WO2016198046A1 - Verfahren für die auswahl eines simulationsmodells zur abbildung wenigstens eines funktionalen prozesses einer antriebsstrangkomponente aus einer optimierten modellmenge - Google Patents
Verfahren für die auswahl eines simulationsmodells zur abbildung wenigstens eines funktionalen prozesses einer antriebsstrangkomponente aus einer optimierten modellmenge Download PDFInfo
- Publication number
- WO2016198046A1 WO2016198046A1 PCT/DE2016/100262 DE2016100262W WO2016198046A1 WO 2016198046 A1 WO2016198046 A1 WO 2016198046A1 DE 2016100262 W DE2016100262 W DE 2016100262W WO 2016198046 A1 WO2016198046 A1 WO 2016198046A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- model
- simulation
- models
- evaluation
- selection
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Definitions
- the present invention relates to a method for the selection of a simulation model for mapping at least one functional process of a drivetrain component from an optimized model set and to a corresponding computer program product for carrying out such a method.
- test benches and their use are associated with a high expenditure of time and correspondingly high costs during operation. Therefore, there is currently a desire to significantly reduce the time for the use of test benches.
- One way of reducing test bench occupancy times is to use a so-called “Design of Experiment”, also called DoE, for the test bench trials, and then to create simulation models from the test bench data that allow a mathematical description of the real functional process of the powertrain components , Accordingly, such simulation models calculate the relationship between input values and output values, so that a calculation or a simulation for at least parts of the functional process of the real drive train components on a computer can be computationally determined.
- a disadvantage of the known solutions for simulation models is to choose the most suitable simulation model.
- mathematical methods are used to create simulation models from basic models on which a basic function space is based, and to adapt them to corresponding relationships between input variables and output variables.
- a multiplicity of possible simulation models are validated and usually evaluated manually with regard to their qualitative evaluation with reference to the real test bench or the real internal combustion engine.
- the actual selection and assessment in qualitative terms for the individual simulation models is still based to a degree on the experience and previous knowledge of an associated test engineer or user.
- the disadvantage here is that sufficient experience is necessary in order to be able to make a targeted selection of the simulation models.
- a method for selecting a simulation model for the simulation of at least one functional process of an internal combustion engine from an optimized model set. Such a method comprises the following steps: Selecting at least one selection criterion for the qualitative assessment of the simulation models of the model set,
- a process is provided upstream, which provides a model set with optimized simulation models.
- data models are conceivable, which can be used as a simulation model.
- data-based regression methods which in particular undergo an optimization step are preferred.
- the result of such optimization steps is usually a plurality of optimized simulation models, so that, accordingly, the optimized model set forms as an input situation for a method according to the invention.
- consistently at least one selection criterion is selected for the qualitative evaluation of the simulation models of the model set. Possibilities for such qualitative selection criteria will be explained later. In particular, it is about evaluating the individual models in terms of their usability for the later use of the simulation model.
- the desired correlation accuracy with actual measurement data of functional processes of a component of a drive train, in particular an internal combustion engine can be used.
- a reduction of uncertainty bands around a corresponding function course of a simulation model can provide a selection criterion.
- the complexity of individual simulation models can be used to provide a qualitative assessment option for the simulation models.
- the step of selection can be done manually, for example, by the respective user of a method according to the invention. In principle, however, it is also conceivable to automate or partially automate these and / or further steps.
- the selection of a grading series for carrying out an efficient method according to the present invention is follow a crucial role.
- the evaluation sequence is decisive for how long a corresponding evaluation and selection process takes with a method according to the invention.
- simply the corresponding total optimized model set and thus all simulation models, accordingly in the evaluation process with the at least one selected selection criterion can be used.
- more complex sequential orders are also conceivable as evaluation order within the meaning of the present invention in order to be able to achieve optimizations, for example with regard to the computing time or the quality of the overall result.
- At least the last two steps of the method according to the invention ie the performance of the qualitative evaluation and the selection of the simulation model, are carried out in an automatic manner.
- the selection of the at least one selection criterion as well as the selection of the evaluation order can be carried out both manually and also automatically or semi-automatically.
- the subsequent steps of the qualitative assessment and the selection of the simulation model with the best qualitative evaluation are automated and essentially carried out directly in time.
- a functional process which is to be imaged by the simulation model in the sense of the present invention, is in particular a process which correlates input parameters of a component of a drive train, in particular of an internal combustion engine, with output parameters.
