WO2008095620A1

WO2008095620A1 - Method for optimizing an electronically controlled automatic transmission for a motor vehicle

Info

Publication number: WO2008095620A1
Application number: PCT/EP2008/000620
Authority: WO
Inventors: Ahmet Satiryigit
Original assignee: Fatec Fahrzeugtechnik Gmbh
Priority date: 2007-02-06
Filing date: 2008-01-26
Publication date: 2008-08-14
Also published as: DE102007006616B3

Abstract

The invention relates to a method for optimizing an electronically controlled automatic transmission for a motor vehicle. During a test bench phase (P), the transmission system behavior during a shifting process between two gear steps is varied for at least one operating point by varying control parameters (SPDOE) for this shifting process, wherein additionally shifting operations are carried out for a second set of operating points, each for a preferred control parameter set (SPBp) only. Algorithms of a transmission model (G) are formed from the variation of the control parameters (SPDOE), the algorithms empirically linking the universal behavior of specific characteristic parameters (KPDOE) to the variation of the control parameters (SPDOE). The algorithms are then applied to the second set of operating points, wherein a numeric simulation of the control parameter sets and a determination of the associated characteristic parameters is carried out based on the algorithms. In order to analyze the quality of a shifting process for a control parameter set, an evaluation factor BF is determined from the characteristic parameters by means of an evaluation function.

Description

Method for optimizing an electronically controlled automatic shifting transmission for a motor vehicle

Description:

The invention relates to a method for optimizing an electronically controlled automatic shifting transmission for a motor vehicle.

Switching operations between the transmission stages of a transmission are carried out by an electronic transmission control (EGS) depending on operating parameters, for example the transmission input torque, the transmission input speed or the transmission temperature, whereby also driver actions, such as a manual change from a comfort setting to a sport setting, can affect the switching operations by the EGS. In order to ensure a satisfactory switching behavior, an individual adaptation is required for motor vehicles for every combination of vehicle body, engine and transmission.

Currently, the optimization of the switching behavior of the electronic transmission control is largely done in driving tests on the road by a subjective evaluation of the test driver and by recording specific measures. The variation of the control parameters responsible for the shifting sequence within the EGS for the various shifting operations, types of shifting and operating points is made manually during the driving test via a suitable interface, for example a notebook. Substantial subjectively or with appropriate measuring devices determinable variables of a driving test are in particular the switching time, which is required for a switching operation (spontaneity), and the temporal behavior of the longitudinal acceleration (comfort).

To reduce the effort associated with driving tests, it is known to model the typical driving conditions on a chassis dynamometer, starting the desired load and speed ranges automatically by a driving robot. At an operating point determined at least by the transmission input torque and the transmission input rotational speed, the various control parameters are varied, wherein for each control parameter set, characteristic parameters are determined with which one is able to improve the quality of the circuit used, for example with regard to comfort and spontaneity grading characterize. In order to limit the scope of the parameter variations, after determining and evaluating the characteristic parameters for a control parameter set, the control parameter set to be investigated next can be done directly with the aid of a fuzzy logic based model, for example, with the correlation between the individual characteristic parameters and the various control parameters was determined on the basis of a comparative measurement. By means of the model, it is possible to determine an adjustment value for each control parameter from the characteristic parameters determined for a parameter set, which is added or subtracted, taking into account the sign, to the old control parameter value, so that a new control parameter value represents the result of an optimization step. An optimization on the basis of a neural network is known from DE 102 08 205 A1. For the neural network to provide reliable values, a complex training phase is required. Relationships are defined by the neural network, which may be different depending on the respective test circuits performed. In addition, approaches are also known in which the parameter optimization of an electronically controlled automatic shifting transmission is carried out by a simulation with a numerical model, wherein the numerical model is formed on the basis of very extensive, previously determined on a dynamometer data sets.

From DE 196 43 305 A1 a method for determining characteristics of an automatic transmission is known in which circuits are carried out between individual gear ratios on a final test bench. It will be determines the time courses of the transmission input speed, the transmission output speed, the transmission input torque and the transmission output torque. From these measured variables, a filling time, a filling pressure, a reaction time, a coefficient of friction of the lamellae and a pressurized fluid charge are determined as characteristic quantities of the automatic transmission for the clutch engaging in the switching operation. These parameters are stored in a memory, with the EGS correcting the pressure level and the switching times as a function of these parameters. The selected parameters are characterized by the fact that they allow direct conclusions about the transmission behavior, in particular control and gearbox tolerances can be detected and corrected. For a precise determination of the parameters, a large number of circuits are to be carried out between the individual gear ratios in the design of the automatic transmission.

