DE102021002318A1 - Method for creating a simulation model, use of a simulation model, computer program product, method for calibrating a control device - Google Patents

Abstract

Method for creating a simulation model for a calibration of a control device, wherein the control device controls at least one functional process of a component of a drive train and wherein the simulation model maps the at least one functional process of the component, use of a simulation model in a control device, computer program product and method for calibrating a control device using at least one simulation model depicting a functional process of a component of a drive train.

Description

The invention relates to a method for creating a simulation model for a calibration of a control device, wherein the control device controls at least one functional process of a component of a drive train and wherein the simulation model maps the at least one functional process of the component, a use of a simulation model in a control device, a computer program product and a method for calibrating a control device using at least one simulation model depicting a functional process of a component of a drive train.

When developing drive train components, test bench tests are carried out with real components. The operation of test stands is associated with a high expenditure of time and corresponding operating costs. The number and duration of test bench investigations are reduced with the help of simulation models, which allow a mathematical description of the real functional process of the drive train components. Such simulation models calculate, for example, a relationship between input values and output values, so that measurement results for at least parts of the functional process of the real drive train components can be computationally determined by mapping or simulation.

The pamphlet WO 2016/198047 A1 discloses a method for creating a simulation model for mapping at least one functional process of a component of a drive train, with an input transformation function for the input values of the simulation model, an output transformation function for the output values of the simulation model, a basic model of a data-based regression method as the basis for the simulation model can be specified. Optimized simulation models are generated in four optimization runs, at least one of which is selected.

One task can be to improve a method for creating a simulation model for mapping a functional process.

• S1) Festlegen von initialen Modellparametern des Simulationsmodells,
• S2) Bestimmen einer Plausibilitätsverteilung der Modellparameter des Simulationsmodells,
• S3) Berechnen eines Gütekriteriums des Simulationsmodells auf Basis der Modellparameter,
• S4) Bestimmen einer Korrektur für das Gütekriterium anhand der Modellparameter und der Plausibilitätsverteilung,
• S5) Berechnen eines korrigierten Gütekriteriums anhand der Korrektur,
• S6) Verändern der Modellparameter, um das korrigierte Gütekriterium zu verbessern,

wobei die Schritte S3, S4, S5 und S6 wiederholt werden.One aspect relates to a method for creating a simulation model for a calibration of a control device, wherein the control device controls at least one functional process of a component of a drive train and wherein the simulation model maps the at least one functional process of the component. The simulation model depicting the at least one functional process is created on the basis of training data, with the following steps:

• S1 ) Definition of initial model parameters of the simulation model,
• S2 ) Determining a plausibility distribution of the model parameters of the simulation model,
• S3 ) Calculation of a quality criterion for the simulation model based on the model parameters,
• S4 ) Determining a correction for the quality criterion based on the model parameters and the plausibility distribution,
• S5 ) Calculating a corrected quality criterion based on the correction,
• S6 ) Changing the model parameters in order to improve the corrected quality criterion,

taking the steps S3 , S4 , S5 and S6 be repeated.

When calibrating a control device, the simulation model can, for example, advantageously serve to replace measurements on the components of the drive train on a test bench with virtual measurements on the simulation model. The virtual measurements provide data that can be used to calibrate the control unit. For example, a large number of virtual measurements can be carried out in order to determine favorable input values in an optimization process, which lead to an improvement in terms of the output values and thus certain properties of the components of the drive train. This improvement in the output values can be represented in a calibration criterion which, for example, also combines the output values of different simulation models with the same input variables but different output variables. This calibration criterion is then improved or fulfilled by varying the input values of the input variables of the simulation model. The optimal input values of the simulation model determined in this way are then transferred to corresponding manipulated variables of a function of the control device, for example an engine controller, by means of transfer software. These manipulated variables are usually stored in the form of maps, characteristic curves or characteristic values, so that the optimal input values of the simulation model are transmitted in this case by data input into characteristic maps, characteristic curves or characteristic values in the engine control.

