DE102017204868A1 - Method and device for adapting a characteristic diagram of an internal combustion engine - Google Patents

Abstract

The invention relates to a method and a device for adapting an initial map which can be addressed by means of one or more input signals and has a plurality of initial map values, stored in a memory of an individual internal combustion engine, having the following steps:
- Determining occurring during operation of the individual internal combustion engine deviations from map values of each associated initial map value and
Creation of a correction map in which the deviations detected during operation from the respective associated initial map value are stored,
- Wherein the creation of the correction map, a division of the detected deviations into a global trend of the deviations descriptive adaptive regression plane and an adaptive correction map is made.

Description

The invention relates to a method and a device for adapting a characteristic diagram of an internal combustion engine.

Modern internal combustion engines are controlled by engine control units, in which the control signals are determined on the basis of software implemented models of the internal combustion engine. These models are usually subdivided into submodels which each describe a specific subsystem of the internal combustion engine and whose interfaces represent certain physical operating variables of the internal combustion engine, for example the engine speed, the intake manifold pressure and the exhaust gas temperature. Such a model of a subsystem can be implemented either as an analytical model by calculating an output of the subsystem from input variables by means of an equation or as a data-based model in which the output variable of the subsystem is read from a one-dimensional or multi-dimensional characteristic map spanned by the input variables.

With the implementation of a model in the engine control software as a map, the qualitative dependency of the output signal on the selected input signals is established. For the quantitative dependency of the output signal of the map on its input signals, a fixed frame is staked in the implementation by the choice of the map size and the resolution and the value range of the map axis and the map values. In general, however, the choice of map values can represent a broad range of complex functional dependencies. This is done as part of the calibration of the engine control unit for use with a particular engine / vehicle type based on simulations and measurements on a reference engine or on a representative group of engines and vehicles. Care must be taken in many maps that the map values in adjacent map points do not have too great a difference - i. E. that the characteristic field is sufficiently smooth - otherwise the output signal fluctuates greatly when driving through the dynamic map, which can have a negative effect on the operating behavior of the engine.

The behavior of the individual engines and vehicles deviates from each other as a result of manufacturing tolerances, aging of the components and the interaction with different, not described in the engine control or not sufficiently detailed operating conditions. For motor operation significant deviations of the operating variables of the individual engine from the behavior of the reference motor are detected with the help of sensors and minimized by control interventions. A control intervention can always take place only in response to an already occurred deviation, so this deviation can only reduce, but not prevent. In order to avoid the deviations occurring when the operating point is approached for the first time when repeatedly approaching an operating point, selected parameters of the corresponding subsystem model can be permanently adapted so that this deviation is minimized. These changed parameters are stored non-volatilely in the control unit, so that they act immediately upon restarting the same operating point in the same or later driving cycles and the observed deviation of the operating behavior does not occur again.

• - die Werte des kalibrierten Kennfelds selbst nichtflüchtig verändert werden, oder
• - ein zweites Korrekturkennfeld über dieselben Eingangsgrößen aufgespannt wird, dessen Ausgang auf den Ausgang des kalibrierten Kennfelds addiert wird und dessen Werte ausgehend vom neutralen Initialwert 0 nichtflüchtig so verändert werden, dass die Abweichung des Betriebsverhaltens minimiert wird, oder
• - ein zweites Korrekturkennfeld über dieselben Eingangsgrößen aufgespannt wird, dessen Ausgang auf den Ausgang des kalibrierten Kennfelds multipliziert wird und dessen Werte ausgehend vom neutralen Initialwert 1 nichtflüchtig so verändert werden, dass die Abweichung des Betriebsverhaltens minimiert wird.

In the case of a model implemented using a map, this can be done by either

• - the values of the calibrated map itself are non-volatilely changed, or
• - A second correction map is spanned over the same input variables, the output of which is added to the output of the calibrated map and whose values are changed non-volatile starting from the neutral initial value 0, that the deviation of the operating behavior is minimized, or
• - A second correction map is spanned over the same input variables whose output is multiplied to the output of the calibrated map and its values are non-volatile changed from the neutral initial value 1, that the deviation of the operating behavior is minimized.

