WO2019169421A1

WO2019169421A1 - Method for determining a vehicle parameter of a vehicle data set of a vehicle, and use of the vehicle parameter at a test bench

Info

Publication number: WO2019169421A1
Application number: PCT/AT2019/060079
Authority: WO
Inventors: Felix Pfister; Stefan GENEDER; Bernhard Schick
Original assignee: Avl List Gmbh
Priority date: 2018-03-09
Filing date: 2019-03-08
Publication date: 2019-09-12
Also published as: AT520827A4; AT520827B1

Abstract

The invention relates to a method for determining the value of at least one vehicle parameter of a vehicle data set of a vehicle, a sub-component or a sub-system of a vehicle, as well as the use of a determined vehicle parameter for parameterising a simulation model (23) of a simulation (6) for carrying out a test on a test bench (10). For this purpose, from the vehicle parameters of the vehicle data set, a first number of parameters are established as input data or input parameters (E_i) in a task-specific manner, and a second number of parameters, preferably a complementary number to the first number, are established as output data or output parameters (A_j), and a vehicle database (1) is used, in which respective values of the input parameters (E_i) and values of the output parameters (A_j) are stored, in the form of vehicle data sets, for a number of validated vehicles that is sufficient for the purpose of model formation, in order to identify a dependence of the output parameters (A_j) on the number of the input parameters (E_i) in the form of a number of mapping models (Mx) (system identification, model formation), and the value of the number of the output parameters (A_j) is determined as the parameter of the vehicle data set based on said number of mapping models (Mx) and predefined values of the input parameters (E_i).

Description

A method of determining a vehicle parameter of a vehicle record of a vehicle and using the vehicle parameter on a test bench

The invention relates to a method for determining the value of at least one vehicle parameter of a vehicle data record of a vehicle, a subcomponent or a subsystem of a vehicle, and the use of a determined vehicle parameter in a simulation of carrying out a test on a test bench.

During the development of a vehicle, a large number of partially redundant parameters are produced which describe the vehicle and all its subcomponents in more detail. Such parameters are generally referred to below as vehicle parameters. Such vehicle parameters are e.g. Dimensions, transmission type, power, consumption, air resistance values, suspension parameters, drive, brake system, etc. The vehicle parameters also include parameters that fundamentally describe the vehicle and are in themselves easily accessible. Such parameters are also often referred to as design parameters and include parameters such as the year of manufacture of the vehicle, a vehicle class, information on the geometry of the vehicle, a mass, details of the type of drive, information on the transmission, etc. Furthermore, metadata such as the brand of the vehicle and / or a specific vehicle model may be design parameters.

In addition, there are measurement data on the vehicle, which also represent vehicle parameters. If the measurement data are characteristic values, i. describing the behavior of the vehicle, one often also speaks of characteristic Kennwer th, known examples are e.g. the coasting curve, a standard consumption, the lap times at the Nürburgring, etc.

It should be noted that the vehicle parameters need not include parameters of the entire vehicle, but may include only parameters of a subcomponent or a subsystem of the vehicle, such as those described hereinabove. Parameters of the drive train, the internal combustion engine, the drive unit in hybrid vehicles, the brake system, etc. Nevertheless, it will be discussed here and without limitation of Fahrzeugpa parameters.

In the development of vehicles or vehicle components, such as drive trains, drive systems, drive units, etc., simulations are increasingly being used. The advantages of a simulation are obvious: simulations can be used to simulate various influences on the vehicle or the vehicle component without the need for a real vehicle or a real vehicle component. This saves expensive prototypes and complex tests on test benches as well as real road tests so that development costs and development time can be reduced. In particular, current trends in legislation in the direction of "real-driving emissions" require efforts in vehicle development, which underscore the advantages of simulation solutions. Especially in early stages of development, simulations have prevailed. Of course, such simulations require suitable simulation models for the vehicle or vehicle component. These are now in high quality IN ANY. Such simulation models include simulation parameters with which the simulation model is adapted to a specific characteristic of the vehicle or of a vehicle subsystem. Before a simulation, the simulation models must therefore be parametrised via the simulation parameters. In this case, simulation parameters may possibly be identical to vehicle parameters. In addition to the simulation model itself, the quality of the simulation is directly influenced by the quality of the simulation parameters and the other vehicle parameters required. The best simulation model provides poor simulation data if the simulation parameters of the simulation model or the required vehicle parameters were parameterized incorrectly or poorly.

In the development of a vehicle, a vehicle component and a subsystem of a vehicle, the entirety of the vehicle parameters and the simulation parameter are often referred to as a vehicle data record. The problem with this is that, of course, only a few parameters of the vehicle data record are known at the beginning of the development of a vehicle. In practice, the correct parameterization of the parameters of the vehicle data record, ie the allocation of specific values for the various parameters, but quite a problem. In particular, the following problem areas occur:

Especially in the early development stage, there is the problem that simulations Runaway leads, but many parameters of the vehicle data set, especially simulation parameters, later in the course of development, since the development Fahrzeugent, eg the construction of the vehicle body or the power plant, and the Validation (eg by simulation) of the properties of the vehicle, eg the fuel consumption, usually take place in parallel ("Simultaneous Engineering" or "Cross Enter Prize Engineering"). Therefore, simulation engineers often have to make appropriate estimates in order to parameterize missing parameters of the vehicle data set with reasonable values. The developers often have incomplete information about the vehicle or the vehicle subsystem, ie they are only aware of a choice of parameters. These are usually simple design parameters, such as vehicle body construction, vehicle dimensions, rated engine power, etc. More precise Fahrzeugparame ter, such as stiffness and damping of the suspension, etc., measurement data or characteristic characteristics are generally not available. These missing parameters of the vehicle data however, it is necessary to parameterize but also to check the plausibility and validation of the vehicle data record. Although the necessary parameters of the vehicle data record are generally available in later stages of development, they are often not available or only available to simulation engineers due to organizational constraints (eg, other departments, other companies).

