WO1998037321A9 - Process and device for controlling the mixture in an internal combustion engine - Google Patents

Abstract

A process for controlling the mixture in an internal combustion engine has the following steps: (a) at least one value linked to the air volume that reaches a combustion chamber of the internal combustion engine is measured (so-called input value); (b) at least one output value which controls the amount of fuel to be supplied is determined depending at least on the input value(s) measured under (a), by means of stored representation information; (c) the amount of fuel is supplied according to the output value under (b); (d) a value which gives information on the thus obtained mixture is measured (so-called real value); (e) a deviation between the real value measured under (d) and a nominal value is determined; (f) the stored representation information is altered depending on the deviation determined under (e) for the state of operation measured under (a), so that when steps (a) to (e) are carried out in the future in the same state of operation, the deviation is reduced. This is implemented by a learning process. Also disclosed is a device for carrying out this process.

Description

 Method for mixture control in an internal combustion engine and apparatus for its implementation.

The invention relates to a method for mixture control in an internal combustion engine and to a device for carrying it out

The mixture control not only plays an important role in the operating behavior of an internal combustion engine, but is crucial for the achievement of lower emissions of harmful exhaust gases So überfuhrt the usual in gasoline engines

Three-way catalyst only satisfactorily harmful exhaust gases (especially hydrocarbons, carbon monoxide and nitrogen oxides) in harmless reaction products, if the so-called lambda (λ) of the mixture within very narrow limits at λ = l (stoichiometric ratio) Also in diesel engines to achieve low emissions and high Efficiencies Specific variable mixture ratios can be achieved very precisely In view of the increasing problems in the area of air pollution control and the ever stricter emission protection regulations, mixture control therefore plays an important role

In the conventional systems, the mixture is generally adjusted by pilot control and superimposed control. For example, the remaining oxygen content in the exhaust gas is measured via a lambda probe arranged in the exhaust gas flow. The desired stoichiometric value of the air / fuel mixture (ie λ = 1) leads to a certain residual oxygen content. On the other hand, a deviation from λ-1 results in a greater or lesser residual oxygen content. The lambda probe generates a corresponding output signal. This serves as an actual value for a PI controller, which changes the fuel quantity to be injected in such a way that the stoichiometric or -egf maintain a different desired mixing ratio In order to keep the control difference, which is unavoidable in a conventional sense, as small as possible, an operating point-dependent z B of the throttle position and the speed dependent feedforward control of the injection quantity is determined by using static maps. The characteristic maps are determined by the engine manufacturer in a comprehensive test on test states stored in the map to compensate for the series deviation of the decisive for the combustion engine parameters As long as no control signal is present, the mixture control uses the value specified in the map value for the injection due to the several working cycles amounting dead time (namely until a certain injection quantity in the exhaust gas at the location of the lambda probe noticeable), it is not possible for this regulation to maintain the desired mixing ratio after load or speed changes. In practice, this is the reason for acceleration and deceleration phases the lambda control is completely inactivated, the injection quantity is then controlled only with the aid of the characteristic map. In dynamic driving, therefore, there are considerable deviations from the lambda nominal value, which leads to correspondingly high pollutant emissions. Strict exhaust gas regulations can not be met or only with difficulty

In the publication U Lenz and D Schroeder Artificial Intelligence for Combustion Engine Control, SAE Paper No. 960328, February 1996, a method for

Determination of the air mass sucked into a cylinder proposed The

Publication U Lenz and D Schröder identification of isolated nonlinear data with

Neural Networks, GMA Technical Committee 1 4 "Theoretical Methods of

Control Engineering ", Workshop in Interlaken, 1996, generally refers to a so-called observer approach based on a neural network in these

Publications becomes a method of artificial intelligence, applicable to

Regulation of internal combustion engines, which describes an "indirect"

By contrast, the present invention teaches a novel Artificial Intelligence method that provides a "direct" approach to regulation by providing the method

Tax law "learns") The invention has the object to provide a method with which the above disadvantages are avoided or at least reduced, so that even strict emission standards can be met. This also includes the provision of a corresponding device

This object is achieved by the method for mixture control according to claim 1, which comprises the following steps: a) measuring at least one variable with which the air mass entering a combustion chamber of the internal combustion engine is related (so-called input quantity), b) determining at least one quantity of fuel to be supplied controlling output quantity as a function of at least the input variable (s) measured in a), with the aid of stored image information, c) supplying the fuel quantity corresponding to the output variable from b), d) measuring a quantity carrying information about the resulting mixture (so-called E) determining a deviation of the actual size measured in d) from a desired value for this variable, f) changing the stored image information as a function of the deviation determined in e) for the operating state measured in a), so that in a future run through of steps a) to e) in the same operating state di e Deviation is smaller, and thus realized a learning process This includes both the stationary and the dynamic operating conditions

Here, the term "measuring" is understood in a broad sense, which includes the actual physical measurement and, if necessary, the derivation of a variable thereof. The measured variable can therefore be an immediately measured magnitude or a quantity derived therefrom. For example, the measured value sensor for the actual quantity can be a lambda probe , which emits a signal corresponding to the residual oxygen content of the exhaust gas. In this example, "measuring" can, in addition to obtaining this signal, also determine the residual oxygen content and, if appropriate, determine the Lambda values include The mixture quantity of interest ia is the mixture ratio (ie the lambda value) It is possible to directly measure a mixture size or make a measurement of an exhaust gas quantity that allows a conclusion on the mixture size of interest (eg the lambda value)

In short, therefore, the invention teaches a learning mixture controller which compares the actual mixture relationship with a target value and, in the event of a deviation thereof, adapts the stored control information so as to achieve less deviation in future running of the same or similar operating point. The learning process of the invention generates one exact mapping of the real conditions under all relevant operating conditions, from which then the control information is derived to avoid any control difference, this is especially true for the dynamic operation of the internal combustion engine So it is, for example, with methods of Artificial Intelligence from mistakes to learn to avoid them in the future

In the following, the mixture formation is explained in more detail

In reciprocating engines, power, fuel consumption and emission are decisively influenced by the mixture ratio (also air ratio) λ This mixture ratio λ is defined as the ratio of air mass ( ^m _ac i) ^and fuel ^mass (m _fcyl ) in the engine cylinder (combustion chamber),

m λ = - acvl

0 ^{) K} X ^m jcvl

where K _{λ is} the factor for stoichiometric combustion (K, = 14 7 for gasoline)

The task of the engine control system is therefore to determine the mixture ratio as a function of

Optimum setting of the operating point according to the above criteria For example, for an emission-optimized operation of a gasoline engine with a three-way catalytic converter, an air ratio of λ = 1.00 can be achieved, which corresponds to an exactly stoichiometric mixture ratio Diesel engines, on the other hand, are operated with a variable, lean mixture, ie with a λ greater than one, at full load the mixture is enriched down to the so-called smoke limit at approximately λ = 1.5

The following illustration of the mixture formation refers by way of example to internal combustion engines with direct injection. As will be explained below, the invention can also be used advantageously in other engines.