In Applicant's engine control units, a desired wastegate position as a function of desired waste gas mass flow and desired pressure ratio is calibrated via the wastegate for controlling a turbocharger in the calibrated characteristic map and an additive correction in the correction map with the same input variables desired waste gas mass flow and desired pressure ratio via the wastegate learned the desired wastegate position. The sum of calibrated desired wastegate position and correction of the desired wastegate position is the corrected desired wastegate position.

In order to ensure that the corrected desired wastegate position is also sufficiently smooth, ie that the corrected desired wastegate positions in adjacent characteristic map points do not have too great differences so that the output signal does not fluctuate greatly during dynamic driving through the characteristic map, it is allowed to adapt each individual map point of the correction map a predetermined maximum allowable difference to all neighboring points in the map are not exceeded, ie the correction map must remain sufficiently smooth. If an adaptation step which was actually useful for reducing the deviation of the engine behavior from the reference engine would lead to exceeding the maximum permissible difference between neighboring points, this adaptation step is not carried out.

The value ranges of the axes of a calibrated characteristic map are usually selected so that all relevant operating points are covered. The sum of all operating points of a system in the two-dimensional space spanned by the input variables of the calibrated map generally does not form a perfect rectangle, so that there are areas in most maps, in particular near the fairleads, which are never approached. In the same way, even in the case of a characteristic diagram with more than two input variables, the sum of all operating points in the higher-dimensional space does not form a right-angled parallelepiped. Even in a characteristic curve usually designated as a characteristic curve with only one input variable, the two ends of the characteristic curve may possibly never be approached. If an adaptive correction map is set up using the same input variables, it follows that operating points exist in this adaptive correction map as well, which are never approached at least stationary, thus never adapted and thus always remain at their initial value. Thus, in the adaptive correction map, there is usually an unskilled area which remains at initial value and a learned area which differs from the initial value.

It follows from the required smoothness of the correction map that the map points are located directly at the edge of the learned map area, i. in the immediate vicinity of a map value standing at initial value, only values from the interval [-1... + 1] * (maximum permissible difference to all neighboring points) can assume. Your neighbor points, which have no direct neighbor points, can only accept values from the interval [-2 ... + 2] * (maximum permissible difference to all neighboring points), etc.

If an individual combustion engine has a greater deviation from the reference engine at an operating point at the edge of the learned range of the adaptive correction map of the target wastegate position than with a correction from the interval [-1... + 1] * (maximum permissible difference to all neighboring points ) can be compensated, then the map adaptation is not able to fully compensate for this deviation with the given smoothness requirements. There is a conflict between the smoothness requirement and the need to learn large correction values at the edge of the operating room.

Engine control units of the same type may be installed in various vehicles and vehicle variants, for example, with different variants of the exhaust system to meet different requirements for pollutant emissions and noise. Traditionally, engine ECUs have been calibrated for the specific application of an engine in a particular vehicle variant, i. The parameters of the motor control are optimized for each application. In this procedure, deviations between individual motors and the reference motor result from component dispersion and component aging.

For the adaptation of the turbocharger drive, this means that the above-described method of adapting the desired wastegate position was sufficient.

In order to save costs, in recent years, not all subsystems of the engine control units for all applications in different vehicles and in different vehicle variants are optimized individually, but for individual subsystems the data determined for a vehicle variant are also used for other vehicle variants and vehicle types. In this procedure, the deviations between individual engines and the reference engine increase the effects of different operating conditions of the engine in different vehicle variants and vehicle types.

For the adaptation of the turbocharger drive, this has the consequence that the need for a correction of the desired wastegate position generally increases in the entire operating range of the described characteristic map for determining the desired wastegate position and especially also at the edge of the learned region of the correction map. The described limitations of adaptation due to the smoothness requirements are no longer acceptable.

The object of the invention is to show a way how these disadvantages can be avoided.

This object is achieved by a method having the features specified in claim 1. Advantageous embodiments and further developments of the invention are specified in the dependent claims 2-10. The claim 11 has a device for adapting an addressable by means of one or more input signals, a plurality of initial map values having stored in a memory of an individual internal combustion engine initial map subject. The claim 12 relates to a use of a method according to any one of claims 1-10 in the control of an internal combustion engine.