Often it is also in practice that certain components of the vehicle, such as e.g. a gearbox, not developed by the vehicle manufacturer itself, but by third parties on behalf of the vehicle manufacturer. For the development, however, further information of the whole system (vehicle or vehicle subsystem) is necessary in order to develop the component (for example the definition of the gear stages) and to be able to validate the desired properties (for example the simulation of the acceleration behavior). The third-party developers usually have incomplete information.

If parameters of the vehicle record originate from another source (e.g., other department or other entity), there is often the problem that it can not be ascertained that these parameters are correct, i. are not arbitrary assumptions, and on the other come from the right vehicle. A review of sustained parameters of the vehicle data set is difficult and often impossible.

Another essential point is that often no measurement data of the vehicle is available with which a parameterization of the parameters of the vehicle data record can be validated. However, the validation of the development parameters is important because it can be used at any time to check whether a development step, for example a specific simulation with a simulation model, provides meaningful values with the entered parameters.

Last but not least, there is a further problem in that, although different parameters of the vehicle data record can be present for a vehicle, they must be mapped for the parameterization of the simulation model to specific simulation parameters which, for example, have no physical correspondence. Such a mapping is often necessary because the existing vehicle parameters do not match the simulation parameters required. A corresponding mapping rule is often unknown to the simulation engineer. On the other hand, it may be that such a picture is not so easy.

Currently, simulation engineers typically use default values to parameterize the vehicle record parameters that are unknown, such as textbooks such as the landing gear manual, textbooks, textbooks, or data from vehicles they have already simulated, or data from similar vehicles produced by the manufacturer Si mulationmodelle mitliefert. Of course, this procedure has the problem that the quality of the parameters of the vehicle data set depends strongly on whether at least the essential vehicle parameters and / or essential simulation parameters, ie the simulation parameters which have a significant influence on the simulation model, are close to the correct value. In addition, it must be taken into account that technical progress changes relationships and parameter values. For example, downsizing has increased the power-to-weight ratio of internal combustion engines; furthermore, the cw values of vehicles have steadily improved.

In this context, DE 100 46 742 A1 discloses a method for a vehicle drafting system in which modeling and calculation programs from different fields are combined. They all work on a common database. First, the geometry of the vehicle is configured. Afterwards, a stiffness calculation is carried out and subsequently vehicle parameters such as vehicle weight, center of gravity, axle load distribution, mass inertia and aerodynamic coefficients are determined on the basis of correlation relationships. In a vehicle performance calculation then the speed, acceleration and stiffness of the vehicle, as well as the consumption determined at predetermined driving cycles. In this way, a first driving data set can be determined from some of the specified geometry parameters. The method uses predefined or known correlation relationships. The method is therefore very limited applicable and can also determine only fixed parameters of the vehicle data set from other specified parameters of the vehicle data set (geometry). For a flexible determination of parameters of the vehicle data set, this method is therefore not suitable.

It is now an object of the present invention to provide a method which allows any vehicle parameters of the vehicle data record of a vehicle to be determined in a simple manner and in a sufficiently high quality for development in order, inter alia, to carry out a test with a vehicle simulation to be able to use at the test stand.

This object is achieved by the method mentioned above ge triggers that from the vehicle parameters of the vehicle data set task-dependent a first number of input parameters and a second, preferably complementary to the first number kom, number of output parameters are set and a vehicle database, in the for a in the sense of modeling sufficient number of validated vehicles each values of the input parameters and values of the output parameters are stored in the form of data sets (vehicle, subcomponent, or subsystem records), used to determine a dependence of the output parameters on the number of Input parameters in the form of a number of imaging models to identify - the so-called system identification or modeling - and on the basis of this number of Ab from education models and predetermined values of the input parameters, the value of the number of output parameters is determined as a parameter of the data set. Conveniently, the value of the number of output parameters with associated statistical parameters (eg confidence interval) is determined and used as the parameter of the data record.

Using data from a vehicle database, which has both vehicle, sub-component and subsystem data, so not yet known Parame ter can be determined and used for the development or simulation activities. Thus, e.g. Input parameters selected by a user are used to generate the - unknown to the user - output parameters and thus a complete data set. In this case, the first number of input parameters that was used for the identification of the imaging models can be partially or completely used - if less than the first number of input parameters is used, there may be some difficulties in tuning all output parameters.

During the processing, the vehicle database is also increasingly extended by the results of the method according to the invention, and the imaging models can thus be adapted and improved. Accordingly, the generation becomes more and more accurate as the amount of data increases. A realistic development, meaningful assessment of the current state of development and timely detection of errors and later problem areas is possible. The parameters determined by the method are, in particular, output parameters which are defined on the basis of the imaging models and the selected / predetermined input parameters. As statistical characteristics are e.g. the confidence interval, the coefficient of determination or the adjusted coefficient of determination or the model significance for the application.

