Method and apparatus for operating a fuel injection valve with the aid of machine learning methods

Patent metadata

Field	Value
Publication number	US-11846244-B2
Application number	US-202118001784-A
Country	US
Kind code	B2
Filing date	Sep 10, 2021
Priority date	Sep 11, 2020
Publication date	Dec 19, 2023
Grant date	Dec 19, 2023

Abstract

A method for operating an injection valve by ascertaining an opening time and/or closing time of the injection valve on the basis of a sensor signal. The method includes: providing an analysis point time series by sampling a sensor signal of a sensor of the injection valve; using a nonlinear, data-based first submodel in order to obtain a first model output on the basis of the analysis point time series; using a linear, data-based second submodel in order to obtain a second model output on the basis of the analysis point time series; ascertaining the opening time and/or closing time as a function of the first and second model outputs.

First claim

What is claimed is:

1. A method for operating an injection valve by ascertaining an opening time and/or closing time of the injection valve based on a sensor signal, the method comprising the following steps:
providing an analysis point time series by sampling a sensor signal of a sensor of the injection valve;
using a nonlinear, data-based first submodel to obtain a first model output based on the analysis point time series;
using a linear, data-based second submodel to obtain a second model output based on the analysis point time series;
ascertaining the opening time and/or closing time as a function of the first model output and the second model output;
wherein the first submodel and the second submodel are configured so that they output an output vector as the first model output and the second model output, respectively, each element of the output vectors being assigned to a determined opening or closing time, the first and second submodels being configured so that they each indicate a value of each element of the output vector according to a probability with which a time determined by an index value of the element corresponds to the opening time or closing time to be output.

2. The method as recited in claim 1 , wherein the opening time and/or closing time is ascertained by correspondingly adding together the elements of the output vectors in an index-wise weighted manner using a defined weighting factor to obtain a total output vector, an index value of an element of the total output vector having a correspondingly highest value being ascertained and the ascertained index value being assigned to a corresponding opening time and/or closing time.

3. The method as recited in claim 1 , wherein the operation of the injection valve is carried out as a function of the opening time and/or closing time, the operation of the injection valve being performed in such a way that an opening period of the injection valve, which is determined by the ascertained opening time and/or closing time, is set to a defined setpoint opening period.

4. A method for operating an injection valve by ascertaining an opening time and/or closing time of the injection valve based on a sensor signal, the method comprising the following steps:
providing an analysis point time series by sampling a sensor signal of a sensor of the injection valve;
using a nonlinear, data-based first submodel to obtain a first model output based on the analysis point time series;
using a linear, data-based second submodel to obtain a second model output based on the analysis point time series;
ascertaining the opening time and/or closing time as a function of the first model output and the second model output;
wherein the first submodel is in the form of a nonlinear neural network and the second submodel is in the form of a linear neural network.

5. A method for operating an injection valve by ascertaining an opening time and/or closing time of the injection valve based on a sensor signal, the method comprising the following steps:
providing an analysis point time series by sampling a sensor signal of a sensor of the injection valve;
using a nonlinear, data-based first submodel to obtain a first model output based on the analysis point time series;
using a linear, data-based second submodel to obtain a second model output based on the analysis point time series;
ascertaining the opening time and/or closing time as a function of the first model output and the second model output;
wherein the first and second submodels are configured so that they each output a time by regression, and the times are added together in a weighted manner as a function of a weighting factor to determine the opening time or closing time.

6. A method for training submodels for a data-based analysis model for determining an opening time and/or closing time of an injection valve, comprising the following steps:
providing training data sets, which indicate a determined opening time and/or closing time for an analysis point time series;
training a first submodel, which takes the form of a nonlinear, data-based model, using the training data sets;
training a second submodel, which takes the form of a linear, data-based model, using the training data sets.

7. The method as recited in claim 6 , wherein a weighting factor is determined for a weighted combining of first and second model outputs of the first and second submodels, respectively, so that, for a quantity of defined training data sets, a correct model output of a time as an opening time or closing time by the second submodel does not change as a result of a model output of the first data-based submodel.

8. An apparatus configured to operate an injection valve by ascertaining an opening time and/or closing time of the injection valve based on a sensor signal, the apparatus configured to:
provide an analysis point time series by sampling a sensor signal of a sensor of the injection valve;
use a nonlinear, data-based first submodel to obtain a first model output based on the analysis point time series;
use a linear, data-based second submodel to obtain a second model output based on the analysis point time series;
ascertain the opening time and/or closing time as a function of the first model output and the second model output;
wherein the first submodel is in the form of a nonlinear neural network and the second submodel is in the form of a linear neural network.

9. A non-transitory machine-readable storage medium on which are stored commands for operating an injection valve by ascertaining an opening time and/or closing time of the injection valve based on a sensor signal, the commands, when executed by a computer, causing the computer to perform the following steps:
providing an analysis point time series by sampling a sensor signal of a sensor of the injection valve;
using a nonlinear, data-based first submodel to obtain a first model output based on the analysis point time series;
using a linear, data-based second submodel to obtain a second model output based on the analysis point time series;
ascertaining the opening time and/or closing time as a function of the first model output and the second model output;
wherein the first submodel is in the form of a nonlinear neural network and the second submodel is in the form of a linear neural network.

Assignees

• Bosch Gmbh Robert

Inventors

• Hopf Andreas
• Tonner Erik
• Kowol Frank
• Barth Jens-Holger
• Groh Konrad
• Woehrle Matthias
• Meister Mona
• Norden Roland

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• F02D41/1405 Primary

  Neural network control · CPC title

• F02D41/2425

  Particular ways of programming the data · CPC title

• F02D2041/1433

  using a model or simulation of the system · CPC title