Based on the stored in the EGS for various switching operations control parameters that have been determined specifically for the present combination of vehicle body, engine and transmission in the context of optimizing the electronic gearbox, it is known in the operation of the vehicle, an adaptive adjustment of the control parameters, for example based on data fields stored in the EGS and determined during the initial optimization. The adaptation can be done for example due to manual settings or as described in DE 100 57 093 A1 for the correction of lifetime loads. From DE 199 16 006 A1 it is known to gradually adjust the actuating pressure of a hydraulically actuated clutch on the basis of an adaptation characteristic field on the basis of external parameters. In order to achieve a quick adaptation as a whole, a corresponding adaptation also takes place for similar switching operations.

The publication DE 102 38 474 A1 relates to a method for adapting shifting sequences of an automatic transmission, in which adaptation values depend on speed of the shift quality determining events takes place. In order to be able to take into account manufacturing tolerances and wear, the adaptation to the individual vehicles equipped with the automatic transmission takes place on the basis of an individual adaptation characteristic determined for the respective vehicle.

The invention has for its object to provide a method for optimizing an electronically controlled automatic shift transmission for a motor vehicle, which allows a reduction in the useful life of a chassis dynamometer or Aggregateprüfstandes to reduce costs.

The object is achieved by a method according to claim 1.

Characteristic parameters in the context of the invention are parameters which are suitable for characterizing the circuit which has been carried out and for describing them by objective numerical values. Typical characteristic parameters are, for example, the duration of the switching process, which is a measure of the spontaneity, and the time course of the longitudinal acceleration or the change in the connection force of a rigidly held on a dynamometer motor vehicle. Each characteristic parameter (KP _x ) can be represented as a function of the various control parameters (SP _j ) and the respective operating point, with respect to the operating point often only the transmission input speed and the transmission input torque and not the transmission temperature are taken into account. Accordingly:

KP _x = f (SPi, ... SP _j , transmission input speed, transmission input torque)

For the optimization with the aid of an aggregate test stand or a roller test bench, in particular the switching and reaction time determined from the time course of the transmission input rotational speed and frequency-weighted Characteristic parameters of the connection force of the motor vehicle to a holder during the circuit particularly meaningful characteristic parameters.

The invention is based in particular on the knowledge that the characteristic parameters behave qualitatively the same with differentiation of the control parameters at different operating points as a function of the variation of the individual control parameters. The difference in the behavior of the characteristic parameters with variation of the control parameters at different operating points is therefore essentially limited to the quantitative classification of their physical units.

In the context of the method according to the invention, the transmission system behavior in the test bench phase for predetermined parameter sets is determined on a transmission on an engine test bench or on a gearbox built into the vehicle on a chassis dynamometer. The vehicle is fixed on the chassis dynamometer by means of a vehicle restraint. Instead of the longitudinal acceleration no longer present on the test bench, the connection force of the vehicle to the vehicle binding is recorded and evaluated.

In the test bench phase, in a shift between two gear stages for a first set of operating points determined by the transmission input speed, the transmission input torque, and the transmission temperature, a set of control parameters is varied on a given matrix in a plurality of shifts, respectively. For each set of control parameters that determine the switching sequence, associated characteristic characteristic parameters are determined. In the control parameter variation for the first set of operating points, the greatest possible information content with respect to the system to be identified is to be determined from the smallest possible number of attempts. In the context of the method according to the invention, it is preferably provided to identify the transmission system behavior by intelligent test planning. This is often spoken of so-called DoE (Design of Experiments) plans, which generate the best possible test plan for different boundary conditions on the basis of various optimization criteria. In particular, in the context of the present invention, a so-called D-optimal test plan described in the specialist literature is suitable for defining the control parameters SP 1 to SP _j for each of the tests to be carried out. For the circuits started under these conditions, characteristic parameters are determined from the connection force signal at the vehicle binding and the transmission input speed, with which the quality of the circuit that has been effected is characterized (objectification). This results in a link between the control parameter sets on the one hand and the associated characteristic parameters on the other.