Advantageously, in addition at least one dependence of an additional output parameter on a number of the input parameters is modeled in the form of a known calculation rule.

In a further variant of the invention, the ascertained data record-or the data record created on the basis of the determined parameters-is compared with a predefined data record and the comparison result is output, wherein preferably a plausibility check is made between the determined and the predetermined data record.

The output of the comparison of the data sets can be output in the form of deviations, a quality value or simply as a result and allows another technical assessment of development quality. Plausibility exists, for example, if values of the parameters match to a certain extent; The dimension can be chosen depending on the application - eg 20% for ADAS or 5% for consumption values.

Conveniently, the models used are polynomial models or neural networks, which are determined by regression and / or training. One part of the data sets for the modeling is determined, the other part for the determination of statistical parameters for determining the model quality (eg coefficient of determination or the other parameters mentioned above).

In order to prevent inappropriate parameters from being used and to adversely affect the development work, in a variant of the invention, the deviations between stored values and values calculated therefrom by means of imaging models are minimized by identifying and not taking into account incorrectly specified input parameters and / or implausible data sets. be used. Preferably, confidence levels are defined for the input parameters on the basis of the data records of the vehicle database, and new input parameters are identified as being false if their values lie outside the confidence ranges. For example, fine, medium and coarse ranges in the form of overshoots and undershoots in percent or absolute values can be defined. If the default values of the input parameters fall short of the smallest value in the database by a predetermined percentage and exceed the largest value (e.g., by 25%), then a wrongly predetermined parameter is present; the limits of acceptable parameters as well as the allowed deviations can be defined application-related.

In order to exclude as far as possible the influence of errors on the data sets, the data sets of the vehicle database are at least partially assigned at least one quality indicator. This can be implemented, for example, as a numerical value, a high indicator is assigned with a particularly accurate measurement - if e.g. a measuring accuracy +/- 5% can be ensured. High quality is then considered more in the formation of the image models and allows a more accurate overall result.

In a further variant of the invention, the values of the input parameters of a given data set of a subset of the validated data sets stored in the vehicle database are used to determine a data record determined using the above-described method. Depending on the task, only subsets of the database or of data records of the database can be used - for example, a restriction of the data records to year numbers, weight classes or equipment variants can take place. According to the invention, at least one vehicle parameter determined by the method can be used as a simulation parameter for parameterizing a simulation model for simulating a vehicle or a subsystem of the vehicle for carrying out a test with a test object and an associated load machine on a test bench, wherein the test attempt at least partially with the simulation ge controls. In this way, a simulation can be parameterized in a simple manner, without exact prior knowledge of specific simulation parameters. This also ensures that the unknown simulation parameters are plausibly parameterized. Thus, simulations can be carried out, which were not possible according to the prior art, because a parameterization - that is, configuration - was not possible in the early stages of development due to lack of data.

Advantageously, the performance of the test on the test bench provides a characteristic value of the test specimen, which is compared with a corresponding determined Fahrzeugparame ter to validate the simulation. This also allows an immediate validation of the determined simulation parameters and a concomitant feedback, whether the determined vehicle parameters can be trusted or not.

The subject invention will be explained in more detail with reference to Figures 1 to 7, the exemplary, schematic and non-limiting advantageous Ausgestal lines of the invention. It shows

1 shows a block diagram of a first variant of the method according to the invention, FIG. 2 shows a block diagram of a second variant of the method,

3 shows a block diagram of a third variant of the method,

4 shows an input mask for a user interface,

5 shows a block diagram of a further variant of the method,

6 shows a device for carrying out the method according to the invention, and FIG. 7 shows a test stand according to the invention for carrying out a test with a simulation.

The subject invention assumes that there is a vehicle database in which known parameters of the vehicle record are stored by different vehicles. These data may include different vehicle parameters and simulation parameters for each vehicle.

As already explained at the outset, the vehicle parameters may include design parameters, vehicle measurement data and characteristic characteristics. Design parameters are vehicle parameters that determine the vehicle (or just a subsystem or a vehicle) Subcomponent) and are in themselves simple, even public, accessible. Measurement data and characteristic parameters, such as parameters such as the suspension spring stiffness, a rolling resistance, a maximum transferable Rei fenkraft, the Ausrollkurve, a standard consumption, etc., are typically not public, or difficult, accessible.

In particular, the design year of the vehicle, the vehicle class, a geometry parameter, a mass parameter, a drive parameter, a drive mode parameter, a tire parameter and a transmission parameter are considered as design parameters. Further, metadata such as the brand of the vehicle and / or a particular vehicle model may also be a design parameter.

The vehicle class describes, for example, whether the vehicle is a small car, Mittelklas sewagen, upper middle-class cars, luxury class cars, luxury class cars, sports cars, SUVs, a SUV, a people carrier, etc., is. As geo metrieparameter information on the geometry of the vehicle are considered, so example, the length, the width, the wheelbase, the track width, or the vehicle overhang before ne. A mass parameter describes, for example, the empty mass of the vehicle. The drive parameter includes parameters that describe the drive of the vehicle, such as the maximum power, possibly at a certain speed, or the maximum torque, possibly at a certain speed. With the drive type parameter can be distinguished between's front-wheel drive, rear-wheel drive, four-wheel drive, hybrid drive, etc. The tire parameter describes the tire in more detail, i. the tire identification, e.g. 185/65 R 15, or the tire diameter. The transmission parameter is for example the number of gear stages or the transmission spread (ratio of largest ter to smallest translation) or others.