For some time, the direct injection is known for diesel engines This also includes diesel engines with vortex or vortex chamber, if they are effectively part of the combustion chamber. Direct fuel injection for gasoline engines was already mass production more than 50 years ago, Daimler Benz military aircraft engines were with a mechanical Equipped with direct injection After the Second World War mechanical direct injection was used in the Mercedes 300 Sports Car and Lloyd small car, but in the end the mechanical outlay was unprofitable In view of increasing environmental protection requirements, such engines, which are expected to generate low fuel consumption, are up to date again

The advantage of direct injection compared to a conventional intake manifold engine in the exact metering of the fuel without the dynamic memory effect due to the local so-called wall film formation

It is expected that the efficiency of gasoline engines in the partial load range can be increased by the replacement of the throttle valve by means of independently controllable intake valves (hydraulic, electromagnetic) in addition to the gasoline direct injection

The mixture formation will be described mathematically below by way of example in various direct-injection piston-type internal combustion engines

The mixture ratio λ in the piston internal combustion engine is determined by the air and the fuel mass in the cylinder Therefore, consideration of the mixture formation is favorable as the intersection of two paths, the air and the fuel path First, the intake behavior, the so-called air path to be described: The engine sucks in open intake valves and running down the piston fresh air from the intake manifold. As a result of the so-called choked-flow effect, the inflow velocity of the air through the intake valves into the cylinders can reach maximum speed of sound. As a result, during the shorter with increasing speed intake stroke no complete pressure equalization between the intake manifold and cylinder; The cylinder thus contains less air mass than would be possible according to its volume in the thermodynamic Saugrohrzustand. Against this background, the suction behavior of the engine can be described mathematically as a pump with non-linear efficiency, the air mass (m _t ) that has entered the cylinder per intake stroke results

m * = JO ^~ r (2)

With:

T _m = temperature of the air mass p _m = pressure of the air mass

V _D = stroke volume of a cylinder

The dimensionless efficiency "volumetric efficiency" η _vol is non-linearly dependent on the intake manifold pressure and the speed of the engine (which determines the duration of the intake stroke) in engines with throttle valve (gasoline engine, conventional, suction or turbo), ie

rj _vol = η _vol (p _m , n) or η _voι = ^τ l _Vo , (P ^ », <x _a ) (4) or l _vo ι ^{= r} l _V o> (P _m ι ^fl, a ( ⁵ ) if the engine is equipped with a camshaft adjustable by a _s or with a variable intake manifold (a _sr ) (ie an intake manifold adjustable in the effective length)

If the intake valves are driven independently of a camshaft, that is, the load control happens z B on the duty cycle a _{a of} the intake valve, so the volumetric efficiency is determined by just this duty cycle and the speed, ie

rj _val = η _vol {a _a , n) (6)

According to equation 2, the air mass in the cylinder after the intake stroke is also determined by the pressure and the temperature in the intake manifold. If the engine is not equipped with a throttle, these thermodynamic conditions correspond to those of the environment, it can be assumed that these are only slow vary, with an adaptation can be dispensed with the measurement of both large If the internal combustion engine, however throttled operated, both large variables are rapidly variable, then z B also pressure and possibly temperature in the intake manifold can be measured to determine the air mass in the cylinder the air mass sucked into the combustion chamber per working cycle thus results in the non-linear dependency

with volumetric efficiency as a function of intake manifold pressure and speed or duty ratio and speed For variable intake manifolds or variable speed

Camshaft Steurungen also occurs the dependence of the volumetric efficiency of these control variables

In the following, the fuel injection, the so-called fuel path is described. Analogously to the intake behavior, the fuel injection by means of a If ideal inertial opening and closing of the injection valves and constant gasoline supply pressure were assumed, then the fuel mass injected into the cylinder per cycle would be calculated at the duty factor a _f

™ _fcy l = Ka _f (8)

with a constant K, determined by Ventiloffnungsquerschnitt and

Fuel supply pressure However, the dynamics of the valve needles and a low supply pressure loss in front of the valves are also a description here

Efficiency η _{f are} introduced. This efficiency depends nonlinearly on the

Off duration from, ie the duty cycle of the injectors _y and the

Drive frequency \ ln. This results in the non-linear dependence

NL _f = m _fcvl = η _} (a _} , n) Ka, (9)

for the fuel mass injected into the cylinder per working cycle

According to Equation 1 and the above derivations for the air path and the fuel path, the mixture ratio (the air ratio λ) is thus determined to be

ι = ^m ^ = J * - (io)

14.7m _fcyl K _λ NL _J

In Fig. L, this relationship is exemplified for a piston internal combustion engine with conventional intake valve control, and although Fig. 1 shows a "Signalflußplan" for the formation of the mixture ratio in direct injection piston internal combustion engines with conventional intake valve control The additional (thin drawn) dependency η _vol = / (.,., _c ) applies to engines with variable camshaft (gasoline and diesel engines) Here, for example, applies for a freischaugenden diesel engine p _m «p ₀ (/? ₀ = ambient pressure) or for a throttled Otto engine p _m <p ₀ In turbo engines, the intake manifold pressure may also exceed the ambient air pressure. If the piston engine in question is equipped with independently driven intake valves instead of a conventional intake valve control, the volumetric efficiency z B depends on the duty cycle a _a and the engine speed n can then be done by varying the duty cycle, the engine sucks unthrottled The Gemischverhaltnis is then formed, for example, according to the illustration of Figure 2, which shows a "Signalflußplan" for the formation of the mixture ratio in direct injection internal combustion piston engine with camshaft inlet valve control

During operation of the internal combustion engine, there is now generally the task of setting a desired mixture ratio by means of a control or regulation. This mixture ratio is determined from the quotient of air and fuel mass in the cylinder. Both air and fuel mass in the cylinder are not measurable and not without great effort (eg by calibration, time variance) can be set by varying the control variables p _m or a _a and a _f. However, the intake air mass or the injected fuel mass is - in the example explained here in more detail