In the test phase, switching operations are also performed only for a preferred set of control parameters, even for a second set of operating points. In order to avoid measurement inaccuracies, for example by a measured value noise, the switching operations of the second set of operating points without a variation of the control parameters can be approached several times in succession, where from the measured values measured then identification parameters are identified and fed to a Ausreißerbereinigung and averaging.

After the test phase, algorithms are determined within the framework of an empirical, numerical model formation from the characteristic parameters determined for the first set of operating points, which link the general behavior of the characteristic parameters with the variation of the control parameters. To determine the algorithms, various methods such as higher order polynomial models or artificial neural networks can be used. The basic goal is to mathematically describe the behavior of the system identified by the test bed measurement. The generated model has a statistical certainty that satisfies the statements Salary quantified. Especially with regard to the transmission of model information to unknown operating areas, ie to the second set of operating points, it is necessary to achieve a particularly high explanatory power of the model. In a preferred embodiment of the method according to the invention, it is provided that the characteristic parameters are represented by a normalization as dimensionless variables. The modeled characteristic parameter behavior is therefore no longer expressed in physical absolute values, but in dimensionless values.

In a further method step, the operating point adaptation, the algorithms previously determined in the modeling are applied to the second set of operating points, whereby a variation of the control parameters is numerically simulated on the basis of the control parameters which were determined for the second set of operating points in the test bench phase and determining parameters for the simulated control parameter sets. The application of the algorithms to the second set of operating points is possible because the qualitative behavior of the characteristic parameters in the variation of the control parameter sets for different operating points matches qualitatively, so that the difference in the Kennparameterverhaltens and the variation of the control parameters at different operating points on the quantitative classification of their physical units is limited.

This means that for an operating point of the second set of operating points, the assigned characteristic parameters must be known only for a preferred set of control parameters (actual state), on which basis the numerical modeling can be carried out with the generally valid algorithms determined during the modeling. From the characteristic parameters, an evaluation factor for the quality of the associated switching operation is determined for each control parameter set and each operating point by means of evaluation functions, wherein the evaluation functions are preferably determined by a fuzzy logic model. In a preferred embodiment of the method according to the invention, it is provided that in an optimization cycle by means of a numerical simulation using the gearbox model, a variation of the operating point-specific control parameters is performed in order to determine a maximum weighting factor, wherein typically the control parameter sets having the highest weighting factors are stored. To verify the numerical simulation, the stored control parameter sets can be checked in a final verification phase by test measurements on the chassis dynamometer or an aggregate test bench, in which case the actually most suitable control parameter sets are selected from the stored control parameter sets.

The invention will be explained in the following by means of an exemplary example with reference to the drawings. They show schematically:

1 shows the structure of the example of a chassis dynamometer,

Fig. 2 is a schematic representation of the inventive method for optimizing an electronically controlled automatically switching

Transmission for a motor vehicle.

Fig. 1 shows a chassis dynamometer 1, which is used in the context of the inventive method in a test phase P to determine the transmission system behavior of a built-in a motor vehicle 2 transmission. The motor vehicle 2 is connected rigidly to a support 4 via a vehicle restraint 3, wherein a force-measuring element for determining the connection force is integrated in the vehicle restraint 3. The temporal course of the connection force during a switching process is an important tuning signal for comfort evaluation, which is quasi-proportional to the longitudinal acceleration in a moving vehicle. The electronic gearbox Controller 5 (EGS) is connected to a control computer 6, whereby the control parameters of the EGS 5 can be adjusted in real time between individual switching operations. In addition to measurement data recorded by the EGS 5, the time course of the connection force is also recorded in the control computer 6. At the same time the driving resistance of a roller 7 of the chassis dynamometer 1 and the settings of a driving robot 8 is varied by the control computer 6, wherein the start of desired load and speed ranges is performed by the driving robot 8, which operates the accelerator pedal and can change the speed levels. A deceleration or acceleration of the motor vehicle 2 is typically carried out by an active intervention in the control of the chassis dynamometer 1, so that the service brake of the motor vehicle 2 remains largely unused. Switching operations are usually not performed by moving the selector lever in manual mode, but by a digital command on the EGS 5, so that time delays between the gear request and tripping are avoided.