This vehicle database can be constructed once and can of course be continuously updated and extended with new or updated vehicle parameters. Data, in particular parameters of the vehicle data record, for the vehicle database can be obtained, for example, from accessible data sheets of a vehicle (such as the type certificate), from vehicle catalogs, from car magazines, from measurements or from calculations or simulations based on other known data. Likewise, data from previous developments can also be included in the vehicle database. It is not necessary for all vehicles in the vehicle database, all possible parameters are known and registered.

According to the invention, dependencies between a first group of a number will now be based on the parameters of the vehicle data record stored in the vehicle database i = 1, ..., n of parameters P, of the vehicle data set, referred to as input parameter E, and a second group of a number j = 1, ..., m of parameters P _j of Fahrzeugdatensat zes, referred to as output parameter A _j , examined. In this case, the dependence on several different output parameters is investigated for at least one input parameter and the dependency is mapped by an imaging model. Data sets for vehicles / vehicle subsystem / vehicle subcomponents in the sense of modeling a sufficient number of validated vehicles, vehicle subsystems and vehicle subcomponents are stored in the vehicle database or are continually updated. At least some of the data sets may be assigned a quality indicator if, for example, they come from validated sources or specially made measurements. A high value for the quality indicator may mean that the data set comes from a particularly accurate measurement. High-quality datasets can be incorporated more into the formation of image models.

It makes sense, of course, if the input parameters and the output parame- ters do not overlap. Preferably, the input parameters are complementary, so under different to the output parameters. If a parameter of the vehicle data set is at the same time an input parameter and an output parameter, then one could set up a direct relationship input parameter = output parameter, and the application of the method according to the invention would not really make sense in this case. An imaging model is a mathematical model that calculates one or more output parameters A _j from the at least one input parameter Ei. Such a mathematical model can be a polynomial model, a neural network, a linear model network, a mathematical function, etc., which is determined by regression and / or training. One part of the data sets can be used for modeling, other parts for the determination of statistical parameters for determining the model quality (eg determination dimension). The mathematical models usually describe dependencies between the at least one input parameter E, and the one or more output parameters A _j , which have no physical correspondence. Such an imaging model can certainly also simultaneously determine a plurality of output parameters A _j as a function of several (but at least one) input parameters E i.

The determination of polynomial models, or another mathematical function, for example, by means of a known regression analysis. There are many known methods that could be used. Generally speaking, a given polynomial will be in the general form P (x) =

+ a _n x ⁿ , n> 0 changes to an error between i-0

see known data and to minimize the calculation result of the polynomial. As a rule, the coefficients a, are varied. As an error, for example, the mean square error can be used. It may also be provided that, as part of a refinement of the polynomial model, model terms are removed which are not statistically significant. If appropriate, a transformation of the output quantity (for example inversion or logarithmization) can also be carried out beforehand in order to better depict the dependencies.

When using neural networks as an imaging model, the neural network is appropriately taught based on existing data (input parameters, output parameters) in the vehicle database. There is also a plethora of well-known structures of the neural network and methods for training a neural network, which are not discussed here. By way of example, but not limitation, backpropagation algorithm, simulated annealing, delta rule, etc. are mentioned here. When determining an imaging model in the form of a mathematical model, it is advantageous if a part of the available data from the vehicle database is not used for modeling , These unused data can then be used to validate the identified imaging model. If records that were not used for the model computation have a similar error to other records computed using the mathematical model, it can be assumed that the mathematical model describes the data with sufficient accuracy. Thus, it can also be checked whether there is an overfitting of the imaging model.

Output parameters can be determined with associated statistical parameters (for example, confidence range) and the associated statistical parameters can additionally be determined as parameters of the data set and also stored.

For example, in a known manner, the confidence interval of the imaging model, more precisely the confidence interval of an output parameter of the imaging model, can be calculated. The confidence interval is a well-known term from statistics that can be calculated using known methods of statistics. The confidence interval gives an indication of how much an output parameter determined by the imaging model can be trusted. This means that the confidence interval, or another statistical parameter, provides information about the reliability of the model. The confidence interval refers to the mean value of the distribution of the y values for any x value in the examined area. The confidence interval of a model determines the uncertainty of the output variable yi in the measured points xi. For the confidence interval:

with the mean value y, the number of measuring points n, the variance s ² and z (1-a / 2) as (1- a / 2) -quantile eg the normal distribution.

For the variance applies:

with the above variables.

In general, the narrower the associated confidence interval, the more trustworthy an output parameter is. Further determinable statistical characteristics are e.g. Determinance measure, adjusted coefficient of determination or model significance.

With the help of the coefficient of determination r2 or of the adjusted coefficient of determination r2adj, as another possible statistical parameter, the quality of the determined imaging models is assessed, for example. The coefficient of determination r2 indicates to what degree the mapping model explains the deviations of the measured values from a constant mean value. It is calculated according to the relation r2 = SSreg / SStot = 1 - SSerr / SStot, where SSreg is the sum of the squared deviations between the predicted values (calculated by regression) yi * and the mean value y, SStot the sum of the quadratic deviations between the measured values yi and the mean value y and SSerr represents the sum of the squared deviations between the individual measured values yi and the calculated or predicted values yi *.