- input parameters may be an air mixture, a temperature ratio, an accelerator pedal position or a load change number.
- a fuel consumption of the internal combustion engine or the gas composition of the exhaust gas can be used as output parameters.
- the simulation models can now be used again in order to avoid or significantly reduce actual test bench time.
- the selection possibility of a possibly optimized simulation model of a broader set of users is now opposite.
- the version of the simulation model with the best qualitative rating selected at the end can be used in a variety of ways. be set. For example, at this time the selected simulation model is being used to set up a corresponding virtual test bench. It is also possible to condition or partially condition vehicle control units using the selected simulation model.
- a simulation model is optimally available for the user with regard to an ideal qualitative evaluation.
- a method according to the invention can be carried out offline and thus independently of a test stand.
- an online application is performed.
- a test stand is in action, so for example an internal combustion engine is operated in different modes.
- the method according to the invention for generating a simulation model is running, wherein the data from the current test bench trial are introduced into the optimization runs.
- a feedback from an optimization run in the current experiment is also conceivable, for example, to re-use areas with high uncertainties with current measured values from the ongoing test bench test.
- a method according to the invention or the simulation model produced can also be used directly in a vehicle, for example in a control unit, for example engine control unit or transmission control unit of a vehicle.
- the selection of the selection criterion is carried out from a plurality of at least two selection criteria, in particular from at least two of the following selection criteria: - Data to fit
- a selection criterion also includes an approximation to the selection criterion.
- data to fit is understood to mean that a simulation model according to this selection criterion is considered particularly good if the correlation between input measurement points as the basis for the optimization of the simulation model with the optimized simulation model is as appropriate. In other words, a corresponding function curve of the simulation model runs as accurately as possible by measuring points which underlie an optimization run of the simulation model.
- the complexity of the simulation model can preferably be included in a corresponding combinatorial for the selection criterion as a mathematical penalty term.
- the complexity of a simulation model the number of parameters / hyperparameters or the size of the basis function space of the base model can be used.
- quality parameters can be determined in an abstract manner, which are preferably co-determined during the optimization runs for the individual simulation models. Thus, these quality parameters already exist at the time of the presence of the model set of all simulation models, so that a pure comparison can carry out the method according to the invention.
- Such quality parameters also allow individual selection criteria to be combined from a technical point of view in a mathematical term.
- An uncertainty of the simulation model is to be understood both punctually and globally.
- stochastic variances are to be determined for the individual simulation models and their functional relationships, which reflect the uncertainty of this simulation model on the basis of the measured values made available.
- Another advantage is when the least number of local optima of the simulation model than at least a selection criterion is used. This leads to an increased uniqueness, which area of a functional relationship should actually be made available later for the control of a control unit as a local optimum for the control specification of a desired-actual loop.
- the data volumes used for the selection criterion are not used to create the optimized simulation models. These are so-called validation data. These are also measured at the test bench but are not included in the training of the models. That They check the prediction accuracy of the models by comparing the output variables determined by the simulation model with the actual measured output variables.
- the model set has at least two models, wherein as evaluation order first the model shares are qualitatively evaluated, after this evaluation the best quality model set is selected, and then from the selected model set the best quality simulation model is selected .
- This is a possibility of an evaluation sequence, which is used in particular when the model set is already designed on the basis of different optimization runs in multiple parts and thus in particular with at least two model shares.
- optimization runs can be used to generate a model set of a multiplicity of simulation models for each optimization run.
- the set of models is composed of a plurality of sets of models, which in turn contain a plurality of simulation models.
- a selection in a first step can be made roughly, which has a significantly lower computational requirement than the qualitative evaluation of all existing simulation models of all model families.
- the best-in-class model has been assessed and the selection made, then the best quality simulation model will be selected from this selected set of models. This is preferably exclusively in the selected model set a corresponding qualitative assessment of the simulation models contained therein performed.
- this two-step process can use different selection criteria for each stage.
- the reduction of the computational effort may lead to a situation in which individual simulation models in a few qualitative sets of models rather than singular simulation models would have a high quality, which would be discarded by the negative evaluation of the model population.
- an embodiment according to this paragraph provides an advantage in terms of the correlation between the evaluation time and the quality of the selection made.
- the model set has at least two model shares, wherein all simulation models of all model shares are qualitatively evaluated as the evaluation sequence, and then the simulation model with the best qualitative evaluation is selected from the model set of all model shares.
