure exhaust gas recirculation 3 is preferably next to a control valve 6 also a cooler 7 arranged. Thereby, the recirculated exhaust gas is cooled to a temperature such that a compressor 8th an exhaust gas turbocharger is able to provide a sufficient air mass flow compressed the internal combustion engine to provide. The compressor 8th downstream is preferably in turn a cooler 7 in order to dissipate the resulting in the compression temperature in the gas can. The thus cooled gas stream can via the high-pressure exhaust gas recirculation 2 then further exhaust gas, which has been cooled by a corresponding radiator, through another control valve 6 be supplied. At an engine entrance 9 Preferably, further sensors are arranged which receive parameters for the control. An exhaust from the diesel passenger car internal combustion engine 1 can then be removed, with a partial mass flow of the high-pressure exhaust gas recirculation 2 or the low-pressure exhaust gas recirculation 3 is supplied. Furthermore, a main flow of the exhaust gas via a turbine 10 the exhaust gas turbocharger used to the compressor 8th drive. Instead of an exhaust gas turbocharger, another charge pressure charging can be performed. For this purpose, for example, a mechanical loader or other means can be used. The turbine 10 downstream is a diesel particulate filter 11 as well as a catalyst 12 , For the sake of simplicity, only the diesel particulate filter 11 shown. Furthermore, a NOx sensor as well as a lambda probe are also arranged in the exhaust gas system.

From 1 Outgoing Diesel Passenger Car Internal Combustion Engine 1 is used in the proposed method as follows: The EGR control concept preferably provides the combination of Hock shown however, may alternatively be implemented with a separate high pressure EGR or low pressure EGR. In the exhaust gas is after the turbine 10 measured with a NOx sensor at position 2 in the circle an exhaust gas concentration. Via an exhaust gas pressure sensor, position 3 in the circle, and an exhaust gas temperature sensor, position 4 in the circle, a state of the exhaust gas, in particular a density, before the control valve 6 , the high-pressure EGR valve measured. A boost pressure P2 is measured with a boost pressure sensor at position 6 in a circle. In the control valve 6 the high-pressure exhaust gas recirculation 2 a position SEGR of the valve is determined. By way of a density of the exhaust gas upstream of the valve and using the differential pressure from P3 and P2 above the valve, an EGR mass flow can then be calculated. Furthermore, an air effort model not described here is available. This is the boost pressure P2 and a Saugrohrtemperatur T2, positions 6 and 7, respectively in a circle, fed. From this, the air effort model can calculate an engine mass flow. In the case of the presence of low pressure EGR and high pressure EGR as shown, it is necessary that the low pressure EGR mass flow is also calculated or measured for this. If, on the other hand, no low-pressure EGR is present, the EGR rate and the fresh air mass flow can also be calculated via the air effort model. A low-pressure EGR mass flow is determined by a differential pressure measurement DP at position 8 in a circle. For this purpose, a pressure drop is preferably determined via an orifice in the exhaust gas line of the low-pressure EGR path. Alternatively, an air mass meter may be used before low pressure EGR delivery, position 10 in the loop, and air mass meter after low pressure EGR supply, position 11 in the circle. From this, the low pressure EGR rate can be calculated. If an air mass meter is present, it may be possible to dispense with the model of a high-pressure exhaust gas recirculation and calculate a high-pressure exhaust gas recirculation rate from a fresh air mass flow, measured at position 11 in the circuit, and an engine mass flow of an air consumption model. For example, an exhaust gas recirculation rate can also be determined via a model, as for example from the DE 102 42 234 which is referred to in this regard.

By means of the sensors and models described above, a calculation of the exhaust gas recirculation rate and using an oxygen content of the recirculated exhaust gas, a calculation of an oxygen concentration of the internal combustion engine gas supplied as a target allows. In this case, an oxygen content of the recirculated exhaust gas can be determined based on a lambda signal of the NOx sensor, for example, at position 2 in the circuit. About an oxygen concentration at the engine inlet 9 can be calculated via correlations, which are described in the above-mentioned thesis, a nitrogen oxide concentration in the exhaust gas. With regard to this correlation, reference is made to the dissertation within the scope of this application. With the structure provided, it is thus possible to operate an adaptive NOx regulator with high-pressure and low-pressure exhaust gas recirculation. For this purpose, a virtual NOx signal is determined by means of a model. This is used as a controlled variable and compared with a NOx setpoint. This setpoint is preferably obtained from a map and indeed as a function of an engine speed and engine load. A so-called model-based predictive controller, hereinafter referred to as MPR, is then used to control a deviation of the virtual NOx signal from the setpoint. A manipulated variable of the MPR for this purpose is a desired EGR mass flow. This can be converted with the EGR model into a corresponding desired EGR valve position.