• - Ermittlung von im Betrieb der individuellen Verbrennungskraftmaschine auftretenden Abweichungen von Kennfeldwerten von einem jeweils zugehörigen Initial-Kennfeldwert und
• - Erstellung eines Korrekturkennfeldes, in welchem die im Betrieb erfassten Abweichungen vom jeweils zugehörigen Initial-Kennfeldwert abgespeichert sind,
• - wobei bei der Erstellung des Korrekturkennfelds eine Aufteilung der erfassten Abweichungen in eine den globalen Trend der Abweichungen beschreibende adaptive Regressionsebene und ein adaptives Korrekturkennfeld vorgenommen wird.

In the method specified in claim 1 for adapting an initial characteristic map which can be addressed by means of one or more input signals and has a plurality of initial characteristic values, stored in a memory of an individual internal combustion engine, the following steps are carried out:

• - Determining occurring during operation of the individual internal combustion engine deviations from map values of each associated initial map value and
• Creation of a correction map in which the deviations detected during operation from the respective associated initial map value are stored,
• - Wherein the creation of the correction map, a division of the detected deviations into a global trend of the deviations descriptive adaptive regression plane and an adaptive correction map is made.

By these measures is achieved in an advantageous manner that the existing in known methods conflict between the smoothness requirements and the need for large corrections is also solved at the edge of the approachable range of the respective map. Both large correction values can be learned at the edge of the operating room and large fluctuations in the output variable of the respective characteristic field during dynamic driving through the operating room can be avoided.

A method with the inventive features can, for example, in connection with a determination of the desired position of the wastegate of a turbocharger from the desired Wastegatemassenstrom and the desired pressure ratio across the wastegate by means of a map with the inputs target Wastegatemassenstrom and target pressure ratio through the wastegate and the Output target position of the wastegate can be used.

In an analogous manner, a method with the features according to the invention can also be used for a different output variable for controlling the turbocharger, for example for a set pressure in a wastegate actuator or for a setpoint current through an actuator.

Furthermore, a method with the features according to the invention can also be used in conjunction with other input variables such as the desired boost pressure or the desired supercharger speed, for only one or more than two input variables or for driving not a turbocharger, but another component of the internal combustion engine ,

• 1 eine Blockdarstellung einer Vorrichtung zur Steuerung eines von einem Abgasturbolader aufgeladenen Verbrennungsmotors,
• 2 eine Skizze zur Veranschaulichung einer bekannten Vorrichtung zur Ermittlung einer Soll-Wastegateposition,
• 3 eine Skizze eines bekannten Vorsteuerkennfelds einer Soll-Wastegateposition (0 = voll geschlossen, 1 = voll geöffnet) als Funktion eines skalierten Soll-Massenstroms und eines skalierten Soll-Druckverhältnisses,
• 4 eine Skizze eines bekannten Adaptionskennfelds zur additiven Korrektur einer Soll-Wastegateposition als Funktion eines skalierten Soll-Massenstroms und eines skalierten Soll-Druckverhältnisses,
• 5 erste Skizzen zur Erläuterung eines erstmaligen Adaptionsparameter-Balancings im Sinne des erfindungsgemäßen Verfahrens,
• 6 eine Skizze zur Veranschaulichung einer erfindungsgemäßen Vorrichtung zur Ermittlung einer Soll-Wastegateposition und
• 7 weitere Skizzen zur Erläuterung eines weiteren Adaptionsparameter-Balancings im Sinne des erfindungsgemäßen Verfahrens.

Further advantageous features of the invention will become apparent from the following exemplary explanation with reference to FIGS. It shows

• 1 a block diagram of an apparatus for controlling a supercharged from an exhaust gas turbocharged internal combustion engine,
• 2 1 is a sketch to illustrate a known device for determining a desired wastegate position;
• 3 a sketch of a known pilot control map of a desired wastegate position (0 = fully closed, 1 = fully open) as a function of a scaled desired mass flow and a scaled desired pressure ratio,
• 4 2 is a sketch of a known adaptation map for the additive correction of a desired wastegate position as a function of a scaled desired mass flow and a scaled desired pressure ratio,
• 5 first sketches for explaining a first adaptation parameter balancing in the sense of the method according to the invention,
• 6 a sketch for illustrating a device according to the invention for determining a desired wastegate position and
• 7 further sketches for explaining a further adaptation parameter balancing in the sense of the method according to the invention.

The 1 shows a block diagram of an apparatus for controlling a supercharged from an exhaust gas turbocharged internal combustion engine.