• S1) Festlegen von initialen Modellparametern des Simulationsmodells,
• S2) Bestimmen einer Plausibilitätsverteilung der Modellparameter des Simulationsmodells,
• S3) Berechnen eines Gütekriteriums des Simulationsmodells auf Basis der Modellparameter,
• S4) Bestimmen einer Korrektur für das Gütekriterium anhand der Modellparameter und der Plausibilitätsverteilung,
• S5) Berechnen eines korrigierten Gütekriteriums anhand der Korrektur,
• S6) Verändern der Modellparameter, um das korrigierte Gütekriterium zu verbessern,

wobei die Schritte S3, S4, S5 und S6 wiederholt werden.Another aspect relates to the use of a simulation model in a control device, the simulation model depicting at least one functional process of a component of a drive train, and the control device controls the at least one functional process, the simulation model is created on the basis of training data, with the following steps:

• S1 ) Definition of initial model parameters of the simulation model,
• S2 ) Determining a plausibility distribution of the model parameters of the simulation model,
• S3 ) Calculation of a quality criterion for the simulation model based on the model parameters,
• S4 ) Determining a correction for the quality criterion based on the model parameters and the plausibility distribution,
• S5 ) Calculating a corrected quality criterion based on the correction,
• S6 ) Changing the model parameters in order to improve the corrected quality criterion,

taking the steps S3 , S4 , S5 and S6 be repeated.

When used in a control device, the simulation model itself is advantageously used to control the functional process. The following statements relate both to the method for calibrating the control unit and to the use of the simulation model in the control unit.

Another aspect relates to a method for calibrating a control device using at least one simulation model depicting a functional process of a component of a drive train, the at least one simulation model being created according to the disclosed method for creating a simulation model.

According to one embodiment of the method for calibrating a control device, it is provided that at least one calibration criterion is defined on the basis of an output variable of the at least one simulation model and the simulation model is used to obtain output values by entering various input values, an optimized input value being sought for which the output value fulfills at least one calibration criterion. The optimized input value of at least one input variable of one of the simulation models is stored, for example, in a characteristic map, a characteristic curve or a characteristic value of the control device.

The features and embodiments described below relate equally to all aspects described above, the steps S3 , S4 , S5 and S6 can be repeated iteratively until a termination criterion is reached, wherein the termination criterion can be, for example, a number of iterations, reaching a specified value for the quality criterion or a combination of termination criteria. A person skilled in the art is able to select a sensible termination criterion based on the application. The model parameters of the simulation model are determined by training the simulation model, the simulation model being trained using the training data with regard to an improvement in the quality criterion. A plausibility distribution of the model parameters of the simulation model is determined and the model parameters of the simulation model are advantageously checked for plausibility using the plausibility distribution by correcting the quality criterion for training the simulation model using the plausibility distribution. This advantageously ensures that, for example, very large or very small values for certain model parameters are avoided in order to avoid implausible behavior of the simulation model, whereas very large or very small values are preferred for certain other model parameters and thus irrelevant input variables are suppressed, as a result of which the simulation model is advantageously simplified.

According to a particular embodiment it is provided that the plausibility distribution in step S2 is estimated on the basis of a comparative distribution, the comparative distribution of known predecessor model parameters being created from a plurality of predecessor simulation models created beforehand for mapping an at least comparable functional process. For example, a known plausibility distribution of predecessor model parameters of one or more of the predecessor simulation models created previously can be used as a comparison distribution for mapping the at least comparable functional process. Alternatively or additionally, repetitive measurements can be carried out on the component of the drive train with unchanged values of the manipulated variables corresponding to the input variables of the simulation model and a distribution of data of a measured variable corresponding to an output variable of the simulation model can be analyzed. An analysis of a distribution of data of input variables and / or output variables of the training data on which the simulation model is based is also conceivable.

According to a further particular embodiment, it is provided that the plausibility distribution in step S2 is adapted depending on a number of training data by shifting and / or scaling in such a way that, with regard to the number of training data, more relevant model parameters are assigned a higher plausibility. This advantageously means that, depending on the number of training data, different requirements are placed on the characteristics of the Plausibility distribution can be realized, for example that with a high number of training data, smaller values of the model parameters are assigned a higher plausibility in order to obtain an overall more plausible behavior of the simulation model.

According to a further particular embodiment, it is provided that the plausibility distribution in step S2 is created on the basis of empirical values, expectations and / or physical relationships.

According to a further particular embodiment, it is provided that the plausibility distribution in step S2 is described using at least one of the model parameters on the basis of at least one sigmoid function. In particular, the plausibility distribution of the at least one of the model parameters is described as a product of sigmoid functions.

According to a further particular embodiment, it is provided that the simulation model is created in an online process, with the component of the drive train being operated on a test bench and with measured values measured on the component being continuously used as training data.

According to a further particular embodiment, it is provided that non-parametric models are used as simulation models. In particular, nonparametric models of a Gaussian process model type, a manifold Gaussian process model type, a warped Gaussian process model type or a manifold-warped Gaussian process model type are used as simulation models.

According to a further particular embodiment, it is provided that a model size of the simulation model is reduced in order to enable a higher execution speed, the reduction of the model size being carried out in particular on the basis of plausibility-checked model parameters.