The Among other things, the EGR model considers the condition at the respective EGR valve and can thereby compensate for example, a changing pressure in front of the turbine provide. An internal position control of the EGR valve regulates a position of the valve and reports the actual Position back to the EGR model, which in turn the current actual EGR mass flow calculated. An EGR controller structure Thus, an EGR model, a NOx model as well as connected the Internal combustion engine and corresponding data streams provide in between. So the pressures go P2, P3 as well the temperature TEGR and the position value SEGR in the EGR model one. In an air effort model again go the temperature T2, a value of the lambda sensor and the pressure P2. From the EGR model is the determined via the model EGR mass flow to Provided. The air effort model calculates from this Further values, in particular a mass flow, of the internal combustion engine is supplied, an exhaust / air ratio and also values of the NOx sensor. From this the NOx model determines virtual NOx signal. This will be provided to the MPR, wherein the MPR associates the virtual NOx signal with a desired NOx value. The as input signal desired NOx value preferably results from a Map. From this, the MPR determined a mass flow of the exhaust gas recirculation, from which turn on the AGR model turn the travel of the respective Exhaust gas recirculation valve results.

From 1 resulting in a virtual EGR rate, referred to therein as the "virtual EGR rate" thus calculable over a mass flow balance. The virtual EGR rate on the high pressure side is model calculated. Such an approach also makes it possible in particular to omit an air mass meter.

A possible EGR controller structure as well as respective adaptation in the context of the AGR or NOx model is described below explained in more detail. However, this is just one of different ways of how to implement a scheme can be.

2 shows a schematic view of an adaptation of an EGR mass flow via a lambda probe. This adaptation is provided as follows: In an air effort model, here called "engine-in-mass-model" go out 1 known pressures and temperatures P2, T2. In the EGR model, here called "EGR model", the values P2, P3, TEGR and SEGR are included. In the EGR model for the low-pressure exhaust gas recirculation comes the pressure difference DP and the position SEGR _LP . LP stands for "low pressure". Below the EGR model for the low-pressure exhaust gas recirculation, a lambda probe or the values determined via the sensors present there are schematically indicated. The structure of the regulation provides that the mass flows determined from the respective models are linked and continued with each other. In order to be able to detect in particular the transient range within the scope of the regulation, a learning function, referred to here as "adaptive learner", is integrated. By means of the learning function stored there, an adaptation of the virtually determined mass flows can take place. Here, as above with respect to 1 beautifully engineered a traced in the low-pressure exhaust gas recirculation EGR mass flow virtually determined and adapted via the learning function. This value is included in the model of high-pressure exhaust gas recirculation, which in turn results in connection with the air effort model, the virtually determined mass flows for the air and the exhaust gas recirculation rate, so that there is a virtual lambda, a virtual oxygen content and an exhaust gas recirculation rate. These are the results that can now be transferred from the virtual exhaust gas recirculation models and from the air mass sensors into the NOx model.

