METHOD AND DEVICE FOR CREATING AN EMISSIONS MODEL OF AN INTERNAL COMBUSTION ENGINE

Abstract

A method for creating an emissions model of an internal combustion engine. The method begins with a provision of a plurality of measurement series at an internal combustion engine. There then follows a filtering of the measurement series using various low-pass filters, and ascertaining, from the filtered measurement series, those measurement series, when provided as an input variable for the emissions model during optimization of the emissions model, the smallest deviation from predicted emissions of the emissions model for measured emissions is achieved.

Claims

1. A method for creating an emissions model of an internal combustion engine, comprising the following steps: providing a plurality of measurement series acquired at an internal combustion engine; filtering the measurement series using various low-pass filters; optimizing a plurality of emissions models for respectively different combinations of the filtered measurement series as input variables for the emissions models, in such a way that a deviation of a predicted emissions of each emissions model for associated acquired emissions for an acquired measurement series is minimized; selecting an optimized emissions model of the plurality of optimized emissions models that achieves a lowest deviation from predicted emissions of the emissions model for measured emissions; and outputting the optimized emissions model having a smallest deviation.

2. The method as recited in claim 1, wherein the various low-pass filters differ in that they have different time constants relative to one another, the time constant of each respective low-pass filter of the various low pass filters characterizing which preceding measurement values the respective low-pass filter takes into account.

3. The method as recited in claim 1, wherein the various low-pass filters are first-order low-pass filters.

4. The method as recited in claim 1, wherein, during the optimization of the emissions models, at least a most recent measurement points of a plurality of the acquired measurement series are taken into account and are used for the input variables.

5. The method as recited in claim 1, wherein each acquired measurement series is assigned to a measurement variable, the measurement variable characterizing a variable of the internal combustion engine, the variable characterizing: an engine rotational speed or relative cylinder air filling or lambda or ignition angle or cylinder ignition information, and wherein the predicted emission is: (i) a particle emission mass and/or number, and/or or (ii) a gaseous emission of NOx or THC or CO, and/or (iii) temperature and/or pressure of an emission.

6. The method as recited in claim 3, wherein, when each low-pass filter is applied to the first measurement values, additional fictive measurement values are provided as an input variable for the low-pass filter, the additional fictive measurement values characterizing a constant operating point of the engine at low load.

7. A method, comprising: creating an emissions model of an internal combustion engine, including: providing a plurality of measurement series acquired at an internal combustion engine; filtering the measurement series using various low-pass filters; optimizing a plurality of emissions models for respectively different combinations of the filtered measurement series as input variables for the emissions models, in such a way that a deviation of a predicted emissions of each emissions model for associated acquired emissions for an acquired measurement series is minimized; and selecting an optimized emissions model of the plurality of optimized emissions models that achieves a lowest deviation from predicted emissions of the emissions model for measured emissions; using the selected optimized emissions model for predicting emissions of the internal combustion engine or at an outlet of an exhaust pipe that is connected to the internal combustion engine, an acquired measurement series being filtered with a selected low-pass filter and the selected optimized emissions model predicting the emissions of the internal combustion engine as a function of the filtered acquired measurement series as input variables.

8. A device configured to create an emissions model of an internal combustion engine, the device configured to: provide a plurality of measurement series acquired at an internal combustion engine; filter the measurement series using various low-pass filters; optimize a plurality of emissions models for respectively different combinations of the filtered measurement series as input variables for the emissions models, in such a way that a deviation of a predicted emissions of each emissions model for associated acquired emissions for an acquired measurement series is minimized; select an optimized emissions model of the plurality of optimized emissions models that achieves a lowest deviation from predicted emissions of the emissions model for measured emissions; and output the optimized emissions model having a smallest deviation.

9. A non-transitory machine-readable storage medium on which is stored a computer program for creating an emissions model of an internal combustion engine, the computer program, when executed by a computer, causing the computer to perform the following steps: providing a plurality of measurement series acquired at an internal combustion engine; filtering the measurement series using various low-pass filters; optimizing a plurality of emissions models for respectively different combinations of the filtered measurement series as input variables for the emissions models, in such a way that a deviation of a predicted emissions of each emissions model for associated acquired emissions for an acquired measurement series is minimized; selecting an optimized emissions model of the plurality of optimized emissions models that achieves a lowest deviation from predicted emissions of the emissions model for measured emissions; and outputting the optimized emissions model having a smallest deviation.