Fig. 2 shows schematically the sequence of the method according to the invention, wherein first in the test bed phase P on the test stand 1 shown in Fig. 1 switching operations for the identification of the transmission system behavior are performed. According to the invention, the further modeling and optimization is carried out using numerical models.

Due to the fact that test bench hours for chassis dynamometers are relatively expensive, the test stands are rather rare and consequently the workloads are very high, it is always necessary to keep the duration of the effective test stand usage as low as possible. Consequently, when identifying the transmission system behavior, methods based on efficient design of the experiment must be used.

In principle, the design of experiments is an instrument with as few attempts as possible to maximize the information salary from the system to be identified. Transferring to the automated tuning of shifting comfort is therefore the challenge of intelligently adjusting the control parameters in the transmission so that a maximum of the transmission system behavior can be identified. Each so-called attempt thus represents a combination of different control parameter settings, the system response of which is described by one or more start-up of an operating point in the form of the changing characteristic parameter behavior.

In the context of the method according to the invention, the transmission system behavior is identified by intelligent test planning. This is often referred to as so-called DoE (Design of Experiments) plans, which generate based on various optimization criteria for each of the different boundary conditions best possible experimental design.

In contrast to purely factorial design plans, in a D-optimal measurement the control parameters are adjusted together - apparently randomly. In constructing these designs, an algorithm works that maximizes the volume of geometric bodies spanned by the experimental points in a given x-dimensional (eg, 9-dimensional) experimental space so that the informational content also assumes a theoretical maximum. In a D-optimal design, the interactions of the various control parameters should be clearly identified. In this case, simple interactions between in each case two of the control parameters are typically determined by a corresponding variation, whereby usually a determination of higher interactions in the simultaneous adjustment of a plurality of control parameters does not usually have to be considered, whereby an efficient determination of the transmission system behavior is possible. For the circuits started under these conditions, characteristic parameters are determined from the connection force signal and the transmission input speed with which one is able to characterize the quality of the circuit that has been made (objectification). The result is therefore a table in which the corresponding characteristic parameters KP _DO E are available for each test circuit line, ie for each test program line and control parameter combination SP _D0 E. Inputs and outputs of the measured system are therefore clearly defined. The first line of the test plan stands for the so-called reference line, ie for the current state of the EGS 5 without the variation of control parameters. In the following, this control parameter combination is called SPD _O E, Ref.

With regard to the optimization of operating points of a circuit type, which are not part of the DoE measurement, the actual conditions with regard to the control and characteristic parameters of the relevant points are still required for the application of the new optimization cycle. In order to avoid measured value noise, the circuits of the new operating points are repeatedly approached without a control parameter variation and the values set in the EGS 5 are identified. From the measured data obtained, characteristic parameters are then identified and fed to an outlier cleaning and averaging. The result for each new operating point are values of the considered control parameters SP _B p and outlier-corrected mean values of the calculated characteristic parameters KP _BP .

The empirical modeling of the transmission model G can be done by numerous methods, such. B. polynomial models of higher orders or artificial neural networks. In general, the goal is to mathematically describe the behavior of the system identified by the test bench measurement. Therefore, the behavior of the parameter KP _DO E with the

Control parameter values SPD _O E of the EGS 5 link, so that functions according to the equation KPD _O E _{, X} = f (SP _DO E _, I, SPD ₀ E _J1 transmission _input speed, transmission input torque) (1)

the goal is.

Each generated transmission model G has a statistical certainty that quantifies the meaningfulness. Especially with regard to the transmission of model information to unknown operating areas, d. H. For new operating points, it is important to achieve particularly high explanatory levels of the models.

A special feature in the application of the process cycle shown in FIG. 2 is the processing of the data necessary for the model creation (control and characteristic parameter values). Modeling and application are carried out with respect to the transmission of the model information to unknown areas preferably in the general, normalized range between zero and one. The modeled characteristic parameter behavior is therefore no longer expressed in physical absolute values, but in dimensionless values.

The operating point adaptation A has the objective of transferring existing information from a detailed operating point to new operating points which were not the subject of the DoE measurement.