3 shows an adaptation of the NOx model over the values determined by means of the NOx sensor. From 2 determined virtual values air _expenditure λ _-Virtual , virtual EGR rate X _EGRVirtuell and the virtual oxygen _amount fraction Ψ _O2virtual is for example used to determine a virtual oxidation air ratio λ _{Ox, virtual} . This enters a particle model. From this, a particle concentration C _PM in the exhaust gas can thus be determined. From the molar fraction Ψ _{O2, virtual of} the oxygen, a corrected mole fraction of oxygen Ψ _{O2, virtual corrected, is added to} a NOx model, taking into account an oxygen amount fraction difference that has been adapted. From this, a virtual mole fraction of NOx can then be determined. The formula for the determination of the oxygen amount fraction, which is virtually corrected, results here from the 3 resulting relationship. From the virtual oxygen quantity fraction and the characteristic diagram determined via a rotational speed N _engine and a load q, a target value of an oxygen substance amount fraction is supplied. The same is done for a mole fraction of NOx as a setpoint from a map, this value is compared with the determined by the NOx sensor mole fraction NOx also. While from the comparison of the amount of oxygen fraction through a correlation, a difference of the NOx mole fraction as a model-based, quickly determined value, the comparison of the NOx mass fraction of the map and the NOx sensor ei NEN second difference value. These are both compared and then provided to a learning function. From this, an adapted NOx value is then made available to an inverse correlation, from which a difference value for the oxygen substance amount fraction in the form of ΔΨ _O2 adapters then results. The correlation, which is preferably used here, results from the abovementioned dissertation, in particular from Equation 2-3 given on page 7. The determined difference value then returns to the comparison with the virtually determined proportion of oxygen substance amount and corrects it. This corrected value goes into the NOx model, whereby the virtual NOx mass fraction Ψ _{NOx, virtual} can now be determined from this NOx model. The aim here is that the NOx value, which is determined by the NOx sensor, indicating an actual state description, and possibly coincides with that value, as the NOx mole fraction Ψ _{NOx, virtual} finally by the NOx model in this way could be determined. Due to the virtually faster available values and the use of the learning function and thus the adaptation, a faster and, in particular, more precise setting of a mass flow at the exhaust gas recirculation can be carried out in order to be able to comply with the desired nitrogen values or particle values.

4 shows another example, in particular based on the 2 resulting system, to determine mass flows using a mass-loading model of a cylinder. In addition to the use of each model takes place at the 4 Implementation of an adaptation of the mass load from the model "engine-in-mass model" on a dargestell te adaptation. The value determined in this way is finally used to obtain a virtual air value. At the same time it is linked with a virtual EGR mass flow and a load, so that the subsequent module can determine the virtual values air _expenditure λ- _virtual , the virtual EGR rate X _EGR virtual and the virtual oxygen _amount fraction Ψ _O2virual .

Out The individual figures each have different parameters, input as well as outputs and links out. These are not described in detail verbatim but can be seen as shown in the figures. The figures and their contents are not limiting, they are Example interpreted. Therefore, parts of the links, Parameters, inlet and outlet sizes also changed, be omitted or supplemented by others. Also can from individual parts or sections as well as parameters, links, One and outlet sizes new schemas put together be used to perform the procedure and control can be put on.

In the following some remarks are made to the used model based predictive controller (MPR), which is used in the 5 is shown schematically. The 5 illustrates the aforementioned diesel passenger car internal combustion engine, as such, a controlled system 13 describes. Furthermore, the illustrated 5 Manipulated variables u, which are in the controlled system 13 come in, and controlled variables y, which the Regestrecke 13 leave as output variables. The manipulated variables u are at least the previously described setting of an exhaust gas recirculation valve, from which a corresponding mass flow of the exhaust gas recirculation results. By contrast, the controlled variables y are at least the previously mentioned virtually determined NOx value, which can be set via an adjustment of the mass flow of the exhaust gas recirculation.

The MPR uses a dynamic model that is previously described with reference to the 1 to 4 includes described models and structures. This dynamic model is in a so-called observer 14 in which the controlled variables y enter and with which the model is tracked to the real engine process. The model describes a so-called state space model, which is used for the prediction of the controlled variables. A vector describing the state space, which emerges from the state space model, thereby goes into a cost function 15 a, which uses this vector to predict the controlled variables. The of the cost function 15 Predicted control variables are composed of so-called free controlled variables - which describe the predicted portion of the controlled variables that can not be influenced by the manipulated variables - and of the uncontrolled portions associated with the free controlled variables resulting from future manipulated variable changes. The prediction of the controlled variables takes place within a so-called prediction window, which is defined by an upper and by a lower prediction horizon. The regulatory goals, z. As regards the behavior of the controlled variable or controlled variables y, are formulated in the cost function. By optimizing the cost function 15 the future manipulated variables are determined so that the cost function 15 takes a minimum. To account for errors in the model and on the process disturbances, this is the observer 14 deposited and used model of the controlled system 13 tracking the real process. For the time window or prediction window is the cost function 15 minimized. The cost function 15 contains the sum of the error squares of the error between a desired value or a plurality of desired values w and the control variable or the predicted control variables within the prediction window. The result of the cost function 15 is a vector describing the manipulated variables u of the future manipulated variable changes, both in the observer 14 as well as in the controlled system 13 received.