Direct injection - always the same for the same size control

Engine operating points when the time variance is neglected as a result of aging or slow clogging of a valve Both

and NL _j are here static (possibly time-variant) nonlinearities and determinable only by a complex calibration The mixture ratio can be measured, novel methods are based on the interpretation of the ionization current in the electric spark ignition (gasoline engine), but conventionally, the measurement by means of faster switching Lambda probe The measurement by means of a lambda probe (eg wideband lambda probe or probe with a jump in the probe voltage at λ = 1) in the exhaust manifold (so-called bifurcated pipe) is, however, preceded by a speed-dependent dead time due to the internal combustion engine operation

From the above explanations, it will be understood that the mapping given for inventive methods is from the mass air mass (input size) to the size of the fuel For reasons of simplification, reference was made here, by way of example, to direct-injection engines. As will be explained in more detail below, the method according to the invention can also be applied to non-direct-injection engines which have a so-called "direct injection" engine. Have fuel storage effect.

According to the a priori not exactly known mapping is not - as in the prior art - mediated by a control, but by a learning control The learning can be done during operation, in which the various occurring in practice operating points (such may, for The method according to the invention is carried out in each relevant, ie achievable, operating point. After all the operating points have been run through several times-which is generally achieved relatively quickly in normal operation-the overall figure for all possible values is .PHI Input size learned

After learning, the controller then delivers instantaneously - ie without any control delay - the correct value of the output variable with high accuracy, even after a change of the operating point. Advantageously, the learning process continues to be carried out in order to enable an ongoing adaptation to disturbance variables These may be, for example, wear-related changes, changes in the intake air temperature, the coolant water temperature, the external air pressure, the oxygen content of the air, etc.

The inventive method is able to ensure the start of new operating points without any λ-deviation after learning the image conventional control approaches always occurs in contrast to a control difference, which is then integrated, for example, by an integrating portion in the controller as long as a control until the difference has become zero

The invention thus has the following advantages. Due to the self-adaptation of the image information, the accuracy requirements with which the image information for the image is reduced are reduced This reduces the effort for the development of a motor control and series development considerably, the control method according to the invention is robust with respect to series variations and time-varying disturbances, the desired lambda value is not only in steady-state operation, but also after a change of operating state (rotational speed). and / or load changes of

Motors) maintained without time delay with great accuracy, overall resulting in lower exhaust emissions and a lesser

fuel consumption

The invention thus contributes to the protection of the environment and the careful use of limited resources

Incidentally, the invention can also be used in the course of a fault diagnosis during operation. If the degree of adaptation required exceeds that of conventional series scattering and interference, let it be concluded that the condition is inaccurate, such as inadmissibly high wear or a defect due to appropriate evaluation of the degree of adaptation , for example, by a vehicle diagnostic system, engine failure or partial failure early -iΌg recognize during operation

In principle, the method can be carried out such that one or more of the method steps a) -f) are carried out averaged over several cycles or cycles of a cylinder. In a multi-cylinder engine with z B a common intake manifold and / or a common lambda probe then contribute to the measured in the cuts a) and / or d) due to the averaging several cylinders and possibly more cycles at preferred, the method is in time with the Working cycles of the individual cylinders carried out (claim 2), ie, the sequence of process steps a) -f) is carried out once in the context of a single stroke of a single cylinder Accordingly, for example, the measurement in step a) takes place during the intake stroke of a cylinder in step b ), the amount of fuel to be supplied on the basis of this measurement (and possibly previous measurements, more details below) is determined The supply of fuel in Schntt c) is then z B in the Immediately following compression stroke of the same cylinder, ie in the same cycle as step a) based on the same cylinder The following Schπtte d) -f) are generally carried out delayed due to dead time effects, but they are the cycle of steps a) -c) and the associated cylinder In principle, it is possible to perform the method not in each working cycle, but only in every second, third, etc. game, for example. Particularly advantageous, however, is an embodiment in which the method with the steps a) -f) is performed at each working cycle of each cylinder once processed

The most common gasoline engine is a throttled engine, which is usually controlled by adjusting a throttle valve disposed in front of a suction pipe in such a motor, the input size or - for several input variables - one of the input variables advantageously the pressure in the intake manifold (claim 3) determines this pressure The cylinder filling may be an engine with or without turbocharging. In the case of a turbocharged engine (eg turbo or supercharged engine), the pressure in the intake manifold may be temporarily or permanently above atmospheric pressure

Some engine designs make use of the dynamics of the intake air for a charge In some of these constructions, the intake system is designed to adapt the dynamic charging various Betnebsbedmgungen (see Automotive Handbook, Bosch, 1991, S 373, "switching intake systems") For example, the effective Saugrohrlange can be adjusted to use acoustic phenomena to increase the filling (keyword resonance charging) In these systems is advantageous a characterizing the instantaneous position of the intake system large (z B the effective Saugrohrlange), the input size or one of the input variables (claim 4)

For example, can be achieved by a camshaft adjustment, a shift of the opening and closing time and / or a change in the Anϊhungsdauer (see Automotive Handbook aa OS 370) In this case, are advantageous or a several valve timing parameters (eg Camshaft rotation and / or axial displacement), the input size or one of the input variables (claim 5)

Another type of engine, which could be of great importance in the future, has freely operable (eg electromagnetically actuatable) intake valves. Such an engine does not need to be throttled; it may be pressurized with atmospheric pressure upstream of the intake valves since the valves are open due to the free selectability of the ports Advantageously, the duty cycle and / or the closing and / or opening time of the intake valve is an input size or one of the input variables (claim 6). The duty cycle is related to the duration of the intake cycle or the total cycle Ofϊhungsdauer

It has already been explained above that the air mass reaching into the cylinder depends on the rotational speed of the internal combustion engine. The rotational speed is therefore advantageously an input variable or one of the input variables (claim 7).

As stated above, gasoline engines with 3-way catalyst usually betπeben with a constant mixture ratio of λ = 1, diesel engines, however, with a variable target mixture ratio The inventive method is suitable with a simple supplement for the latter Betnebsweise, and Although in the target mixture ratio the role of an input variable is assigned (claim 8) While in an Otto engine, the input large space z B is two-dimensional (it is clamped about by intake manifold pressure and speed), comes in this embodiment for a diesel engine as For example, third input variable adds the desired mixture ratio, so that here a three-dimensional input large space is clamped (for example by intake manifold pressure, rotational speed and target mixture ratio).