This is possible because the characteristic parameter behavior behaves qualitatively similar when measuring different operating points. The difference in the characteristic parameter behavior with variation of the control parameters at different operating points is therefore limited to the quantitative classification of their physical units.

This means that, for a new operating point, the actual state with regard to the control parameters SP _B p and characteristic parameter KP _B ρ must be known once, in which case the optimization is carried out with generally valid (dimensionless)

Kennparamermodellen can be done. The absolute characteristic parameter value KP _B p The new operating point must therefore be normalized. However, the characteristic parameter limits are needed for the normalization of the absolute characteristic parameter value KP _BP . To be able to save measurements of the new operating point on a test bench and still have knowledge of the Kennparametergrenzen, transformations of the existing information material are completed.

From the system identification of the test bench examinations, the control parameter values of the DoE test plan (SP _DO E) set in EGS 5 are available. This is a matrix with i rows (i tries, ie i different control parameter combinations) and j columns (j varied control parameters). The first line of this matrix has the name SP _Ref and stands for the so-called reference line, in which the gear is not adjusted. From these data and the associated parameters KP _DOE of each of the i test bench measurements, the empirical characteristic model models are generated on a standardized basis.

CD CD CD

^OJ DoE (2, \) ^OI DoE (I 2) - ^{C> r} DoE (2, j)

SP DaE ~ with SPR ₆ PJ = SPD ₀ E _{01 J)} (2)

CD CD ^ ¹ DoEOA) ^o - * DoE (ι, 2) • • • ^{o /} DoE (ιj)

Also known is the information of a new operating point, which is not the subject of DoE measurement and has been approached several times on the test bench. After the statistical processing of these so-called calibration measurements, the control parameter values SP _B p = [SP _B pi SP _B p.2 ■■■ SP _B p ,, J and, on the other hand, associated characteristic parameter values KP _B p of the new operating point.

The goal is now the calculation of a standardized characteristic parameter value KP _B p, n (0 <KPBP, _n <1), which is required for the evaluation process B and the optimization process O. However, knowledge of the minimum and maximum characteristic parameter values of the new operating point for normalization is necessary for this purpose. It therefore applies below to calculate these Kennparametergrenzen the new operating point.

In this case, the construction of an operating point-specific test plan is started. It should be carried out virtually, ie on the basis of the model applications in the office for the new operating point, an identification of the transmission system behavior. Thus, for the new operating point, a new design is called, called SPD _O E.BP.

So that the model information can be transferred to the new operating point, it is necessary to create the same conditions with regard to the reference line. A control parameter starting value, in which z. If, for example, the ratio between the pressure to be switched on and off is five times higher than the original DoE measurement on the test bench, the characteristic parameter behavior will be different overall, while maintaining the same percentage control parameter variation, than one in which the ratio will be much lower. It is therefore necessary to make the reference line of the new test plan (new operating point: SP _B p, R _e f) comparable with the actual measurement (SP _Ref ).

In order to calculate the control parameter _{relations of} the reference line (SP _Rβf ), a start value is assumed which is compared with the other control parameter values and factored. It has been shown that this preferably the control parameters are to be selected, which due to their high Sensitivity already have significant parameter changes at low adjustments. The reference line is:

SJ ³ Re / = [SP DoK (IJ) SP DoE (1, 2) ■■■ SP DoH (Ij)] ⁼ [SP Ηef, I SP _Re f ₂ ... SP 'R _e fJ (5)

In this example SPD is _O E ( _I , ₂ )

the highly sensitive control parameter, hereinafter referred to as SP _Ref, s. Starting from the identified control parameters SPBP of the new operating point, the transmission of the control parameter ratios takes place with the aim of calculating a reference line of the new point. Starting from this line, the operating point-specific test plan is set up. The operating point-specific reference line is calculated by the relationships between the sensitive control parameters of the known and new operating point and the original reference line. It therefore applies generally to the control parameter j:

* - "/» ', Ref, j _ * - "ßf, Ref, S / gx

VD CD * '

^Or Ref, j " ⁷ IUf, p

As a result, the operating point-specific reference line is calculated as follows:

SP ^ ⁷ BP ₁ RCf ₁ S ς, _p '- »' gP.Ref.S _Qp S * BP. Re (, S ς, _p

H ₁ P.Ref ς, p ^{■ O7} Ref, l _cp ^{"ύ /} Ref, 2 ^*"' o _p ^{' Or} Ref, j (7)