Since measurements on the test bench are time-consuming and expensive, it is advantageous not to test all possible combinations of manipulated variable variations, but rather to carry out selected measurements. The simulation model is, for example, a data-driven model with the manipulated variables as input variables of the model. The simulation model created is used, for example, to determine the manipulated variables of the component in an optimization process in such a way that the output variable of the simulation model, for example the fuel consumption, assumes a value that is as favorable as possible. Since no real measurements have to be carried out during optimization, but only the simulation model is evaluated, expensive measurements on test stands can be saved.

In the calibration method, a distinction is made between embodiments as an offline method and an online method. Measurement data are generated for the creation of the simulation model. For this purpose, a test plan is usually created with a list of measuring points to be measured, which are then measured. With the offline method, this takes place in two separate steps, one after the other, i.e. first the measurements are carried out on the test bench, then the simulation model is created without the test bench, in particular with the help of a computer. With the online method, a simulation model is created after each measured point or after a certain number of measurements on the test stand, which is preferably used to determine the next points to be measured on the test stand in a more targeted manner.

A drift correction is to be understood as a correction of a drift. Drift describes a systematic error in measurement data that increases over time. The drift correction is used to compensate for such a drift in data, in particular in training data, the drift in particular being removed from the data. An example of drift is the systematically increasing deviation in emission measurement data due to the constant sooting of a measuring device.

In the context of the invention, model parameters are understood to mean so-called hyperparameters. The shape of a Gaussian process model is determined by its covariance function and mean value function. For manifold and warped Gaussian process models, the shape is also determined by the transformation functions. These functions are parameterized by the hyperparameters. With given functions of the respective Gaussian process model, the course can be influenced by the choice of the values for the hyperparameters. When training the Gaussian process models, suitable values are determined for the hyperparameters, so that a specific quality criterion is improved.

Training data are an essential part of a Gaussian process model. Training the Gaussian process model means adapting the hyperparameters in such a way that they match the training data. This is expressed by the quality criterion likelihood. After adapting or optimizing the hyperparameters, one speaks of a trained Gaussian process model.

The so-called likelihood, the marginal likelihood, the logarithmized likelihood or the logarithmized marginal likelihood is understood as a quality criterion for the purposes of the invention. When training Gaussian process models suitable values are determined for the hyperparameters, so that one of the quality criteria is improved. The quality criterion is corrected based on the plausibility distribution. Training can also be referred to as determining parameters. Concrete numerical values of abstract parameters are referred to as parameter values.

The plausibility distribution assigns a plausibility to a hyperparameter. In general, this can be called a prior. When it refers to hyperparameters it is called hyperprior. If the hyperprior is represented as a probability distribution, then this is referred to as the actual hyperprior. If this is not the case, then it is an improper hyperprior. In order to calculate the model evidence of a Gaussian process model, an actual hyperprior is required.

A class of data-driven models is referred to as a Gaussian process model. This includes different model variants. These model variants are differentiated with regard to various criteria: the model type, the model structure, the plausibility distribution of the hyperparameters, the preprocessing or adaptation of the training data, i.e. the transformations, the drift correction and the outlier removal.

A distinction is made between standard Gaussian process models, Manifold Gaussian process models, Warped Gaussian process models, Maniold-Warped Gaussian process models and other model types. The model structure is determined by the choice of the covariance and mean value functions. In the case of manifold and warped Gaussian process models, the model structure is also determined by the manifold or warped transformation function.

Historical data or models that already existed in time before the method according to the invention was applied are referred to as predecessor simulation models.

Another aspect relates to a computer program product comprising program codes, the program codes being stored on a computer-readable medium in order to carry out the disclosed method for creating a simulation model or to carry out the disclosed method for calibrating a control device when the program is executed in a computer.

Another aspect relates to a further computer program product comprising program codes, wherein the program codes are stored on a computer-readable medium and wherein the program codes contain the simulation model according to the disclosed use in order to use the simulation model for controlling the at least one functional process of the component of the drive train, if the Program is executed in the control unit.

• 1 ein Ausführungsbeispiel eines Verfahrens zur Erstellung eines Simulationsmodells in einem Flussdiagramm.

Embodiments of the invention are explained in more detail with reference to the accompanying figure. It shows

• 1 an embodiment of a method for creating a simulation model in a flowchart.

With reference to the 1 an exemplary embodiment of a method for creating a simulation model for a calibration of a control device, wherein the control device controls at least one functional process of a component of a drive train and wherein the simulation model maps the at least one functional process of the component, is initially explained in an application-independent manner using a flowchart.