Preferably, in the internal combustion engine, the fuel supply by injecting Advantageously, the output quantity controlling the amount of fuel to be supplied then one or more of the following Great Einspntzdauer, Tastverhaltnis Einspritzventiloffhung, injection pressure, degree of opening of the injection valve (claim 9) With the duty cycle is also the Ofϊhungsdauer related to the Duration of a work cycle or a working cycle The degree of opening of the Einspntzventils can be controlled z B on the stroke of the valve needle

In today's most popular petrol engines, the fuel injection is not carried out directly in the cylinder, but in the suction pipe upstream them The fuel wets here first the suction tube wall (so-called wall-wetting effect) and then required for the transition to the gaseous phase a certain time. As a result, only a portion of the injected fuel to an operating cycle amount of fuel in this cycle the associated cylinder The remaining part is gaseous later and is therefore at later power strokes and - with common injection for several cylinders - partly burned in unallocated cylinders The invention Method is particularly advantageous in an engine used, which does not have such a fuel storage effect since no such delay needs to be considered here (claim 10) In particular, this is a direct-injection engine Even engines in which the injection takes place in a pre- or auxiliary chamber, have none Fuel storage effect, if after the combustion stroke practically no fuel in the pre- or secondary chamber remains

The absence of a fuel storage effect allows a particularly simple embodiment of the

Method in which the determination of the output variable (s) in step b) is based only on variables which belong to the currently performed working cycle - that is to say in particular on the result of the input size measurement carried out in the immediately preceding step a)

H)

The method according to the invention can also be used advantageously in internal combustion engines with a fuel storage effect. The amount of fuel injected into a cylinder at a particular power stroke generally depends on the quantity of fuel injected during preceding cycles. Preferably, in such an engine, large quantities of one or more preceding power strokes are also possible The determination of the output variable (s) in step b) may be involved. This may be, for example, the fuel quantities injected in these preceding cycles, which still have an effect in the current cycle. Their consideration allows a very precise control of the injection rate in the current cycle Kraftstofϊmenge Advantageously, the stored image information is in the form of a map containing the output size directly or indirectly. (Claim 13) A map is obtained, for example, by discretizing the (generally multi-dimensional) input size space and providing each input size cell formed by the discretization an output size Assignment of several output variables The discretization of the input space does not have to be regular, likewise the size of the input cells must not be constant. Learning takes place in such a way that in step f) the stored values of one or more adjacent input cells corresponding to a detected deviation be adjusted that a smaller deviation occurs in a future operation in the same input large cell

Particularly preferred is the mapping of the input size / s to the output size / s and the adaptation of the stored image information by a neural network algorithm (claim 14). Due to the parallelization of the data processing in so-called neurons forming the network, the use of a neural network - Algorithm the calculation time of the output size based on the stored image information compared to other interpolation methods (eg splines, linear or polynomial interpolation) significantly shortened using matched hardware While using known interpolation already by selecting the interpolation (possibly nonexistent Given that prior knowledge of the relationship to be approximated is assumed, a neural network can do without any prior knowledge. In particular, neural networks are suitable for the present adaptive method due to their evaluation and adaptation algorithm can be used very advantageously

Thus, according to this preferred embodiment, a neural network NN is set so as to map the control path of the air path (eg, intake manifold pressure) to the control path of the fuel path so as to achieve a desired mixture ratio. The control path of the air path is also considered from a historical point of view considered the control size, with which the driver affects the power output of the internal combustion engine The adaptive neural network NN is intended depending on the driver-influenced control size for the air path and possibly from Engine operating state (eg speed) the control size, here z B α _y for the

Output fuel path The output of the neural network thus corresponds to the duty cycle of the injection valves a _f , the input space of the neural network NN consists in this example of intake manifold pressure p _m and engine speed n, so we write

a _f = N N (p _m , n) (11)

The aim of the control is to have a desired air-fuel ratio for each operating point

m λ acvl (12)

^K X ^m y yl

In addition, we neglect the constant factor K _A. With the neural network NN as control according to a _t = et _f (/>",, //) = NN (p _m , ^ή ), the above equation can be considered as

_λ

If a certain mixture ratio λ _{oll is to be} achieved, an image must be learned in the neural network NN such that m _{ac l} = λ _∞n m _fcvl , ie

NL _a ( _Pm , n) = λ _should NL _f [NN

(14)

The output quantity is advantageous in this case by linking a the stored mapping information representing Stutz value vector and one of the / the A large gear ^¬ / n dependent activation vector formed (claim 15) This linkage is preferably linear, and has in particular the form of a dot product or - in the case of multidimensional output size - of a vector Matnx product (claim 16). According to the example above, one then obtains the control quantity a _f

a _f (P _m , n) = A A (p _m , t ή, (15)

where Θ is the trimming value vector and A (p _m , n) is the activation vector

It is advantageous that the activation vector is normalized and depends only on the distance of the input size / s to the support points on which the vector representation is based (claim 17). The standardization condition is z B

A (x) ^τ \ = l, (16) ie the sum of all components of the activation vector is always equal to one

The dependence of the activation vector A only on the distance of the input variable (s) to the support points can be expressed by the fact that its components A _t depend only on a variable which corresponds to the distance to the component belonging to the considered support point

4 = 4 ⁽ 4 ^{) (17)}

with the distances

d, = || -, || (18)

where ς _{t are} the bottlenecks, ie the locations of the neurons in the input space

Particularly advantageous is an embodiment in which the mapping of the input variable (s) to the output variable (s) is substantially local (claim 18). This means that for a given value of the input variable (s) substantially only the truncation value e contribute to the mapping to the output size (s) immediately adjacent to the input size (s). This is advantageously accomplished by having only the n component (s) of the activation vector obtains appreciably large values at close range to the input / s lie / lie, while components are negligibly small or disappear at a greater distance (claim 19)

Particularly advantageously, the components of the activation vector depend on the distance of the input variable (s) to the associated nozzle location according to a center function (claim 20), examples of advantageous center functions

~ -

A, ~ e ^{σ '} or A _^ , where σ is a (possibly variable) width parameter l + σ - d;