- a total quantity is formed from all simulation models of all model shares, which is completely fed into the evaluation process.
- the model shares are brought together in the common model set and thus the qualitative assessment is actually carried out for each individual simulation model which is present in all model sets.
- the increased computational effort achieved in this way is justified by the fact that an optimized selection option for the best-quality simulation model of all existing simulation models of all model shares can now actually be carried out.
- a further advantage may be that, in a method according to the invention, the model set has at least two model shares, the evaluation order chosen being the best-quality family of models from each model family, and then the best quality simulation model being selected from all family simulation models.
- This is preferably a combination of the embodiment of the two preceding paragraphs.
- a qualitative evaluation is carried out for each model family and thus also for all simulation models of the entire model set.
- a two-stage process is also provided here, so that, in particular, a time-parallel processing can take place. If, for example, a first model family is provided and generated in a first optimization run, then the second optimization run for generating a second model family is then started. tet.
- the selection of the at least one selection criterion and / or the selection of the evaluation sequence is based on at least one base parameter, in particular designed as hyperparameter, of the basic model of the model model simulation models.
- this is an automatic or semi-automatic selection.
- a corresponding basic parameter can take into account possible disadvantages or uncertainties of the corresponding function.
- Such basic parameters can thus lead to eliminate or reduce possible disadvantages in the selection of the corresponding function in an automated manner by skillful selection of the selection criterion.
- this correlation takes place in an automated manner, so that no dependence on a potentially existing level of knowledge of the user is more necessary.
- At least one selection criterion is switched on and / or off before selecting at least one selection criterion.
- the pre-switching of a purely manual selection step in the form of a switch-on and a switch-off makes it possible, despite a high degree of automation of the procedure to provide manual adaptability. This combines high flexibility in the use of a method according to the invention with the advantages described by the degree of automation.
- the simulation models of the model set are data-based regression models, in particular using Bayes regression methods, preferably Gaussian process models.
- data-based regression methods it is possible to significantly reduce the necessary measurement data for the optimization runs and thus to minimize the necessary test bench time before the optimization runs are carried out.
- the use of data-based regression methods is also simple and inexpensive in the actual implementation of a method according to the invention.
- At least two different optimization runs for a basic model of a simulation model are carried out for generating the model set, wherein each optimization run generates a model family of optimized simulation models.
- at least a double optimization is carried out, preferably a multi-stage optimization, it being possible to use a very wide variety of basic models and / or a wide variety of transformation functions explained below, as well as combinations thereof.
- a further advantage of the present invention is that these transformation functions are now not created separately before the optimization runs for the simulation model, rather, the optimization run is adapted to a combination of the base model with the respective transformation function.
- At least one input transformation function for input values of the simulation model and / or one output transformation function for output values of the simulation model are taken into account in at least one combined optimization run of the base model of the simulation model and the respective transformation .
- the use of an input transformation function in a combined optimization run can basically be referred to as a "manifold" basic model or "manifold” Gaussian process.
- the output transformation function may be understood as a "warped” functionality for a corresponding base model and, in particular, a Gaussian process.
- the optimization run then runs over the total combination of the corresponding transformation functions and the base model.
- a combinatorics takes into account all possible combinations, ie in particular four optimization runs are carried out.
- a first optimization run will optimize the base model alone, a second combined basic model and input transformation function optimization, a third combined basic model and output transformation function optimization, and a final, fourth combined optimization model basic model and input transformation function and output transformation function .
- the four model families which now yield the model set with a large number of simulation models, now provide the basis on which the evaluation and selection steps according to the invention are carried out.
- a computer program product stored on a computer-readable medium for the selection of a simulation model for imaging at least one functional process of a drivetrain component from an optimized model set, comprising:
- computer readable program means for causing a computational unit to automatically select the simulation model from the model set with the best qualitative score.
- the computer program product is preferably provided with computer readable program means for causing a computing unit to perform the steps according to a method having the features of claims 1 to 12.
- a computer program product according to the invention brings the same advantages as have been explained in detail with reference to a method according to the invention.
- FIG. 1 shows a representation of a simulation model
- FIG. 2 shows a possibility of an optimization run for the generation of a model family
- FIG. 3 shows another possibility for generating a model family
- FIG. 4 shows another possibility for generating a model crowd
- FIG. 5 shows a further possibility for generating a model family
- FIG. 6 shows a possibility of an evaluation sequence
- Figure 7 shows another possibility of an evaluation order
- Figure 8 shows another possibility of an evaluation order.