SSerr is thus a measure of the variation of y = P (x), which is not mapped by the regression equations. With the adjusted coefficient of determination r2adj, besides the model calculation quality, the given degrees of freedom of the model equation can be assessed, ie whether the measured values are close to the calculated model and whether sufficient measured values are available in order to be able to determine the model equation. It is calculated according to the context r2adj = 1 - [(SSerr / (n-p)] / [SStot / (n-1)] where n is the number of measurements and p is the number of independent model regression coefficients.

The used imaging models can be redetermined or refined at any time, for example, when new data is added to the vehicle database.

In addition to the mapping model, a dependency between the number of input parameters Ei and an output parameter A _{j may be given} by a physical or functional relationship, generally referred to as a calculation rule. A calculation rule in the form of a physical relationship could be, for example, the calculation of individual gear ratios from the number of gear stages and the gear ratio. A functional relationship can be set up, for example, for the Cw value as a function of the vehicle length and the type of vehicle. Such functional relationships can also depict relationships that can not be calculated physically without further ado, or that have no direct physical correspondence, but in which, due to the wide variety of constraints, there are nevertheless relationships between input parameter E, and output parameter A _j .

A calculation rule is thus a known mathematical function which calculates an output parameter A _j as a function of a number of input parameters E i.

The step of determining the at least explaining an image model for at least one output parameter A _j, preferably a plurality of imaging models for various gait parameters from A _j is based on the Fig.1. In a vehicle database 1, various parameters of the vehicle data set are stored for k different vehicles F1, F2,..., Fk. In this case, a vehicle data record comprises z parameters (vehicle parameters and / or simulation parameters). However, it should be noted that values for all parameters P _{kZ of} the vehicle data record do not always have to be stored in the vehicle database 1 for each vehicle Fk.

From the parameters P _{kZ of} the vehicle data set, the i input parameters E, and the j output parameters A _{j are} selected. As input parameter E, typically simply available or known vehicle parameters are used, that is to say in particular the design parameters described above. The output parameters A _j are those parameters P _{kZ of} the vehicle _{data record} which are to be determined (in the sense of being predicted).

From the available data of the vehicle database 1, in a model calculation unit 2 for at least one output parameter A _j, the mapping model M _{x is determined} as described above. determined, which maps the input parameter E, the output parameter A _j . For this purpose, of course, only such vehicle data sets from the vehicle database 1 are of course used, to which the selected input parameters E, and from input parameter A _j are present, since only then appropriate dependencies can be determined. Of course, several imaging models Mx for several Ausgangssparame ter A _j can be determined. It is also conceivable that several output parameters A _j are simultaneously calculated with an imaging model Mx.

It should be noted, however, that not all imaging models Mx must contain the same input parameter Ei. For this purpose, preferably a part of the vehicle data sets contained in the vehicle database 1 of the vehicles Fk (training data) is used for the determination (training) of the imaging models Mx. The rest of the vehicle data sets (validation data) contained in the vehicle database 1 can then be used for the validation of the imaging models Mx. For this purpose, the associated output parameters A _{j are} calculated using the specific mapping models Mx with the input parameters E, the validation data, and the validation error between the calculated output parameters A _j and the output parameters stored in the validation data is determined. If the validation error is within a specified tolerance range, then the mapping models Mx are validated. In addition, certain output parameters A _{j can} also be calculated on the basis of known calculation rules as a function of the input parameters E i. For this purpose, NEN known calculation rules for images can be stored in the model calculation unit 2. These direct mapping rules can be taken directly into the Mx imaging model. However, it can also be provided that a known calculation rule is also input via a user interface by a user.

In this process, an additional verification step can be provided, as will be described with reference to FIG. Therein, the selection of the input parameters E, (wherein the input parameters E, can also be predetermined) and output parameters A _j at a suitable user interface 4 a is shown. For example, a user can easily select the desired parameters of the vehicle data record on an input mask via the user interface 4a. Furthermore, provision can be made for a user to be able to specify previously known calculation rules for modeling the dependence of the output parameters A _j on the input parameters via the user interface 4 a. After the determination of the image models Mx in the model calculation unit 2 to another user interface 4b, the same user interface may be 4 as the selection of the input parameters E, and output parameters A _j that he results of the validation described above and / or a confidence region of the mapping models Mx. If an imaging model Mx could not be validated, or if there is too much confidence, then it may be provided that other or additional input parameters E, must be selected and the step of determining the imaging models Mx must be repeated, as shown by the arrow back , This step can also be automated if a tolerance range for a validation error and / or a confidence interval are specified. Otherwise, the determined imaging model Mx can be used.

Ideally, all parameters of the vehicle data record which are not input parameters Ei are regarded as output parameters A _j and, in each case, an imaging model Mx is determined or a calculation rule specified. This can be determined by specifying certain values of the input parameters E, the remaining parameters of the vehicle data record on the basis of the mapping models Mx or the additional calculation instructions. In this way, therefore, a whole vehicle data set, but at least a part thereof, from a few input parameters E, are created. In the determination of the input parameters, a first number of input parameters are preferably selected from and, for the later determination of complete data records, this first number of input parameters is also partially or completely predetermined.