^Or Ref, S ^or Ref, S ° ^r S / \ Ref, S

On the basis of a factorization of the existing test plan of the test bench test (SP _DoE ) and the subsequent multiplication of this ratio with the new reference line, an operating point-specific test plan results: SR

SP, DoE

DoE, BP ■ SP ß, / ^> , Re f, thus (8)

SR Ref op ep pp o ^r DoE, BP (\, \) ^OJ DoE, BP (\, 2) ^{• • • or} DoE, BP (\, j)

CD cp O D

^OJ DoE, BP (2, \) ^OJ DoE, BI '(2,2) ^{• • • o /} DoE, BP (2, j)

SP D, OE, BP O) o ^r DoE, BP (ι \) ^ ¹ DoE, BP (ι, 2) SP D, OE, BPOJ)

The matrix SP _DO E, BP has - just like SPD _O E - i rows and j columns. In addition, the newly measured control parameter combination of the test bench (SP _DOE.P _B P) is included in this new design as row (i + 1). For this line, the characteristic parameter value is already known by the measurements, so that the value KPBP is also adopted.

This newly created matrix, ie this newly created operating point-individual experimental plan with (i + 1) rows, is normalized for every column, ie for each control parameter, with 0 <(SPD _O E, BP _· j X j £ 1.

For the calculation of the model answers, this new test plan is virtually "traced" by using the characteristic model models. The result is a vector with normalized characteristic parameter values of length (i + 1) named

KP _D oE, BP, n with 0 <(KP _DoE , BP, n) l. j <1 -

Furthermore, via an inverse transformation, i. H. By dividing all lines with the last (i + 1-th row), the absolute characteristic parameter values of the new operating point are calculated. The normalized characteristic parameter vector is multiplied by the knowledge of the characteristic parameter value (KPBP) measured on the test bench and transformed back into an absolute value:

KR

KP, DoE, BP.n

DuE, BP KP

KR B _> P (10)

DoE.BP, n (i + l) With the help of this vector - filled with absolute values - one obtains the minimum and maximum characteristic parameter limits of the new operating point. Then the measured characteristic parameter value of the new operating point can be normalized. The result is KP _B p, _n (0 <KP _B p, _n <1), which finds use in evaluation process B.

The evaluation process B has the task of evaluating the quality of the simulated switching operations and in particular the suggestions of the optimization cycle O, so that based on these results new control parameter values can be calculated, which gradually approach an optimum value.

The evaluation is based on the principles of fuzzy logic, whereby a weighting factor BF is calculated, which quantifies the switching processes simulated in the office by calculating a key figure.

Inputs of the evaluation system are the normalized characteristic parameter values KP _BP . _Π of each characteristic parameter considered (eg switching time, delay time, gradients, peak-to-peak values, ...). For each standardized characteristic parameter value KP _B p, _n , a degree of membership is calculated which takes input into the set of rules (inference). The result is a weighting factor BF, which quantifies and outputs the relevant Kennparamterkombination.

The weighting factor BF is thus - in addition to the formulated relationships in the set of rules - dependent on the calculated degrees of membership of the inputs, which in turn depend on the nature and location of the membership function. If one moves the membership functions of the inputs on the abscissa to the right or to the left, one likewise shifts the target value of the relevant characteristic parameter. A high weighting factor BF is only obtained if the characteristic parameters input into the fuzzy system find that meet these targets. By actively shifting the membership functions of the input, it is thus possible to set the target specifications. This process is called abscissa shift. A high weighting factor BF therefore means that the parameter combination found corresponds very well with the formulated targets.

In the context of the new optimization cycle, preference is given to working with so-called fuzzy actual and fuzzy target systems, which differ only by the shifts in the membership functions. The actual state is set by knowledge of the normalized characteristic parameters KP _B p, _n (= KPBP, _n , ιsτ) _> the target state corresponds to the target specification, ie the optimization target.

The optimization cycle can be carried out in the context of the inventive method based on an evolution model. It belongs to the class of stochastic search methods and also allows the treatment of complicated problems, which are no longer manageable due to excessive computational effort with traditional optimization methods.