In principle, the "range" of learning does not have to match that of the figure. For example, it is possible to design the learning process locally (ie to adapt only sample value components in the vicinity of the deviation point), but to perform the mapping non-local (ie Advantageously, however, both ranges are chosen to be substantially the same, ie the adaptation of the image information takes place essentially in the same range of the distance from a deviation point, which is also reflected in the image of a is included at this point input size to the output size (claim 21)

Advantageously, the adaptation of the image information is carried out substantially locally to the deviation point (claim 22). This is let through, for example

Δ -Θl 'Pm «« Δ -Θ (19)

where ΔΘ is a trimming value correction vector. Particularly advantageously, both the adaptation and the image occur locally and with the same range More precisely, the adaptation of the image information is preferably carried out by adding to the bleed value vector a bleed correction vector which is proportional to a combination of the deviation value with the activation vector. (Claim 23) The linkage is in particular the product of the deviation value with the activation vector Adaptation eg according to

Θ = ηe <J> _m , r>) A (p _m , n) (20)

The factor η in this equation represents the learning step size. The large e is the measured mixture error at the operating point that corresponds to

When a broadband lambda probe is used, it is possible to deduce the magnitude of the deviation, whereas in the case of a lambda probe with an abrupt characteristic this information is only in the form of a statement "too rich" or "too lean". Actually speaking, by activation (P "^ ^fl ), the individual truncated values are selected, which depend on the relationship

_Q new _{= Q} old ₊ ^ Q ^

be adapted

If an adaptive method such as control of internal combustion engines is used, the learning process is done solely for safety and acceptance reasons preferably with an adaptive law with verifiable stability and parameter convergence (claim 24). Convergence here is the convergence against a global minimum of the distance between λ and λ _thus understood in all operating points approached for learning (not just convergence against a local minimum). Convergence then means that the learning process can only be completed when the trimming value vector opposes the Converged Only Possible, But Unknown Trimming Value Vector Reference is made to the following description of a proof of stability and convergence of a preferred law of adaptation.

In terms of the device, the object stated at the outset is achieved by a device for controlling mixture in an internal combustion engine according to claim 24, comprising: a) at least one device for measuring at least one quantity with which the air mass entering a combustion chamber of the internal combustion engine is related (so-called input variable b) at least one adjusting device for supplying fuel, c) at least one device for measuring a quantity, which carries information about the resulting mixture or its combustion (so-called actual size); d) at least one memory for receiving the variable

Mapping information; e) and a programmed for carrying out the method according to claim 1 and / or hard-wired computer

With respect to advantageous embodiments of the device, reference is made to the above explanations of the method

The invention will now be illustrated by means of exemplary embodiments and the attached drawing. In the drawing

Fig. 1 is a signal flow diagram for the formation of the Gemischverhaltnisses in a direct-injection piston internal combustion engine with conventional intake control;

2 is a signal flow chart corresponding to FIG. 1, but for an unthrottled internal combustion engine with freely actuable intake valves,

3 is a schematic representation of a direct injection gasoline engine with conventional intake valve control, FIG. 4 is a signal flow diagram corresponding to FIG. 1, but showing a neural network algorithm; 5 shows a so-called parallel representation of a neural network algorithm, FIG. 6 shows a signal flow diagram of a neural network algorithm with an input dimension,

7 shows diagrams representing the network output as a function of a one-dimensional input value, for illustrating the locality of the sample value adaptation, FIG. 8 shows an illustration of the interpolation and extrapolation behavior of a neural network algorithm,

10 is an illustration of an exemplary non-linearity for the air path, FIG. 11 is a fuel path diagram corresponding to FIG. 10, FIG. 12 is a graph of the time course of the output to start 13 is a diagram according to FIG. 12, but in the course of learning,

FIG. 14 is a diagram according to FIG. 12, but after the learning process is almost completed,

15 is a diagram of the mixture error in the course of the learning process, and

16 shows a comparison of the learning and the learned relationship

In the application of approaches with a learning process is the safety idea, the question is whether learning always leads to the correct result without wrong paths, a mathematical proof of stability and convergence should be feasible. The exemplary procedure to be described below becomes a special neural network Depending on certain measurable state variables of the internal combustion engine and / or the driver's request, the pulse duty cycle of the injection by direct injection (solenoid) valves as a control size is expediently the neural network next fixed stationary prior knowledge recognized as a correction element, which is adopted online so that always the desired Gemischverhaltnis is adapted.

With regard to the formation of Gemischverhaltnissen reference is made to Figures 1 and 2 and the associated previous embodiments The following description of exemplary embodiments will initially be made on a direct injection gasoline engine with conventional intake valve control with fixed control times according to FIG

The load control in this internal combustion engine type is done via the position angle of the throttle, the driver controls the air mass flow m _at in the intake manifold with open intake valves of the cylinder sucks during the downward movement of the piston in the intake stroke the air mass flow m _m until after closing the fresh air mass m, in the cylinder is available for combustion The fresh air mass in the cylinder depends decisively on the thermodynamic Saugrohrzustand pressure p _m according to equation 2 Finally, the driver so determined via the throttle position, the air pressure in the intake manifold and thus indirectly the load in the cylinder

From the above, it is apparent that the air mass located after closing the intake valves in the combustion chamber _{aal is} non-linearly dependent on the one hand of the

Engine speed and on the other hand, the intake manifold pressure

Namely, the engine speed determines the time available for intake, so that a rule of thumb is that at lower engine speed, more fresh air can be sucked in the same thermodynamic state as in the intake manifold

Due to the constructive design of the intake manifold and the intake manifold, the effect is achieved that a higher filling is achieved by acoustic phenomena (interference) in certain speed ranges

Internal combustion engines with variable intake manifold geometry exploited to increase the torque

In the case of rapid downflow of the piston with the intake valves open, there is a large sub-dive inside the cylinder relative to the intake manifold, so that the cylinder is out According to the thermodynamic flow equation for compressible media, however, the flowing medium can reach no higher than the speed of sound, so that the mass flow through the inlet valve for various operating ranges of the engine with increasing pressure difference can not increase further (laval ratio, choked-flow effect )

Overall, therefore, the suctioned air mass m, in this example depends continuously nonlinear from the intake manifold pressure and speed, with variable Ansauggeomtrie also of the control angle or the adjustable intake camshaft Mathematically we write so

"O = NL _a {p _m , n) (23)

Similar considerations can be made for direct fuel injection Due to the inertia of the valve needles, there is also a nonlinear relationship between injected fuel mass and duty cycle together with the sampling frequency (reciprocal of half the speed in four-stroke engines). Another effect that does not linearize the injected fuel mass According to these considerations, the injected fuel mass depends statically non-linearly on the valve opening time and the injection frequency (2 / n). According to these considerations, the injected fuel mass is dependent on the fuel injection system in a rotationally dependent manner.