- 1 shows schematically how a number of input values E can be correlated with an arbitrary number of output values A by means of a simulation model 10.
- This simulation model is preferably a data-based regression method based on stochastic fundamentals.
- a Gaussian process model is used here.
- FIGS. 2 to 5 show ways in which a corresponding training can be carried out by optimizing the basic model 14 on the basis of data measured at a test bench or also data from a computer experiment.
- a basic model 14 is given as a Gaussian process model, as shown in FIG. 5. With a corresponding optimization run O, it becomes possible to optimize this basic model 14 and to provide a model family 22 with a multiplicity of optimized simulation models 10.
- FIG. 6 shows a possibility in which, in a first step, the individual model shares 22 are qualitatively evaluated from the model set 20.
- the result is a model family 22, which is regarded as the best-quality model family 22.
- the selection of the simulation model 10 with the best qualitative evaluation now ensues from this selected model family 22.
- the entire model set 20 can also be evaluated with regard to all model shares 22, so that the best high-quality simulation model 10 can be made available directly following this now one-step process with the aid of the selection criterion 30.
- the entire model set 20 is already evaluated with regard to the individual model shares 22.
- the best simulation model 10 is selected.
- the corresponding selection criterion 30 can be selected specifically for the individual model shares 22.
- the evaluation and selection for this last model family 22 is now carried out, so that in the second step, in turn, a significantly lower number of four simulation models 10 in this example is used to select the best quality simulation model 10.
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
Claims
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE112016002594.7T DE112016002594A5 (de) | 2015-06-10 | 2016-06-10 | Verfahren für die auswahl eines simulationsmodells zur abbildung wenigstens eines funktionalen prozesses einer antriebsstrangkomponente aus einer optimierten modellmenge |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE102015109222.0 | 2015-06-10 | ||
DE102015109222 | 2015-06-10 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2016198046A1 true WO2016198046A1 (de) | 2016-12-15 |
Family
ID=56686599
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/DE2016/100262 WO2016198046A1 (de) | 2015-06-10 | 2016-06-10 | Verfahren für die auswahl eines simulationsmodells zur abbildung wenigstens eines funktionalen prozesses einer antriebsstrangkomponente aus einer optimierten modellmenge |
Country Status (2)
Country | Link |
---|---|
DE (1) | DE112016002594A5 (de) |
WO (1) | WO2016198046A1 (de) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102016203073A1 (de) | 2016-02-26 | 2017-08-31 | Airbus Defence and Space GmbH | Werkzeugsystem und Verfahren zur Herstellung eines Faserverbundhalbzeugs sowie Verfahren zur Herstellung eines Faserverbundbauteils |
CN110263363A (zh) * | 2019-04-25 | 2019-09-20 | 珠海格力电器股份有限公司 | 设备配管选型方法和装置 |
DE102021002319A1 (de) | 2020-06-06 | 2021-12-09 | FEV Europe GmbH | Verfahren zum Auswählen eines Simulationsmodells, Computerprogrammprodukt und Verfahren zur Kalibrierung eines Steuergeräts |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102012202802A1 (de) * | 2012-02-23 | 2013-08-29 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Verfahren zum auswählen eines lösungsvorschlags für einen vorbestimmten prozess |
-
2016
- 2016-06-10 WO PCT/DE2016/100262 patent/WO2016198046A1/de active Application Filing
- 2016-06-10 DE DE112016002594.7T patent/DE112016002594A5/de not_active Withdrawn
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102012202802A1 (de) * | 2012-02-23 | 2013-08-29 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Verfahren zum auswählen eines lösungsvorschlags für einen vorbestimmten prozess |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102016203073A1 (de) | 2016-02-26 | 2017-08-31 | Airbus Defence and Space GmbH | Werkzeugsystem und Verfahren zur Herstellung eines Faserverbundhalbzeugs sowie Verfahren zur Herstellung eines Faserverbundbauteils |