As part of the optimization cycle, a standardized characteristic parameter target value is calculated:

^ BP, n, IUI = KPBPJ, ^{(1 1)} ^KP _S _BP, SOU = KPBP _S JST ^{■ F} o _P, F with the optimization factor 0 _<Opt <1. (12)

In advance, a theoretically maximum possible weighting factor (BFMAX, theoretical) is calculated. This results from the generation of a full-factorial design of all involved parameters, which is included in the assessment system as part of a pre-simulation. The value of maximum found from these calculations is as BF _MA x. _t h _eo r _et i _sch . The definition of the parameter-specific optimization factors Fo _pt depends on the respective user specification. If, for example, the optimum control parameter set is searched for a switching operation that is as comfortable as possible with the same switching spontaneity, then the optimization factors for the switching times are set equal to one, while the force differences or gradients determining comfort start at values of less than one.

With regard to this example, the optimization factors of the comfort determining parameters are started with relatively small factors, and it is assumed that there is no solution in the search space corresponding to these targets, which is expressed by small weighting factors. After the optimization process, the evaluation factors BFM _A X found are compared with the theoretically maximum possible weighting factor and the associated control parameter combination is stored as soon as the value is greater than 90% of BF _M Ax, theoretical. For the next run, the optimization factor is increased and approximated to the actual state. As soon as the characteristic parameter setpoint value exceeds the actual value by the continuous increase of the optimization factors, the optimization cycle O is ended.

The fact that not only one set of control parameters is saved, but several sentences, is based on the fact that the office uses several models, all of which have a model error, ranging from good to very good statistical confidence on average at typically six to six ten percent moved. Furthermore, the assumption that existing information is applicable to other operating points inevitably leads to inaccuracies covered by verification measurements. For example, the ten best optimization results are secured and verification by a verification phase V on the test bench.

By the method according to the invention, the test bench utilization times with respect to the known from the prior art method can typically be reduced to one third, with an even greater time savings can be achieved with increasing number of gear stages, since a parameter optimization always for all upshifts and downshifts between must be performed in each case adjacent transmission stages.

Claims

claims:

A method of optimizing an electronically controlled automatic shift transmission for a motor vehicle,

wherein in a test phase the transmission system behavior for predetermined parameter sets is determined on a transmission on an engine test bench (engine-transmission test bench) or on a vehicle-mounted transmission on a chassis dynamometer,

wherein in the test phase in a switching operation between two gear stages for a first operating point or a first set of operating points, which are determined by at least the transmission input speed and the transmission input torque, each of a plurality of switching operations, a set of control parameters that determine the switching sequence, based on a predetermined matrix and characteristic characteristic parameters are determined for each control parameter set,

wherein switching operations are carried out in the test bench phase for a second set of operating points in each case only for a preferred set of control parameters,

wherein algorithms are then determined from the characteristic parameters determined for the first operating point or the first set of operating points, which empirically link the general behavior of the characteristic parameters with the variation of the control parameters, wherein the algorithms are applied to the second set of operating points and, based on the control parameters determined for the second set of operating points in the test bench phase, a variation of the control parameters is numerically simulated and wherein characteristic parameters are determined for the simulated control parameter sets and

wherein an evaluation factor for the quality of the associated switching operation is determined from the characteristic parameters for each control parameter set and each operating point by means of a weighting function.

2. The method of claim 1, wherein the characteristic parameters are represented by a normalization as dimensionless quantities.

3. The method of claim 1 or 2, wherein the evaluation function for determining the weighting factor is determined by a fuzzy logic model.

4. The method according to any one of claims 1 to 3, wherein the algorithms which empirically link the behavior of the characteristic parameters with the variation of the control parameters, are determined from the mutually associated quantities by means of an artificial neural network.

5. The method according to any one of claims 1 to 4, wherein new control parameters are proposed in an optimization cycle, wherein for these control parameters with the algorithms that empirically link the behavior of the characteristic parameters with the variation of the control parameters determined by a numerical simulation characteristic parameters and wherein a weighting factor is determined for these characteristic parameters.

6. The method of claim 5, wherein in the optimization cycle certain preferred simulated control parameter sets are stored and stored in one Verification phase can be checked by measurements on the chassis dynamometer or the unit test bench.