To create the simulation model, follow the steps S1 ) Definition of initial model parameters of the simulation model, S2 ) Determining a plausibility distribution of the model parameters of the simulation model, S3 ) Calculation of a quality criterion for the simulation model based on the model parameters, S4 ) Determining a correction for the quality criterion based on the model parameters and the plausibility distribution, S5 ) Calculating a corrected quality criterion based on the correction and S6 ) Changing the model parameters in order to improve the corrected quality criterion, running through the steps S3 , S4 , S5 and S6 be repeated. The individual steps are explained in more detail below.

In a first step S1 initial values for the model parameters are set first. In a subsequent second step S2 the plausibility distribution, the so-called hyperprior, is then determined for each model parameter of the simulation model. This can be done, for example, by means of historical data from similar simulation models, in particular predecessor simulation models, in which it is shown that the model parameters are primarily concentrated in a certain value range. Or it is known, for example, that for the model parameters of transformation functions only values in a certain range generate acceptable transformations of the data. This information is used to set up a function that assigns a corresponding value of plausibility to a value for a model parameter, namely implausible values of the model parameters a low plausibility value and plausible values of the model parameters a high plausibility value. The next, third step is then based on the initial values of the model parameters S3 , a quality criterion is calculated that describes how well the simulation model is fits the training data. The so-called logarithmic marginal likelihood is usually used here for Gaussian process models. In order to take into account the fact that different values of the model parameters have different plausibilities, a fourth step is then used S4 a correction of the quality criterion is calculated based on the plausibility distribution, which is then calculated in a further fifth step S5 is used to correct the quality criterion. In a next sixth step S6 the value of at least one model parameter is then changed, with the aim of improving the corrected quality criterion and thus the simulation model. After that then the steps S3 until S6 iteratively repeated until a defined termination criterion is reached.

In the following, an advantageous use of the described method for calibrating a directly injecting Otto engine with regard to its particle emissions is described as an example as a function of the input variables load, speed, rail pressure, start and end of injection. This functional process of the gasoline engine as a component of the drive train is mapped as part of the calibration process using a simulation model in the form of a Gaussian process model for describing the particle emissions with the input variables, which is created using the method described. In the calibration process, measurement data are initially generated for the creation of the simulation model by operating and measuring the gasoline engine on a test bench according to a test plan with a list of measurement points to be measured. In the embodiments, this can be carried out as an offline method and as an online method. The data obtained through the measurements are used to create the simulation model. So-called training data, which are also obtained through the measurements, are used to train the simulation model or to check how well it matches the training data.

A particular challenge, for example, is the mapping of the behavior of the particle emissions of the direct-injecting gasoline engine as a function of the beginning and end of the injection, since the behavior of the particle emissions changes particularly quickly or is particularly non-linear. In order to be able to reproduce this behavior, it is advantageous to use transformations of the input and output variables. The use of warped, manifold or warped manifold Gaussian process model types has proven to be particularly advantageous. Here, however, the challenge arises that warped and manifold transformations tend to take on extreme forms, characterized by extreme values of the corresponding model parameters or hyperparameters, which is usually not desired. This problem can be solved with the procedure described. By using the plausibility distributions for the plausibility check of the model parameters, it is possible to prefer plausible values, that is to say in particular moderate values of the model parameters, and to evaluate extreme values of the model parameters as particularly implausible. In this way, those values for the model parameters are preferred during training that allow a plausible form of the simulation model to be expected and, in particular, in this example also the form of manifold and / or warping transformations.

Another example in which the method described can be used advantageously is the calibration of an engine in an online method on the test bench. Simulation models, and in particular Gaussian process models, are used to describe certain properties of the engine. The simulation models can then be used to determine a next measurement point. However, since every measurement on the test bench is time-consuming and costly, it must be ensured that as few nonsensical measuring points as possible are suggested due to poor simulation models in order to keep the number of measuring points as low as possible. It is therefore a challenge that the creation of the simulation model here has to be particularly robust, although at the same time, particularly at the beginning of the calibration on the test stand, particularly little training data is available. These challenges are solved with the method described. In order to be able to train a simulation model, and in particular a Gaussian process model, in a robust manner even with little training data, implausible values of the model parameters, and thus implausible forms of the simulation model, are avoided. For this purpose, a plausibility distribution across the model parameters is used, which assigns extreme values a low plausibility and moderate values a high plausibility. As a result, when training, simulation models with moderate values for the model parameters are preferred, which represent the system to be measured reliably with a higher degree of certainty.