This device contains a drive train 100 , one of a turbocharger 20 charged combustion engine 10 contains. The internal combustion engine 10 has one or more cylinders 11, each having an inlet valve 12 assigned.

To the exhaust gas turbocharger 20 belong to a turbine 21 and a compressor 22 , The compressor 22 is input side to a fresh air duct 1 connected and provides the output side compressed fresh air. This is via a charge air cooler 40 supplied to a charge air passage 2 and from there via a throttle valve 30 to a suction pipe 3 passed. From the intake manifold, the compressed air through the respective inlet valve 12 in the respective cylinder 11 passed. There, the introduced into the cylinder fuel is burned with this fresh air.

The exhaust gases formed in this combustion process pass through an exhaust pipe 4 on the turbine 21 the exhaust gas turbocharger 20 , There they drive a turbine wheel, which is connected via a shaft with a compressor wheel arranged in the compressor, which is thus driven by the turbine wheel via said shaft.

Furthermore, the device shown has a wastegate channel 51 on, which by means of a wastegate flap 50 can be adjusted continuously between a fully open state and a closed state.

Furthermore, in the 1 1, a motor control device 60, which is designed to control the internal combustion engine 10 or a plurality of actuators of the internal combustion engine 10. This motor control device 60 is supplied with a plurality of sensor signals se1, ..., sen. The engine control device 60 determined using these sensor signals, a stored work program and more, in a memory 62 stored data control signals st1, ..., sty, which are provided for controlling said actuators of the internal combustion engine. To those in the store 62 stored data include, among other maps, each describing a particular subsystem of the internal combustion engine and whose interfaces relate to certain physical operating variables of the internal combustion engine, such as the engine speed, the intake manifold pressure and the exhaust gas temperature. Such a model of a subsystem can be implemented as a data-based model, in which the output variable of the subsystem is read from a one-dimensional or multi-dimensional characteristic map spanned by the input variables.

• - Ermittlung von im Betrieb der individuellen Verbrennungskraftmaschine auftretenden Abweichungen von Kennfeldwerten von einem jeweils zugehörigen Initial-Kennfeldwert und
• - Erstellung eines Korrekturkennfeldes, in welchem die im Betrieb erfassten Abweichungen vom jeweils zugehörigen Initial-Kennfeldwert abgespeichert sind,
• - wobei bei der Erstellung des Korrekturkennfelds eine Aufteilung der erfassten Abweichungen in eine den globalen Trend der Abweichungen beschreibende adaptive Regressionsebene und ein adaptives Korrekturkennfeld vorgenommen wird.

The engine control device 60 is for carrying out a method for adapting an addressable by a plurality of input signals, having a plurality of initial map values, in the memory 62 formed of an individual internal combustion engine stored three-dimensional initial map, in which the following steps are performed:

• - Determining occurring during operation of the individual internal combustion engine deviations from map values of each associated initial map value and
• Creation of a correction map in which the deviations detected during operation from the respective associated initial map value are stored,
• - Wherein the creation of the correction map, a division of the detected deviations into a global trend of the deviations descriptive adaptive regression plane and an adaptive correction map is made.

In this method, the existing in known methods conflict between smoothness requirements to the output of the map and the need for large corrections also at the edge of the approachable area of the respective map is achieved in that the learned correction values in the correction map in a global trend reflecting regression plane and Residual correction map are split.

This is explained below using an exemplary embodiment in which a determination of the desired position of the wastegate of a turbocharger from the desired waste gas mass flow and the desired pressure ratio via the wastegate by means of a characteristic map with the inputs Wastegatemassenstrom and desired pressure ratio via the wastegate and the output Target position of the wastegate is made.

In this embodiment, the adaptive correction map with its output correction of the desired wastegate position (= z) and its inputs Wastegatemassenstrom (= x) and desired pressure ratio across the wastegate (= y) implements a mathematical function z = f (x, y ). Analogously, a function z = f (x) can also have only one input variable or a function z = f (x, y, w,...) Can have more than two input variables.

The map has m nodes x_1 ... x_m in x-direction and n nodes y_1 ... y_n in y-direction. Each map value z_k (x_i, y_j) with i∈ [1,..., M] and j∈ [1,..., N] is either equal to the initial value 0 (hereinafter referred to as z_ki) ) or has moved away from it (hereinafter referred to as z_kg).