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• - WO 2008/131789 A1 [0003, 0010]
• - DE 102004057929 A1 [0019]
• - DE 10242234 [0029]

Claims (21)

A method for adjusting a mass flow of an exhaust gas recirculation of an internal combustion engine taking into account a NOx behavior, wherein a control provides a coupling of a virtual NOx determination with a real NOx control, characterized in that the control is a model-based predictive controller for controlling a virtually determined NOx Uses values. Method according to claim 1, characterized in that that a control variable of the NOx control means a virtual NOx sensor is determined. Method according to claim 1 or 2, characterized that is a virtual NOx-dependent controlled variable compared with a determined from a map NOx target value becomes. Method according to claim 1, 2 or 3, characterized that as a manipulated variable for a virtual determined NOx control variable an EGR mass flow is being used. Method according to one of the preceding claims, characterized in that the control is based on an exhaust gas recirculation model recourse. Method according to one of the preceding claims, characterized in that an exhaust gas recirculation model the control in a quasi-stationary operating state the internal combustion engine by means of a signal of a real NOx sensor is adjusted. Method according to one of the preceding claims, characterized in that a virtually determined NOx control variable adapted within the framework of an adapted regulation. Method according to claim 7, characterized in that that the adapted control uses a real NOx sensor. Method according to one of the preceding claims, characterized in that the control an inner and an outer Cascade operates, with the inner cascade relies on a lambda probe, while the outer cascade is on a real NOx sensor uses. Method according to claim 9, characterized in that that the inner cascade carries out a comparison of an air determination. Method according to claim 9 or 10, characterized that the outer cascade is an adjustment of a NOx determination performs. Method according to one of the preceding claims 7 to 11, characterized in that checked is whether a prerequisite for applying an adaptation the scheme is still in existence and, in the event of an omission, the Condition the adaptation of the control is suspended. Method according to one of the preceding claims, characterized in that the virtual NOx determination faster takes place as the determination of a NOx value via a real NOx sensor. Method according to one of the preceding claims, characterized in that using a virtually determined Oxygen content determined a parameterizing a particle flow parameter becomes. Application of a method according to one of the preceding Claims to a diesel passenger car engine. Diesel passenger car engine with at least one exhaust gas recirculation, a boost pressure charging, a diesel particulate filter, a catalyst, a NOx sensor and a first regulation regarding a return rate an exhaust stream for adjusting a NOx value, wherein the first Control first control means comprising a virtual NOx sensor simulate, second control means having an adapted control of the virtual NOx sensor, and third control means having a NOx control implement, wherein the first control is constructed such that the virtual NOx sensor is a default for the first scheme, the first scheme as a regulation according to one of claims 1 to 14 executed is. Diesel passenger car engine according to claim 16, characterized in that the first control as a parent outer rules cascade which is a second, inner control cascade with a faster control time than that of the outer rule cascade. Diesel passenger car engine according to claim 16, characterized in that the first control as a parent outer rules cascade which is a second, inner control cascade with a slower control time than that of the outer rule cascade Diesel passenger vehicle internal combustion engine according to one of the preceding claims, characterized in that determining means for determining an oxygen concentration at an engine inlet of the Diesel-Verbren combustion engine and are provided for determining an oxygen content of a recirculated exhaust gas. Diesel passenger car engine according to one of the preceding claims, characterized that an adaptation to adapt one or more models of the scheme intended to match virtually determined values of the models is, with the adaptation for this purpose with a signal flow of the lambda probe as well as from a real NOx sensor is connected. Diesel passenger car engine according to one of the preceding claims, characterized that a low-pressure exhaust gas recirculation and / or a high-pressure exhaust gas recirculation are provided, each deposited as a model in the scheme.