CN110263363A (zh) * | 2019-04-25 | 2019-09-20 | 珠海格力电器股份有限公司 | 设备配管选型方法和装置 |
DE102021002319A1 (de) | 2020-06-06 | 2021-12-09 | FEV Europe GmbH | Verfahren zum Auswählen eines Simulationsmodells, Computerprogrammprodukt und Verfahren zur Kalibrierung eines Steuergeräts |
Also Published As
Publication number | Publication date |
---|---|
DE112016002594A5 (de) | 2018-05-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
AT512977B1 (de) | Methode zur Ermittlung eines Modells einer Ausgangsgröße eines technischen Systems | |
WO2016198047A1 (de) | Verfahren für die erstellung eines simulationsmodells zur abbildung zumindest eines funktionalen prozesses einer antriebstrangkomponente | |
DE102019124018A1 (de) | Verfahren zum Optimieren von Tests von Regelsystemen für automatisierte Fahrdynamiksysteme | |
WO2017114883A1 (de) | Verfahren zum konfigurieren einer co-simulation für ein gesamtsystem | |
DE102007006616B3 (de) | Verfahren zur Optimierung eines elektronisch gesteuerten automatisch schaltenden Getriebes für ein Kraftfahrzeug | |
WO2016198046A1 (de) | Verfahren für die auswahl eines simulationsmodells zur abbildung wenigstens eines funktionalen prozesses einer antriebsstrangkomponente aus einer optimierten modellmenge | |
DE102020003428A1 (de) | Verfahren zur Erstellung eines Simulationsmodells, Verwendung eines Simulationsmodells, Computerprogrammprodukt, Verfahren zur Kalibrierung eines Steuergeräts | |
WO2004102291A1 (de) | Auslegung von werkzeugen und prozessen für die umformtechnik | |
WO2021259980A1 (de) | Training eines künstlichen neuronalen netzwerkes, künstliches neuronales netzwerk, verwendung, computerprogramm, speichermedium und vorrichtung | |
DE102020213888A1 (de) | Computerimplementiertes Verfahren zum Bestimmen von Kritikalitätswerten eines technischen Systems | |
WO2020207789A1 (de) | Verfahren und vorrichtung zum ansteuern einer technischen einrichtung | |
WO2011082908A1 (de) | Verfahren zum ermitteln von funktionsparametern für ein steuergerät | |
DE10222699B4 (de) | Computergestütztes Verfahren und Vorrichtung zur Optimierung von Fahrzeugstrukturen und -komponenten | |
DE102012102373A1 (de) | Verfahren zur Abschätzung eines Ressourcenverbrauchs bei der Erzeugung eines Steuergeräteprogrammcodes | |
DE102022112606B3 (de) | Computerimplementiertes Verfahren zur Kalibrierung eines technischen Systems | |
DE102019132624A1 (de) | Verfahren, Vorrichtung, Computerprogramm und computerlesbares Speichermedium zum Erstellen eines Motion Cueing Algorithmus | |
DE10325513B4 (de) | Verfahren und Vorrichtung zum Erstellen eines Verhaltensaspekts einer Schaltung zur formalen Verifikation | |
DE102018217139B4 (de) | Verfahren und Vorrichtung zum Konfigurieren einer Kennfunktion eines Steuergeräts eines Kraftfahrzeugs | |
EP3575976A1 (de) | Verfahren zum bestimmen einer physikalischen verbindungstopologie eines für die steuergerätentwicklung eingerichteten, echtzeitfähigen testgeräts | |
AT524858B1 (de) | Erstellungsverfahren für eine Erstellung wenigstens eines System- Einzelmodells für ein Brennstoffzellensystem zur Modellierung von Zusammenhängen von System-Eingangsparametern und System-Ausgangsparametern | |
DE102020133654B3 (de) | Computerimplementiertes Verfahren zur Modifizierung einer Komponente eines computergenerierten Modells eines Kraftfahrzeugs | |
EP3491517B1 (de) | Signalflussbasiertes computerprogramm mit direct-feedthrough-schleifen | |
DE102006034018A1 (de) | Verfahren und Vorrichtung zur Produktionsplanung | |
WO2022242812A1 (de) | Computerimplementiertes verfahren und system zur bestimmung von optimierten systemparametern eines technischen systems mittels einer kostenfunktion | |
DE102018112803A1 (de) | Verfahren zum Bestimmen einer physikalischen Verbindungstopologie eines für die Steuergerätentwicklung eingerichteten, echtzeitfähigen Testgeräts |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 16751464 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 112016002594 Country of ref document: DE |
|
REG | Reference to national code |
Ref country code: DE Ref legal event code: R225 Ref document number: 112016002594 Country of ref document: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 16751464 Country of ref document: EP Kind code of ref document: A1 |