For any state of the adaptive correction map with any number of 0 different map points z_kg, taking into account the maximum allowable difference to the values of neighboring points of arbitrary values, there is exactly one regression plane through that cloud of map location points z_kg other than 0. The regression level is a level for which a rating criterion assumes a minimum value. As such an evaluation criterion, for example, the sum of the squares of the distances of the map values in the z-direction between the plane and all non-zero map points can be used. Another possible evaluation criterion is, for example, the sum of the amounts of the distances of all of the 0 different map points from this level. This plane can be uniquely described with three parameters a, b, c, so that the following relationship applies: $$\text{z\_e}=\text{a}*\text{x}+\text{b}*\text{y}+\text{c},$$

In the case of an adaptive map with only one input variable, instead of the regression plane, a regression line with the corresponding parameters a and c occurs with the equation z_e = a * x + c. In the case of an adaptive map with more than two input quantities, the regression plane is a plane in multidimensional space with corresponding parameters a, b, c, d and the equation z_e = a * x + b * y + d * w + c, etc. All Designs for the two-dimensional map also apply correspondingly to the one-dimensional and multi-dimensional maps.

If a set of regression plane parameters a_old, b_old, c_old has already been determined at least once, one can also determine for each arbitrary state of the adaptive correction map exactly one new regression plane through the cloud of the sum of old regression plane and correction map in the map points different from zero , Assuming that the parameters of the old regression plane a_old, b_old, c_alt are initialized to 0, this is the general case, which also describes the first determination of a regression plane.

• - an allen Positionen den Wert z_r(x_i,y_j)=0 hat, an denen das Korrekturkennfeld den Wert 0 hat, und
• - an allen anderen Positionen x_i, y_j gleich z_r(x_i,y_j)=z_kg + (a_alt * x_i + b_alt * y_j + c_alt) - (a * x_i + b * y_j + c) ist.

For each state of the adaptive correction map, each parameter set a_old, b_old, c_old of an old regression plane and each parameter set a, b, c of a new regression plane, one can unambiguously determine a residual map z_r of dimension mxn which

• - has the value z_r (x_i, y_j) = 0 at all positions at which the correction map has the value 0 has, and
• - at all other positions x_i, y_j is equal to z_r (x_i, y_j) = z_kg + (a_old * x_i + b_old * y_j + c_old) - (a * x_i + b * y_j + c).

The sum of this residual map z_r and the new regression plane is equal to the sum of the correction map and the old regression plane z_r (x_i, y_j) + (a * x_i + b * y_j + c) at all points where the correction map is different from 0. = z_kg + (a_old * x_i + b_old * y_j + c_old).

This conversion of an old regression plane and an old correction map into a new regression plane and a residual map is referred to as adaptation parameter balancing. After each adaption parameter balancing, the mean value of the values in the remaining map is smaller than in the old correction map. In order to specifically minimize the amounts of the values at the edge of the learned range of the remainder map, the regression level can not be optimized to the entire map area other than the initial value, but to the edge of the map area different from the initial value, ie Evaluation criterion minimizes the sum of the squares of the distances of the plane of all non-zero map points, which at least one adjacent map point with the value 0 to have.

The following procedure is adopted for adapting the desired wastegate position during operation of the individual engine:

During the production of the control unit, both the adaptive correction map and the plane parameters a, b, c are initialized with the initial value 0.

In the first driving cycles of the individual engine, map points of the adaptive correction map are learned until a predetermined number of initial points of various points, for example 20 points, are reached, i. until there is a sufficient number of initial values of different map points for a meaningful calculation of a regression plane.

At a greater time interval, for example, always at the beginning of the follow-up phase of the engine control unit after stopping the engine, for the sum of the old repression level and the correction map, the parameters a, b, c of the regression plane by the edge of the initial value 0 different range of the correction map z_k and remainder map z_r recalculated. The parameters of the regression plane are stored non-volatile and the adaptive map z_k is overwritten with the rest map z_r and also stored non-volatile.

In engine operation, the sum of regression plane and correction map z_k is used as correction of the desired wastegate position in all operating points.

As the engine continues to operate, the map points of the adaptive correction map are further learned, taking into account the maximum permissible difference with neighboring points.