Another example in which the method described can be used advantageously is the calibration of a function of autonomous driving, such as what is known as Adaptive Cruise Control (ACC), which regulates the vehicle speed taking into account the distance to a vehicle driving ahead. Here, a function for speed regulation is stored in a control unit, which is calibrated on the basis of several manipulated variables. Possible calibration variables are, for example, a speed exponent or a so-called coolness factor. These manipulated variables are varied and the vehicle behavior is tested and measured either in vehicle simulations or in real drives, for example values for the requested acceleration are measured. On the basis of these values, for example, a data-driven simulation model can be created that represents the maximum requested acceleration as a function of manipulated variables, such as a speed exponent or a coolness factor. In order to obtain a usable simulation model based on as few measurements as possible, it is advantageously possible to influence the training of the simulation model by specifying plausibility distributions in such a way that plausible model parameters are selected. The trained simulation model can then be used in an optimization process, for example to determine the speed exponent and / or the coolness factor in such a way that the maximum requested acceleration is as low as possible. The values for the speed exponent and the coolness factor can then be transmitted via a suitable interface to the control unit that executes the ACC.

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent literature cited

• WO 2016/198047 A1 [0003]

Claims (14)

Method for creating a simulation model for a calibration of a control device, wherein the control device controls at least one functional process of a component of a drive train and wherein the simulation model maps the at least one functional process of the component, wherein the simulation model is created on the basis of training data, with the following steps: S1) Definition of initial model parameters of the simulation model, S2) determining a plausibility distribution of the model parameters of the simulation model, S3) Calculating a quality criterion for the simulation model based on the model parameters, S4) determining a correction for the quality criterion based on the model parameters and the plausibility distribution, S5) Calculating a corrected quality criterion based on the correction, S6) Changing the model parameters in order to improve the corrected quality criterion, steps S3, S4, S5 and S6 being repeated. Procedure according to Claim 1 , characterized in that the plausibility distribution is estimated in step S2 using a comparison distribution, wherein the comparison distribution is determined in one or more of the following method steps: comparable functional process; - Carrying out repetitive measurements on the component of the drive train with unchanged values of input variables of the simulation model corresponding manipulated variables and analyzing a distribution of data of an output variable which corresponds to a measured variable of the simulation model; - Analyzing a distribution of data of input variables and / or output variables of the training data on which the simulation model is based. Method according to one of the preceding claims, characterized in that the plausibility distribution in step S2 is adapted as a function of a number of training data by shifting and / or scaling such that more relevant model parameters are assigned a higher plausibility with regard to the number of training data. Method according to one of the preceding claims, characterized in that the plausibility distribution is created in step S2 on the basis of empirical values, expectations and / or physical relationships. Method according to one of the preceding claims, characterized in that the plausibility distribution is described in step S2 via at least one of the model parameters on the basis of at least one sigmoid function. Method according to one of the preceding claims, characterized in that the simulation model is created in an online process, wherein the component of the drive train is operated on a test bench and wherein measured values measured on the component are continuously used as training data. Method according to one of the preceding claims, characterized in that a test planning of measurements on the component is created with the aid of the trained simulation model, the component being operated for the measurements on a test bench, in a vehicle or as a computer simulation of the functional process. Method according to one of the preceding claims, characterized in that non-parametric models are used as simulation models, Procedure according to Claim 8 , characterized in that nonparametric models of a Gaussian process model type, a manifold-Gaussian process model type, a warped-Gaussian process model type or a manifold-warped-Gaussian process model type are used as simulation models. Use of a simulation model in a control device, wherein the simulation model depicts at least one functional process of a component of a drive train and wherein the control device controls the at least one functional process, characterized in that the simulation model is created on the basis of training data, with the steps: S1) Define of initial model parameters of the simulation model, S2) determining a plausibility distribution of the model parameters of the simulation model, S3) calculating a quality criterion of the simulation model based on the model parameters, S4) determining a correction for the quality criterion based on the model parameters and the plausibility distribution, 55) calculating a corrected quality criterion based on the correction, S6) changing the model parameters in order to improve the corrected quality criterion, steps S3, S4, S5 and S6 being repeated. Computer program product comprising program codes, the program codes on a computer-readable medium are stored in order to perform the method according to one of the Claims 1 until 9 when the program is run in a computer. A method for calibrating a control device using at least one simulation model depicting a functional process of a component of a drive train, the at least one simulation model according to a method according to one of the Claims 1 until 9 is created. Procedure according to Claim 12 , wherein at least one calibration criterion is determined on the basis of an output variable of the at least one simulation model and wherein the simulation model is used to obtain output values by entering various input values, an optimized input value being sought for which the output value meets the at least one calibration criterion. Procedure according to Claim 13 , wherein the optimized input value of at least one input variable of one of the simulation models is stored in a characteristic diagram, a characteristic curve or a characteristic value of the control device.

Images