Mathematically, we write

m _fM = NL _J (a _f , n) (24)

The idea of mixture control now consists in setting up a neural network NN so that the control path of the air path (here the intake manifold pressure) is mapped to the control amount of the fuel path so as to achieve a desired mixture ratio The adaptive neural network ΛW should output the control variable (here a _} ) for the fuel path depending on the engine operating state (in this case the rotational speed) and the driver-influenced control variable for the air path (here the intake manifold pressure).

The output of the neural network thus corresponds here to the duty cycle of the injection valves a _f , the input space of the neural network ΛW consists of

Intake manifold pressure p _m and engine speed n, so we write

a _f = NN (p _n ή ή (25)

The aim of the control is to have a desired air-fuel ratio for each operating point

πι χ ₌ - _____ _, (26)

to reach. In the following, we neglect the constant factor K _λ with the ΛW as control, after a _f - a _f (p _m t) - N N (p _m ή can the above equation be

to be written. If a certain mixture ratio λ _{oll is to be} achieved, an image must be learned in the NN, so that m _Ml - λ _M m _{} cvl} holds, ie

NL _a (P _m , n) = λ ^ NL _j [NN (/, "" //), "] (28) As an example of a multitude of adaptive neural network architectures, we use the so-called THANN, which makes the chosen approach particularly plausible. In addition, this algorithm allows a simple argument for stability and parameter convergence as well as a meaningful interpretability of the learned knowledge. The THEN is a particularly rchenzeitoptimale embodiment of a so-called RBF network with local support value effect. A detailed description of the THEN follows below.

So we start

^a f (> m, ⁿ ) = ®. ^T ά (p _m >") (29)

where the support values Θ of the neural network have to be adapted in order to compensate for both nonlinearities in equation 28 at the operating point (p ", ή.

This should be with the adaptation approach

® = ψ (P _m >) Ä (p _n "») (30)

with guaranteed stability. The factor η in this equation represents the

Learning step size. Figuratively speaking, the activations A (p _m , n) are used to select the individual support values, which depend on the relationship

Θ = Θ? + ΔΘ, (31)

be adapted. Now the support values at THEN work mainly locally, without a big mistake we can

ΔΘ - ηe (p _m , n) A (p _m , n) = ΔΘ | _n = ψ (32) (where the size to the right of the corresponding sign is the component of the Θ vector closest to the location (p _m , n)).

Because of the tact discrete working method of the piston internal combustion engine we replace in this derivation the continuous derivative (after the time) by the notation with delta sizes, which should correspond to a derivation after the power strokes of the internal combustion engine.

In order to mathematically determine the stability proof, we investigate the effect of this local adaptation according to Equation 32 on the mixture error e at the operating point, which increases

^e (P _m , n) = ^λ (P _m , ») - λ, _{o U} (33)

is defined as a deviation from the setpoint assumed to be constant here. So we decide

from equation 27

Ae (p _m , n) ₌ Aλ (p _m , n) a _f p _m >)

(35) ΔΘl -p, "Δα, (A.,") ΔΘI, _κ

follows. First we want the first quotient of the right side from above

A 1 Consider equation: is less than or equal to zero! This can easily be explained

Aa _f : At a constant air mass in the cylinder, the air ratio λ decreases with increasing injected fuel _mass m _fcyl , the mixture becomes fatter And with increasing

Duty cycle a _f , so longer valve opening duration at a constant speed, is of course also injected more fuel, but at least the same amount, less would be absurd It is therefore according to equation 27

because yes

Of course, it's big (or maybe equal to zero)

Now consider the second quotient of the right side of Equation 35. From the properties of the neural network THEN follows

^a _f (P _m ^, ")

> 0 (38)

ΔΘl

which means that when adjusting a Stutz value up at a certain point in the input space of the DANN / GRNN the output of the network at this point, of course, also increases

These derivations are possible because in the trim value adaptation because of the locality of the trimming effect of these networks

_n * ΔΘ = e (p _m) A _(Pm, n) (40)

applies.

We can now summarize

Ae (p _m , n) = Aλ ( _Pm , n) = ^ n ^ ΔΘ | ^ "(41)

and with the law of adaptation follows

Ae (p _m , n) = Aλ ( _Pm , n) = ~ ^ - ΔΘ ^ "= -kψ { _Pm , n) k ≥ 0 (42)

'Pm - ⁿ

-k, k> 0

This is the relationship

^A e (p _m , n) = -kη e (p _m , n) k Q (43)

derived, which proves the stability of the adaptation approach 18 For this we interpret equation 43 verbally

The sign of the error change is thus always inverse to the sign of the error, in other words, if the error is less than zero (e <0), its value increases (Δe> 0), its magnitude becomes smaller its value falls, so here too the amount is smaller Thus, the error can only converge to zero

In addition to the convergence of the mixture error at the operating point to zero, the properties of the THEN are followed by the convergence of the truncation values in the learned operating points against the only possible truncation values with which λ = _λsoU can be achieved This derivation applies locally, for the operating point (p _m , n) of the internal combustion engine is logical, again because of the locality of

Stutzwertewirkung that only in these operating points of the mixture error is compensated with high accuracy, in which the controller can learn To consider is still a learning correlation, namely a dead time because of the four-stroke work process. In the above derivations, we had assumed in the adaptation that in the mixture formation equal the resulting mixture error can be used to adapt the injection control Due to the four-stroke operation of the internal combustion engine, however, when using an exhaust gas lambda probe dead time (compression This is shown in Figure 4 This figure shows a signal flow plan for the formation of the mixture ratio in throttled direct-injection piston internal combustion engines The additional (thin drawn) dependence η _ιl = ( _,, cc _{i ir} ) applies to engines with variable camshaft or variable intake manifold Between mixture formation and lambda measurement is a dead time of three cycles because of the four-stroke working method of the internal combustion engine

For learning correlation we therefore have to consider this dead time in the adaptation. This is done expediently by a simple delay of the input values (p _m and) during the adaptation, so that the error signal and the truncation value to be adjusted correlate with each other. However, to form the injection signal (ct _j ), the unzeroed Input values are used