• - In allen gelernten Korrekturkennfeldpunkten ist die Summe aus adaptiver Ebene und adaptivem Kennfeld vor und nach jeder Ebenenadaption gleich.
• - Die Summe aus Regressionsebene und Restkennfeld extrapoliert die gelernten Werte in den ungelernten Bereich. Und liefert damit auch im nicht unmittelbar adaptierten Kennfeldbereich eine Korrektur der Wastegatesollposition.
• - Die Beträge der Werte am Rand des gelernten Bereichs des adaptiven Korrekturkennfelds sind kleiner als vor der Bestimmung einer neuen Regressionsebene. Damit sind die Differenzen im adaptiven Korrekturkennfeld zu den nicht gelernten, auf Initialwert stehenden Kennfeldpunkten kleiner. Damit kann bei unveränderten Glattheitsanforderungen an das Korrekturkennfeld am Rand des gelernten Bereichs des adaptiven Kennfelds weiter gelernt werden.

The effect of an adaptation parameter balancing, ie the determination of a new regression plane and a residual map from an old regression plane and a learned correction field, is the following:

• In all learned correction map points, the sum of adaptive plane and adaptive map before and after each plane adaptation is the same.
• - The sum of regression level and residual map extrapolates the learned values to the unskilled area. And thus also provides a correction of the wastegate target position in the map area that is not directly adapted.
• The amounts of the values at the edge of the learned range of the adaptive correction map are smaller than before the determination of a new regression plane. Thus, the differences in the adaptive correction map to the non-learned, on initial value map points are smaller. Thus, with unchanged smoothness requirements to the correction map at the edge of the learned range of the adaptive map continue to be learned.

The described method solves the conflict between the smoothness requirement on the correction field and the need to learn large correction values at the edge of the operating space. Both large correction values can be learned at the edge of the operating room and large fluctuations in the desired wastegate position during dynamic driving through the operating room can be avoided.

The present invention, after all, deals with data-based models of subsystems implemented as a map. The method has been described above using the example of determining the desired position of the wastegate of a turbocharger from the target Wastegatemassenstrom and target pressure ratio across the wastegate using a map with the inputs target waste gas mass flow and desired pressure ratio through the wastegate and the output target position of the wastegate.

• - für einen Soll-Druck in einem Wastegateaktuator oder eine Soll-Stromstärke durch einen Aktuator, oder
• - für andere Eingangsgrößen wie beispielsweise den Soll-Ladedruck oder die Soll-Laderdrehzahl, oder
• - für nur eine oder mehr als zwei Eingangsgrößen mit einer entsprechenden Anpassung, oder
• - zur Ansteuerung nicht eines Turboladers, sondern einer anderen Komponente des Verbrennungsmotors.

However, the method can also be applied to another output for controlling a turbocharger, for example

• - for a desired pressure in a Wastegate Actuator or a desired current through an actuator, or
• - For other input variables such as the target boost pressure or the target supercharger speed, or
• - for only one or more than two input variables with a corresponding adaptation, or
• - To control not a turbocharger, but another component of the engine.

Based on 2 to 7 The procedure in a previously known method and the procedure of the method according to the invention are explained in more detail below.

The 2 shows a sketch to illustrate a known device for determining a desired wastegate position in a turbocharger with position-controlled wastegate.

In this device, the desired position of the wastegate to be determined results from the sum of the output signal of a pilot control 110 with the output signal of an adaptation of pilot control 120 and the output signal of a charge pressure deviation minimizing a current boost pressure deviation 130 as follows: $$\begin{array}{c}\text{POSN\_BPA\_SP\_RGL}=\text{FAC\_POW\_PUT\_CTL\_OPL}+\text{FAC\_PUT\_AD\_ADD\_WG}+\\ \text{FAC\_POW\_PUT\_CTL\_CLL}\text{,}\end{array}$$

• POSN_BPA_SP_RGL die zu ermittelnde Wastegate-Sollposition,
• FAC_POW_PUT_CTL_OPL das Ausgangssignal der Vorsteuerung,
• FAC_PUT_AD_ADD_WG das Ausgangssignal der Adaption und
• FAC_POW_PUT_CTL_CLL das Ausgangssignal des Ladedruckreglers.

Where:

• POSN_BPA_SP_RGL the wastegate target position to be determined,
• FAC_POW_PUT_CTL_OPL the output signal of the pilot control,
• FAC_PUT_AD_ADD_WG the output signal of the adaptation and
• FAC_POW_PUT_CTL_CLL the output signal of the boost pressure regulator.