Description

BRIEF DESCRIPTION OF THE DRAWING

[0022] FIG. 1 schematically shows a flow diagram of a specific embodiment of the present invention.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

[0023] FIG. 1 schematically shows a flow diagram of a method (10) for creating an emissions model of an internal combustion engine.

[0024] The method (10) begins with step S11. In this step, measurement series of a wide variety of measurement variables are acquired at the internal combustion engine that characterize the internal combustion engine, preferably a state of the internal combustion engine. One of the measurement variables can for example be an engine rotational speed. In addition, the emissions of the internal combustion engine associated with the respectively acquired measurement variables are also acquired as a measurement series. These acquired data are subsequently used as training data.

[0025] There then follows step S12. This step can be designated “filtering.” In this step, the raw data of the measurement series from step S11 are prepared. In addition to other calculations, they are filtered using a plurality of different low-pass filters. The outputs of the various low-pass filterings are then provided to an optimization software, in addition to the classical inputs. The optimization software is preferably ETAS ASCMO. It is possible that the various low-pass filterings are added to the feature matrix in the form of further rows.

[0026] Preferably, the low-pass filters differ in that they filter a different number of measurement points within the measurement series. Particularly preferably, the low-pass filters are first-order low-passes.

[0027] Preferably, the low-pass is described by the mathematical equation:

out.sub.t=out.sub.t-1+(in−out.sub.t-1)*dT/T

[0028] In the next following step S13, using the optimization software an optimal combination of the various low-pass filterings and classical inputs is ascertained, with the result that the emissions model predicts as precisely as possible the likewise associated acquired emissions of the internal combustion engine from step S11. For this purpose, for example the optimization software can combinatorially test which combinations of the different low-pass filtering result in the best emissions model. This can be carried out for example by selecting a combination of the various low-pass filterings and, using these as input variables for the emissions model, optimizing this model, so that a deviation between the predicted emissions of the emissions model and the acquired emissions from step S11 is minimized.

[0029] Preferably, the emissions model is a GP-NARX, which is trained for example using supervised learning. Other machine learning systems, and other model types, are also possible. It has turned out that neural networks, in particular RNN, are also particularly suitable.

[0030] It is to be noted that the selection of various low-pass filtering and subsequent optimization of the emissions model is carried out multiple times for combinations, different in each case, of the low-pass filtering. Thus, a plurality of emissions models are optimized, each having different input variables.

[0031] In a further specific embodiment, in step S13 in addition time constants of the low-pass filter can be optimized using e.g. a genetic algorithm.

[0032] From the plurality of optimized emissions models, the emissions model is then selected that achieves the smallest deviation between the predicted emissions of the emissions model and the acquired emissions from step S11.

[0033] After step S13, the optional step S14 can be carried out. In this step, the selected emissions model can be tested with validation data. The validation data are further measurement series that were not contained in the training data and are thus used to test the emissions model as to whether the training data were thoroughly learned. If step S14 yields the result that the selected emissions model was not correctly trained, then step S13 can be carried out again. It is possible that this testing with validation data be carried out in step S13 in accordance with specifiable optimization steps.

[0034] This selected emissions model, having the smallest deviations, is then outputted in step S15 as the most accurate emissions model. Preferably, in step S15 the associated measurement variables are also outputted whose assigned measurement series were used as input variables for this emissions model. Preferably, in addition instep S15 the associated low-pass filters are also outputted whose filtered measurement series were used as input variables. Preferably also an associated feature matrix.

[0035] After step S15 has been carried out, the method (10) ends. The outputted emissions model from step S15 can then be used for various applications. For example, this emissions model can then be used to predict emissions of the internal combustion engine. It is also possible that this emissions model be used for the prediction of emissions at the outlet of an exhaust pipe connected to the internal combustion engine. Here, for example a further model can be used with which the predicted emissions of the engine are then further processed by taking into account physical effects of the exhaust gas system in order to then calculate the emissions at the output of the exhaust gas system.

[0036] In the use of the emissions model for the prediction of emissions, measurement series can be acquired that are assigned to the associated outputted measurement variables from step S15 and are filtered with the outputted low-pass filters from step S15, and are then used as input variables for the emissions model.