The example described above relates to the synthesis of a tax law for injection in direct injection internal combustion engines The description was based on a throttled gasoline direct injection with constant target mixture ratio If the target mixture ratio as for example in diesel engines assume a variable value, this setpoint can be as additional Similarly, with camshaft-driven intake valves whose control signal (duty a _a ) instead of the intake manifold pressure can be used as a network input is useful in a real implementation of the approach of the neural network as a correction element in addition to z off-line certain prior knowledge in the form of about one linear mapping p _m -> a _} This can significantly reduce the learning time The following table gives an overview of the nomenclature used here

^m _acyi mass fresh air in cylinder kg n _at mass flow fresh air through restriction kg s ^~ "m _m mass flow fresh air into cylinder kg ^s " ϊflf _cyi mass fuel in cylinder kg n speed of piston internal combustion engine s'

Po Ambient pressure of the air (slowly changing) Pa p _m Air pressure in the intake manifold Pa

R gas constant Jkgm

To ambient air temperature (slowly changing) K

Tχ Time constant of lambda probe s

T _m Air temperature m Suction pipe A ^* V _D Stroke volume of a cylinder m ³ a _cs Control angle with variable camshaft control (VANOS) rad a, _h Position angle of the throttle valve rad η _vo ι Volumetric efficiency - λ Air ratio - A _m Measured air ratio -

In the following, the so-called DNA is explained in more detail as an example of a suitable adaptive neural network algorithm. It combines the advantages of normalized RBF networks with the advantage of computing time and memory optimization

When using the so-called Distance Activation Neural Network THEN in FIG. 5 which shows a parallel representation of a THEN with two input dimensions, the convergence of parameters during the learning process guarantees parameter convergence and the problem of so-called persistent excitation bypasses parameter convergence means that the neural network Therefore, there is an unambiguous assignment of learned context and truncated values of the ΛW, this makes the learned knowledge interpretable. This decisive advantage compared to other adaptive algorithms is based on the form of the RBF network, especially the THEN the parameters to be adjusted during learning act mainly locally and the network thus provides a continuous output function above the input space with defined interpolation behavior. The extrapolation behavior of the THEN at points at which no knowledge in the form of learned parameters is available, is defined as well as meaningful: the output of the network corresponds to a (weighted) average of the learned knowledge in the learned environment; the THEN differs significantly from the original RBF network. The network consists of locally activated neurons, ie mainly the neurons in the immediate vicinity of the network input x are activated. The structure of the THEN is subdividable into activation and weighting. This illustrates the signal flow diagram representation of a THEN shown in Figure 6 with an input dimension (ie, scalar input x), A (x), and 9; are

vectors; y (x) - # A ( ^x ) ■ This makes it accessible to learning for the so-called delayed activation process.

Let's briefly describe the algorithm of THEN.

A scalar estimate y at a location x e 9T with a given set of q

Data points (support values θ at nodes ς) (θ "ς), / e [l, f /] and ς e 91" are defined by the equation at THEN

• _-i ι = l '1 + σd; - T

= 9A (x) (44)

1

Σ:., 1 + σd

with the distances

d, = l x - ς, l (45)

Equation 44 guarantees the boundedness of activation A x), which becomes normalization

^ (x) ^r l = l (46) That is, the sum of the activations of all neurons is always equal to one, and the activation of a single neuron is always a value between zero and one. From equation 44 it is clear that with very small smoothing factor σ almost only a single neuron is activated, Thus, almost only a single truncated value contributes to the result of the evaluation. If a truncated value is adjusted, the network output is also changed only in its surroundings. We call this situation "locality" of the pruning effect. For illustration, FIG. 7 shows the location of the pruning effect, the network output only changes in the environment of the adjusted stem value. This finding is important for the stability proof of learning in the method according to the invention

Equation 44 defines a continuous arbitrary non-linear output function y = f (x) (denoted estimated or adjustable magnitudes) with defined interpolation and extrapolation behavior as shown in FIG. 8. This figure shows the interpolation and extrapolation behavior of THEN, the crosses denote the existing data points Im learned area, the curve approximated by the truncated values (crosses) agrees with the sine function to be learned beyond the learned range, the estimated value tends to increase with increasing distance from the nearest truncated value to the average of all existing knowledge (truncation values) to the stored knowledge (data points) will result in the mean of the existing knowledge, in the immediate vicinity of a data point determines this mainly the network output If the nozzle values in their weight θ, defined as adjustable, then the simplest, shown in Figure 9 online structure to This representation applies to continuous-time systems The THEN shown can learn all static (without internal states such as memory) non-nite-data to group the activation method apart from a small approximation error due to the finite number of samples. For example, because of the above-mentioned memory effect, the information about previous events is included in the learning and imaging behavior. The learning structure illustrated in FIG. 9 is based on a known mathematical error model for which stability according to the direct method according to Lyapunov has been proven A one-dimensional nonlinearity to learn, as in Fis; 9 as the scalar product of y (t) to the THEN form 3_ = ^τ A (x (t)) + d be represented, where x (t), y (t) e 91 and

_, "__ e 91", the constant vector 3 is the unknown parameter vector of dimension n, where n parent values distributed over the input space can represent the real nonlinearity to be learned up to the error d - »0 with n -» ∞ 9, the observer error is calculated as e (t) = y (t) - y (t)

Parameter error vector defined as Φ (t) = (t) - 3_, the error equation is given

e (t) = A (x (t)) ^τ (t) (47)

After the error model 1 can

Φ (/) = -e (t) A_ (x (t)), (48)

are chosen as the guaranteed stable law of adaptation, because of the properties of the THEN (equation 44) the vector 4 _. ((0) ^{ιn a "en} its components is limited

This leads to a monotone decreasing function | Φ (t) | , which means lιm, _{→ OT} v9 (t) = 3

In the present embodiments, the illustrated learning ability of the THEN is used to synthesize a fuel injection control law by making the error between the target mixture ratio and the actual mixture ratio zero during learning

The above embodiment will now be demonstrated on a simulation. It shall be shown the ability of the approach that an image can be learned so that NL _a ( _Pm , n) = λ ^ NL _j [NN ( _Pm ή,] (49)

For demonstration, λ _{soU is} equal to 1 For purposes of illustration, for NL _a (p _m , ri), the static steady nonlinear relationship

NL _a (p _m , n = [1 - exp (-4 ^ _m )] exp (- ») (50), for NL _f (a _f , ή) we assume

NL _f (a _J , n) = 1 - - exp (-0 5 //) (51)

2 + a,

, a nonlinear function that increases monotonically with α _y

Air path and the fuel path assumed non-linearities are shown in FIGS. 10 and 11.