The output signal of the pilot control 110 and the output of the adaptation 120 depend, among other things, on the setpoint boost pressure, the exhaust gas temperature and the ambient pressure.

The output signal of the boost pressure regulator 130 depends on the target boost pressure and the actual boost pressure.

An adaptation of the precontrol is necessary because an individual motor for setting a desired operating point in the vast majority of cases requires a different wastegate position than the reference motor, by means of which the parameters of the precontrol have been determined. This is due to the fact that in the production of the turbocharger or engines series variations occur that during the service life of the turbocharger or engines aging effects deteriorate the response of the engine and that often use a common data set for different vehicles (exhaust systems, bodies, .. .) he follows. The adaptation of the pilot control 120 learns these differences in individual map points and thus creates an adaptive map or adaptation map.

The 3 shows a sketch of a known pilot control map. It can be seen that this pilot control map, which is determined by means of a reference engine, has a "waterfall shape".

The 4 shows a sketch of a known adaptation map. It can be seen that this adaptation map has a "mountain chain shape".

A problem with the known approach is that when there is a great need for adaptation, i. a large deviation of the fed feedforward control to the control requirement of the individual engine, large adaptation values can be learned in individual characteristic map points, while in direct neighboring points of the characteristic map one has not yet adapted.

In order to avoid oscillations of the control occurring from strongly different adaptation values in adjacent characteristic diagram points when driving through the adaptation characteristic field, when writing an adaptation value, a maximum permissible difference C_FAC_PUT_AD_GRD_MAX is specified for all 8 neighboring points of the characteristic field.

Since the shape of the actually approached map area is never rectangular, in practice there is always an unadapted map area in which the map values are equal to the initial value zero. If large adaptation values are required at the edge of the approached map area, in the case of known methods, the range of values achievable by the adaptation is limited to +/- C_FAC_PUT_AD_GRD_MAX by the neighborhood with the map areas not approached.

This restriction is to be canceled by the present invention, so that even at the edge of the approachable map area large adaptation values can be learned.

The 5 shows first sketches for explaining a first adaptation parameter balancing in the sense of the method according to the invention. In this method, a learned adaptive map AK0 is first split into an adaptive regression plane AE1 and a residual map AK1 in a control device tracking. Before this first adaptation parameter balancing, the adaptation value AW0 is equal to the learned adaptive characteristic diagram AK0 in the entire map area.

For the adaptive regression level, the following relationship applies: $$\text{AE1}=\text{a}*\text{x}+\text{b}*\text{y}+\text{c}\text{,}$$

The adaptation value AW1 as the sum of the adaptive plane AE1 and the adaptive correction map AK1 is equal to the value of the old adaptive correction map AK0 in all learned map points.

The 6 shows a sketch for illustrating a device according to the invention for determining a desired wastegate position.

In this device, the desired position of the wastegate to be determined results from the sum of the output signal of a pilot control 110 with the output signal of an adaptation of the pilot control 120 'and the output signal of a boost pressure regulator 130 as follows: $$\begin{array}{c}\text{POSN\_BPA\_SP\_RGL}=\text{FAC\_POW\_PUT\_CTL\_OPL}+\text{FAC\_PUT\_AD\_ADD\_WG}+\\ \text{FAC\_POW\_PUT\_CTL\_CLL}\text{,}\end{array}$$

• POSN_BPA_SP_RGL die zu ermittelnde Wastegate-Sollposition,
• FAC_POW_PUT_CTL_OPL das Ausgangssignal der Vorsteuerung,
• FAC_PUT_AD_ADD_WG das Ausgangssignal der Adaption und
• FAC_POW_PUT_CTL_CLL das Ausgangssignal des Ladedruckreglers.

Where:

• POSN_BPA_SP_RGL the wastegate target position to be determined,
• FAC_POW_PUT_CTL_OPL the output signal of the pilot control,
• FAC_PUT_AD_ADD_WG the output signal of the adaptation and
• FAC_POW_PUT_CTL_CLL the output signal of the boost pressure regulator.

The output signal of the pilot control 110 and the output of the adaptation 120 ' depend, among other things, on the setpoint boost pressure, the exhaust gas temperature and the ambient pressure. The in the 6 The parameters FAC_APPROX_0_BPC_AD, FAC_APPROX_1_BPC_AD and FAC_APPROX_2_BPC_AD correspond to the parameters of the regression level c, b and a.