With the mapping models Mx determined in this way, the unknown values of the output parameters A _j linked via the mapping models Mx and optionally via the calculation rules can now be used, based on predetermined input parameters E, as described with reference to FIG. For this purpose, the values of the input parameters Ei at a user interface 4c (possibly the user interface 4a or 4b from above) can be input via an input mask 3, as shown by way of example in FIG. 4, or also loaded automatically from a present data file. In the exemplary embodiment described, the input mask 3 includes as vehicle parameter the year of construction P_BY of the vehicle, the vehicle class P_VC, a parameter group of geometry parameters P_G of the vehicle, a parameter group of a mass parameter P_M of the vehicle, a parameter group of drive parameters P_D of the vehicle, a parameter group of Drive type parameters P_DT of the vehicle, a parameter group of tire parameters P_T of the vehicle, a parameter group of Getriebepa parameters P_GB of the vehicle, a vehicle brand P_B, a vehicle model P_Mo and a steering ratio P_L. The input mask 3 comprises all possible input data. parameters E ,, which are logically grouped in the named parameter groups. Each parameter group may contain a plurality of vehicle parameters. The input parameters Ei can be input via this input mask 3 as shown in FIG. 4, whereby not all input fields of the input mask 3 have to be filled. This means that not all values of input parameter E, must be set.

By means of the imaging models Mx, at least one imaging model Mx, the associated output parameters A _j are then calculated. The output parameters A _j may include all common vehicle parameters such as design parameters KP, simulation parameters SP and measurement data or characteristic values MP, this classification being irrelevant to the invention. It is also possible to calculate further of the parameters described above. The output parameters A _j can then also be transferred to the vehicle data record in the vehicle database 1.

Thus, based on a few input parameters E, as default values, a whole vehicle data set, at least a part thereof, can be determined with which the development of the vehicle or a part thereof can then be started or continued. On the basis of the possibility of validation and / or the confidence levels of the Abbildmo models Mx can be assumed from a sufficient accuracy of the parameters of the Fahrzeugda tensatzes. For this purpose, it is also possible to repeat the step of determining the mapping models Mx at any time, if you get in the course of development be certain parameters of the vehicle data set or if it changes parameters of the vehicle data set. Thus, the accuracy of the estimation of the unknown output parameters A _j can be continuously improved.

With Figure 7 is a known test stand 10 for a test specimen 12 and thus, for example, by means of a test stand shaft 11, connected loading machine 15, for example, a dynamometer represented. The loading machine 15 generates the load for the test piece 12 for the test run to be performed. The specimen 12 is in the example shown Ausfüh an internal combustion engine and the test bench 10 so that an engine test.

Of course, the test specimen 12 could also be an entire vehicle or a belie biges subsystem of the vehicle, such as a drive train, an electric motor, to a drive battery, a control unit, etc., and the test stand 10 was a matching test, such as a dynamometer, a powertrain test stand, a Elektromotoren test stood, a hardware-in-the-loop test stand, etc. In the case of a battery as a test specimen 12, the loading machine 15 would be electrical, for example in the form of an electric battery tester. Ge suitable loading machines 15 for various specimens 12 are well known and available, which is why it need not be discussed in detail here. On the test bench 10, a test bed automation in the form of a 20 Prüfstandautomatisie unit (hardware and software) is provided which controls the test bench 10 durchzufüh-generating virtual test drive (= test run) and all the necessary facilities (ie in particular the required actuator) of the test bed 10 according to Defines the specifications of the test run. In particular, the test bed automation unit 20 can also control the test object 12 and the loading machine 15 by presetting required setpoint values or manipulated variables. The loading machine 15 is often on the test bench 10 of its own load machine controller 14, in turn, receives from the test bench automation unit 20 according to the specifications of the test run setpoints, to the DUT 12, for example certain, often transient, load moments M or certain, often transient, speeds to regulate. The loading machine controller 14 may also be integrated in the test bench automation unit 20 as software and / or hardware.

To carry out the test run, here for example, a speed measuring device 16 and / or a Momentenmesseinrich- the corresponding actual values of the test object 12 and / or the load are on the test bed 10 usually measuring devices 13, processing 17 is provided, machine 15, _st, for example, the load torque Mi and Speed _{n.st of} the test piece 12, measure as measured variables and make the test bed automation unit 20 available. Of course, for other specimens 12, or test stand types, ande re or additional measurements, such as an electric current or an electrical cal voltage, measured and the test bed automation unit 20 are supplied.

To carry out the test run, a simulation 6 of a vehicle or a part or a component thereof is provided, the operation of which is simulated according to the specifications of the test run. For this purpose, a simulation model is provided. The simulation 6 simulates, for example, a virtual drive of a virtual vehicle along a virtual route. For this purpose, the component of the vehicle constructed on the test stand 10 (for example the internal combustion engine as the test object 12) is actually operated on the test stand. The simulation 6 also receives feedback (for example in the form of specific measured values) for carrying out the simulation 6 from this real operation. The simulation 6 with the implemented simulation model is executed by a simulation unit 21 and can also process measured variables from the test stand 10 for this purpose. The simulation unit 21 may be integrated in the test bench automation unit 20 as hardware and / or software, but may also be separate from the test bed automation unit 20, for example in the form of its own simulation hardware and simulation software. The test run is specified by a test run unit 22. The test run is, for example, a time course of certain variables, such as vehicle speed, slope of a route, curve, etc., and can be specified for example from an external source. The test unit 22 can also with Interfaces to controls of a vehicle (accelerator pedal, steering wheel, brake pedal, etc.) to be executed so that the test run, or a part thereof, can be specified directly on the test bench 10 by Be dienerführung. With the specification of the test result results in the interaction of the system of Simulation 6, 10 test, 10 and 12 Prüfemaschi ne the test run, which is performed on the test bench 10.