If these dependencies are known, then the mapping ΛW., Can be calculated analytically, so that equation 49 is satisfied with NN = ΛW ,,. The relationships according to equations 50 and 51 result in the representation AW ₍ soll

^a r (P _m , n) = N N "= 2 (52)

^/ ^ ^» 'Shall be l - (l - exp (-4 /;)) exp (-05 //)

which is to be learned by the controller synthesis

Speed n is normalized to maximum speed, as well as the intake manifold pressure to ambient pressure, so these both fluctuate between 0 and 1 For learning simulation, the input space »e [02,08] and p _m e [02,08] is now traversed throughout, in FIG this becomes clear in the learned control surface. FIGS. 12, 13 and 14 show the control curve according to equation 52 and the online learned from the network course of a _/ during the learning process These figures show the optimal course of the control size (dashed) and the network learned course (solid) Fig 12 shows the beginning of learning (which was started without any prior knowledge FIG. 13 shows the control curve course during learning, and in FIG. 14 the learning is virtually completed. The mixture error, which rapidly becomes smaller during learning, is shown in FIG.

The figure shows the relationship to be learned and the learned context, in each case the control surface. In the target context, only the area above the input space to be traversed during learning is represented by a simple Of course, it was only possible to learn in the entranced entrance area, so that only there is the knowledge meaningful

This simulation demonstrates that, without any prior knowledge, even after a large order of 10,000 cycles (distributed over the various operating points), an entire map for injection control is "learned" with very great accuracy

Claims

claims

Method for controlling mixture in an internal combustion engine, comprising the following steps: a) measuring at least one variable with which the air mass entering a combustion chamber of the internal combustion engine is related (so-called input quantity); b) determining at least one output variable controlling the amount of fuel to be supplied as a function of at least the one measured in a)

C) feeding the amount of fuel corresponding to the starting size from b), d) measuring a quantity carrying information about the resulting mixture (so-called actual size), e) determining a deviation of the value measured in d) Is big of one

Setpoint for this large, f) changing the stored image information as a function of the deviation determined in e) for the operating state measured in a), so that in a future run through steps a) to e) in the same operating state, the deviation becomes smaller, and thus realizing a learning process

Method according to Claim 1, in which steps a) to e) are carried out in cycles, the steps d) and e) being assigned to the clock in which steps a) to c) are carried out

The method of claim 1 or 2, wherein the internal combustion engine is a motor with or without charging and the input size or one of the input variables is the pressure in the intake manifold of the internal combustion engine Method according to one of claims 1 to 3, wherein the internal combustion engine is equipped with a variable intake system, and a position characterizing its size is the input size or one of the input variables

Method according to one of claims 1 to 4, wherein the internal combustion engine having an intake valve control with adjustable valve timing and the input size or one of the input variables are one or more valve timing parameters

Method according to one of claims 1 to 5, wherein the internal combustion engine has freely operable inlet valves and the input size or one of the input variables are the duty ratio and / or the closing and / or opening time of the intake valves

Method according to one of claims 1 to 6, wherein the input size or one of the input variables is the rotational speed of the internal combustion engine

A method according to any one of claims 1 to 7, wherein the variable ratio engine is controlled, and therefore, the target mixture ratio is one of the input quantities

The method of claim 1 to 8, wherein the output is one or more of the following sizes injection duration, duty ratio of Einspritzventiloffnung, injection pressure, opening degree of the injection valve

Method according to one of claims 1 to 9, in which an engine is used as the internal combustion engine, which has substantially no fuel storage effect, in particular a direct-injection engine

Method according to Claim 10, in which the determination of the output variable (s) in step b) is based only on variables which belong to the currently available working cycle Method according to one of Claims 1 to 9, in which an internal combustion engine with a fuel-storage effect is used, and in determining the output quantity / n in step b), also variables from one or more preceding operating cycles are included

A method according to any one of claims 1 to 12, wherein the stored mapping information is in the form of a map containing the output size (s) directly or indirectly, and in which one or more stored values are changed in step f)

Method according to one of Claims 1 to 13, in which the mapping of the input variable (s) to the output variable (s) is performed by a neural network algorithm

Method according to Claim 14, in which the output variable is formed by combining a trimming value vector (Θ) representing the stored image information and an activation vector (A) which depends on the input variable (s)

The method of claim 15, wherein the link is linear, and in particular has the form of a dot product

A method according to claim 15 or 16, wherein the activation vector (A) is normalized and depends only on the distance of the input size (s) to the trellis points underlying the vector representation

The method of any one of claims 14 to 17, wherein the mapping of the input quantity (s) to the output quantity (s) is substantially local

Process according to Claim 18, in which only those components (s) of the activation vector (A) are given appreciably large values, which in the at a small distance from the input size (s), while components at a greater distance are negligibly small or disappear

Method according to one of claims 15 to 19, wherein the components A, the activation vector (A) from the distance d _{t of} the input size / n to the associated support point i according to a Zentrumsfünktion, in particular

A ₁ ~ e - ^{σ "} or A _^ ι depend, where σ is a width parameter

A method according to any one of claims 14 to 20, wherein the variation of the mapping information occurs substantially in the range of the distance from a deviation location which has also been included in the mapping of the input quantity (s) at that location to the output magnitude (s)

Method according to one of Claims 14 to 21, in which the change in the image information takes place substantially locally to the point of deviation

Method according to one of Claims 15 to 22, in which the image information is modified by adding to the trimming value vector (Θ) a trimming value correction vector (ΔΘ) which is associated with the activation vector (especially the product of the deviation value (e)). A) is proportional

Method according to one of claims 14 to 23, wherein the adaptation of the stored image information or the learning process with a

Adaptation law with stability and parameter convergence happens

Device for mixture control in an internal combustion engine, comprising a) at least one device for measuring at least one variable, with which the air mass reaching into a combustion chamber of the internal combustion engine in the

Context stands (so-called input size), b) at least one adjusting device for supplying fuel; c) at least one device for measuring a variable that carries information about the resulting mixture or its combustion (so-called actual size); d) at least one memory for receiving the variable

Mapping information; e) and a programmed for carrying out the method according to claim 1 and / or hard-wired computer.

26. The apparatus of claim 25, wherein the computer for executing one or more embodiments of the method according to claim 2 to 24 programmed or wired.