The output signal of the boost pressure regulator 130 depends on the target boost pressure and the actual boost pressure.

In the adaptation of the pilot control 120 ' unlike in the 2 the known device shown above, the division of the adaptive map described above into an adaptive regression plane AE and an adaptive residual map AK made.

The output signal of the adaptation of the precontrol 120 ' results from a summation of the adaptive regression plane AE with the adaptive residual map AK.

The 7 shows further sketches for explaining a further adaptation parameter balancing in the sense of the method according to the invention, in which an updating of the adaptive regression plane taking place in a control unit tracking is illustrated.

Here are in the top line of 7 Sketches illustrating the conditions immediately before a recalculation of the adaptive regression plane and in the bottom line of FIG 7 Sketches illustrating the relationships immediately after a recalculation of the adaptive regression plane are shown.

Before the said recalculation of the regression plane, there are an adaptive regression plane AE1 and an adaptive residual map AK1, which, in their addition as a function of the inputs x and y, result in an adaptation value AW1.

After the above recalculation of the regression plane, there is an adaptive regression plane AE2 and an adaptive residual map AK2, which, in their addition as a function of the inputs x and y, result in an adaptation value AW2. In all map points in which the adaptive map A1 is not equal to 0, AW1 = AW2, d. H. the adaptation values as the sum of the adaptive regression plane and the adaptation map are equal before and after adaptation parameter balancing.

The adaptive regression plane AE2 was recalculated in such a way that the adaptive residual map AK2 was adjusted so that the evaluation criterion is minimal. The adaptive residual map AK2 has been adapted so that the adaptation values AW remain the same as the sum in the learned points.

The 7 shows that with unchanged smoothness requirements to the correction map, the amounts in the adaptive residual map after an adaptation of the regression plane are smaller than before. There are smaller differences to the unskilled area. This advantageously has the consequence that it is possible to continue learning.

Claims (12)

Method for adapting an initial characteristic map which can be addressed by means of one or more input signals and has a multiplicity of initial characteristic values, stored in a memory of an individual internal combustion engine, comprising the following steps: - Determining occurring during operation of the individual internal combustion engine deviations from map values of each associated initial map value and Creation of a correction map in which the deviations recorded during operation are stored by the respectively associated initial map value, - Wherein the creation of the correction map, a division of the detected deviations into a global trend of the deviations descriptive adaptive regression plane and an adaptive correction map is made. Method according to Claim 1 , characterized in that the deviations occurring during operation of the individual internal combustion engine minimize a selected criterion. Method according to Claim 2 , characterized in that the selected criterion is the amount of a boost pressure deviation. Method according to one of the preceding claims, characterized in that a plane is determined as the adaptive regression plane, for which an evaluation criterion is minimal. Method according to Claim 4 characterized in that the evaluation criterion is the sum of the squares of the adaptive regression plane distances of all the adaptive map characteristics map values different from the initial map value. Method according to Claim 4 or 5 , characterized in that the following relationship applies to the determined adaptive regression plane z_e: $$\text{z\_e}=\text{a}*\text{x}+\text{b}*\text{y}+\text{c}\text{.}$$

where z_e is the value of the adaptive regression plane at the operating point (x, y), x is the desired pressure ratio via the wastegate or a variable derived from the desired pressure ratio via the wastegate, y is the desired mass flow via the wastegate or a variable derived from the desired mass flow via the wastegate and a, b and c are parameter values. Method according to one of the preceding claims, characterized in that the maps are multidimensional maps. Method according to Claim 7 , characterized in that the maps are three-dimensional maps containing each support points in the x and y direction and map values in the z direction. Method according to one of the preceding claims, characterized in that in follow-up phases of a motor control device, a determination of a new regression level and a new correction map is made. Method according to one of the preceding claims, characterized in that by determining a new regression plane map values are extrapolated into a region of the map in which no individual adaptation of the map values has been made so far. Device for adapting an initial characteristic map, which is addressable by means of one or more input signals and has a multiplicity of initial characteristic values, stored in a memory of an individual internal combustion engine, characterized in that it comprises a motor control device (60) for controlling a method according to one of the preceding claims is trained. Use of a method according to one of Claims 1 - 10 in the control of an internal combustion engine.

Images