During the implementation of the test 10 measurements are often carried out at the test bench 10 with Messein directions and measurements on the test specimen 12 to make certain statements about the behavior of the specimen 12 can. Typical and frequently provided measurements include the emission behavior of an internal combustion engine, the consumption or power requirement of the test object, the generated power of the test object, etc. Such measurements also provide, for example, measured data or characteristic parameters as vehicle parameters, either directly or through appropriate processing of the test object measurement data.

In the example shown in FIG. 7, the test run defines the time-based predefinition of the lane of a vehicle in the form of the gradient RG and of the speed curve v. This test run is given to the simulation unit 21, in which a simulation model, for example, a model of a vehicle, which is moved along a route, implemen ted, which performs the simulation 6. On the basis of the current engine speed n, _st and the current torque Mi _{st of} the test specimen 12 and the specifications of the test run the simulation unit 21 calculates the accelerator pedal position a _so n as a manipulated variable for the engine as the DUT 12, and the target torque M _so n as the setpoint for the The setting or conversion of the manipulated variable and the setpoint values on the test bench 10 leads to a specific state of the test object 12. Depending on the test bench 10 and the test object 12 and simulation 6, other predefined variables, of course, can also be used. Actuating variables and measured variables are used. In the illustrated example, an emission measuring device 18 is provided as measuring device 18 to sen during the execution of the test run emission quantities in the exhaust gas of the internal combustion engine.

For carrying out the simulation 6 for the test run on the test bench 10, it is necessary to parameterize the simulation model with simulation parameters SP. For this purpose, at least one output parameter A _j calculated via an imaging model Mx as described above can be used as the simulation parameter SP in order to condition a parameter of the simulation model, as shown in FIG.

If characteristic values MP are also determined as output parameters A _j , these can optionally also be used to validate a result of a simulation 6, as also indicated in FIG. 3. The simulation model 23 was used for simulation 6 parameterized with determined simulation parameters SP. When carrying out the test on the test bench 10 can result in the interaction of the system from Simulation 6, 10 test with test specimen 12 and loading machine 15 as a result, a characteristic value MPM. This can for example be measured directly or can be derived from other results or measurements of the test. Such a characteristic value MPM, which was determined during the test, can be compared with a characteristic parameter MP of the output parameters Aj, for example a standard consumption or an emission quantity, which was calculated using an imaging model Mx, for validation in a comparison unit 19. The comparison unit 19 (hardware and / or software) may be integrated in the test bed automation unit 20, or may be separate therefrom. If the deviation is within an acceptable range specified for the application, the simulation parameters SP determined using the imaging models Mx can be considered as valid.

An output parameter Aj and / or the result of the simulation 6, possibly together with the result of the validation, e.g. the deviation between the measured characteristic value MPM and the predicted characteristic characteristic value MP can then be output in a suitable manner as user feedback to a suitable output unit 7. Likewise, it is possible to store an output parameter Aj in the vehicle database 1.

With Fig.5 yet another application of the invention will be described. Here, the input parameters E are presented to a user interface 4c (eg as shown in FIG. 4). The output parameters Aj, preferably all other unknown parameters of the vehicle data set, are again calculated on the basis of the mapping models Mx. This vehicle data set determined in this way, from the input parameters E and the output parameters Aj, can be compared with a predefined vehicle data record 8 to validate it. The result of the validation, for example in the form of deviations of the individual output parameters A _j in the two vehicle data sets, can then be output in a suitable manner to a suitable output unit 7 as user feedback. In this way, the plausibility of an existing vehicle data record can be easily and quickly checked.

In other words, therefore, the deviations between stored values and values calculated therefrom by means of imaging models Mx can be minimized by identifying incorrectly specified input parameters E, and / or implausible data records for vehicles, subcomponents and subsystems of a vehicle and subsequently not taking them into consideration. For this purpose, for example, confidence ranges for input parameters Ei can be defined, so that parameters can be identified as being false if their values lie outside these confidence ranges. If, for example, the values of the input parameter Ei fall below the smallest value in the database by a specified percentage or exceed the largest value, then a wrong or incorrectly predefined parameter is present. The permissible undershoot or overshoot can be defined in fine, medium and coarse ranges.

This opens up the following application possibilities, for example: a) A vehicle manufacturer wants to develop a new vehicle model. Due to simple basic information in the form of input parameters Ei, such as length, width height of the vehicle, vehicle class, by means of the mapping models Mx, based on vehicle data of existing vehicle models of the vehicle manufacturer he was witnessed, already a first vehicle data set for the new vehicle model are created , With this vehicle data set, the development, for example, based on simulations can be started. With the mapping models Mx, the behavior of the new vehicle model can also be predicted. b) From freely searchable data of a particular vehicle (for example data from manuals and publications), ie from input parameters Ei, other parameters of the vehicle data set (output parameters Aj) can be automatically generated based on existing or determined mapping models Mx. c) If a simulation engineer constructs a vehicle data set for a simulation, other parameters of the vehicle data set (output parameter Aj) can be averaged based on known design parameters (input parameters E,) and mapping models Mx, with which a simple plausibility check of the constructed vehicle data record for the vehicle Simulation can be performed. d) Engineers at test benches receive a vehicle data record from the simulation department. With the possibility of validating the vehicle data record, you can easily and quickly check whether the vehicle data record is valid. It can be prevented that is wasted by defective vehicle records development time.

An exemplary implementation of the method is shown in FIG. 6:

For the realization of the method, at least one data-keeping unit 100 and a test-bench automation unit 20 are necessary. The test bed automation unit 20 has, for example, an application unit 200, a test run unit 22 and at least one simulation unit 21. Furthermore, a user interface 4 and an output unit 7 can be provided in the form of an input and output device, for example for parameter setting. ration, evaluation and presentation of output parameters Aj and / or results of the test or parameter determination. The individual units can communicate with one another via a suitable data interface 11 (eg shared memory, TCP, CAN, etc.), wherein different units can also communicate via different interfaces. Each of the units may further comprise at least one suitable user interface and local memories.

The data storage unit 100 has the at least one vehicle database 1, T with the data records for vehicles, subcomponents and subsystems of the vehicles. The application unit 200 includes a model calculation unit 2 for determining the imaging models Mx, a parameter calculation unit 9 for predicting non-existent output parameters Aj, as well as possibly also a data verification unit 10 for validating vehicle data sets, in particular foreign or external predefined driving data records. The simulation unit 21 has simulation models 23 and a calculation unit 24 for executing the simulation models 23, which can also take place in real time. For example, a simulation model 23 includes at least one or more of the following models: vehicle model, driver model, road or distance model, wheel model, environment model. Thus, a virtual journey of a virtual vehicle with a virtual driver along a virtual route can be simulated. In the data holding unit 100 and / or the test bed automation unit 20, or subcomponents thereof, such as the application unit 200 and / or simulation unit 21, also input and output devices 4, 7 may be provided. The different units can be on different hardware or, at least partially, out on the same hardware leads.

Claims

claims

A method for determining the value of at least one vehicle parameter of a vehicle data record of a vehicle, a subcomponent or a subsystem of a vehicle, wherein a first number of input parameters (E,) and a second, preferably complementary to the first number, of the vehicle parameters of the vehicle data record, Number of output parameters (Aj) and a vehicle database (1) in which, for a model-forming number of validated vehicles, respectively values of the input parameters (E,) and values of the output parameters (Aj) in the form of known vehicle data sets (Aj) Vehicle, subcomponent, or subsystem datasets) is used to identify a dependency of the output parameters (Aj) on the number of input parameters (E,) in the form of a number of mapping models (Mx) (system identification, model formation ), and on the basis of this number of imaging models (Mx) and of predefined values of the input parameters (E,), the value of the number of output parameters (Aj) is determined as a vehicle parameter of the vehicle data record.

2. The method according to claim 1, characterized in that in addition at least one dependence of an additional output parameter (Aj) of a number of input parameters A (Ei) is modeled in the form of a known calculation rule.

3. The method of claim 1 or 2, characterized in that at least one output parameter (Aj) is a simulation parameter, with the implementation of a Simu lation a simulation model of the vehicle, a vehicle component, a Teilkompo component or a subsystem of a vehicle is parameterized.

4. The method according to any one of claims 1 to 3, characterized in that the determined vehicle data set is compared with a predetermined vehicle data set and the comparison result is output, preferably a plausibility check between the determined and the predetermined vehicle data set is made.

5. The method according to any one of claims 1 to 4, characterized in that as imaging models (Mx) polynomial models or neural networks are used, which are determined by regression and / or training.

6. Method according to one of claims 1 to 5, characterized in that the deviations between stored values and values calculated therefrom by means of imaging models (Mx) are minimized by incorrectly predefined input parameters (E,) and / or implausible vehicle records identified and not taken into account / used.

7. The method according to claim 6, characterized in that for the Eingangsparame ter (Ei) based on the vehicle data sets of the vehicle database (1) confidence ranges defi ned and new input parameters (E,) are identified as wrong if their who te outside the confidence ranges ,

8. The method according to any one of claims 1 to 7, characterized in that the vehicle data sets of the vehicle database (1) at least partially at least one Quali tätsindikator assigned.

9. The method according to any one of claims 1 to 8, characterized in that the values of the input parameters (E,) of a predetermined vehicle data set of a subset of the validated th vehicle data sets stored in the vehicle database is used to determine the vehicle data set.

10. Use of at least one vehicle parameter determined by the method according to claims 1 to 9 as simulation parameter (SP) for parameterizing a simulation model (23) for simulating a vehicle or a subsystem of the vehicle in a simulation (6) for carrying out a test with a test piece (12) and an associated loading machine (15) on a test stand (10), wherein the test is at least partially controlled by the simulation (6).

1 1. Use according to claim 10, characterized in that the implementation of the test on the test bench (10) provides a characteristic characteristic value (MPM) of the specimen (12) and at least one with a mapping model (Mx) determined vehicle parameters (MP) of the corresponding characteristic Characteristic value and the two characteristic parameters are compared in order to validate the simulation (6).

12. A test stand for performing a test on a test specimen (12) which is connected to a loading machine (15), wherein a test bench automation unit (20) is provided, the test specimen (12) and the loading machine (15) for performing the test after Furthermore, a simulation unit (21) for simulating (6) a vehicle or a subsystem of the vehicle with a simulation model (23) is provided and the test bed automation unit (20) for carrying out the test is a result of the simulation (FIG. 6) at least partially used for the control of the test piece (12) and / or the loading machine (15), and where in at least one simulation parameter (SP) of the simulation model (23) is a determined according to the entitlements 1 to 9 vehicle parameters.