Method and device for operating a fuel injection valve

Abstract

The invention relates to a method for training a data-based evaluation model for determining the opening or closing time of an injection valve (6) based on a sensor signal (S), the method comprising the following steps: · measuring (S1) the operation of the injection valve (6) in order to determine at least one sensor signal (S) and the associated opening or closing time; · sampling the sensor signal (S) at a sampling rate in order to obtain a sensor signal time series with sensor signal values; · determining (S4) a plurality of training data sets in such a way that a plurality of evaluation point time series generated from the sensor signal time series are assigned to the opening or closing time associated with the sensor signal, wherein the evaluation point time series has a lower time resolution than the sensor signal time series; · training (S5) a data-based evaluation model based on the determined training data sets.

Description

Technical Field

[0001] The present invention relates to a method for operating a fuel injector of an internal combustion engine using a data-based model, in particular for determining a closing time or estimating an injected fuel quantity. Background Art

[0002] Electromechanical or piezoelectric injection valves are used to meter fuel in combustion engines. These injection valves enable a direct and precisely measured supply of fuel to the cylinders of the combustion engine.

[0003] The challenge is to control the combustion process as precisely as possible in order to improve the operating characteristics of the combustion engine, particularly with regard to fuel consumption, efficiency, pollutant emissions, and smooth operation. For this purpose, it is crucial to operate the injection valves in such a way that the injected fuel quantity can be metered with high repeatability, even with varying operating pressures and, if necessary, multiple injections per operating cycle.

[0004] Injection valves can have electromagnetic or piezoelectric actuators, which actuate the valve needle to remove it from its valve seat and open the outlet opening of the injection valve to release fuel into the combustion chamber. Due to structural differences and varying operating conditions, such as temperature, fuel pressure, and fuel viscosity, there is uncertainty in determining the exact opening and closing times of the injection valve. The opening time is the time at which fuel passes through the injection valve and reaches the combustion chamber of the cylinder, and the closing time is the time until which fuel passes through the injection valve and reaches the combustion chamber of the cylinder. Summary of the Invention

[0005] According to the invention, a method for operating a fuel injection valve is provided according to claim 1 , as well as a device and an injection system according to the independent claims.

[0006] Further embodiments are described in the dependent claims.

[0007] According to a first aspect, a method for training a data-based evaluation model for determining an opening and / or closing time of an injection valve based on a sensor signal is provided, the method comprising the following steps:

[0008] - measuring the operation of the injection valve in order to determine at least one sensor signal and the associated opening and / or closing time;

[0009] - sampling the sensor signal at a sampling rate to obtain a sensor signal time series having sensor signal values;

[0010] - determining a plurality of training data sets by assigning a plurality of evaluation point time series generated from the sensor signal time series to the switch-on or switch-off instants associated with the sensor signal, wherein the evaluation point time series has a lower time resolution than the sensor signal time series;

[0011] -Training a data-based evaluation model based on the determined training dataset.

[0012] Although the injection valve is actuated according to a predetermined course of the actuation signal, the resulting variations in the opening and closing movement of the injection valve make it impossible to precisely predetermine the actual opening and closing times for starting and ending fuel injection. This is due to the complex dependence of the valve movement on the current operating point.

[0013] To monitor the valve movement, a piezoelectric sensor is provided in the injection valve. This piezoelectric sensor is designed as a pressure sensor to detect changes in the fuel pressure triggered by actuation of the injection valve and to provide a corresponding sensor signal. The measured sensor signal can then be evaluated to determine the actual opening and closing times of the injection valve in order to adapt the actuation of the injection valve accordingly.

[0014] However, the sensor signal is also noisy and depends, inter alia, on the actual fuel pressure in the fuel supply and the duration of the actuation to be measured.

[0015] The sensor signals can be evaluated with the aid of a data-based evaluation model to determine the opening and / or closing times of the injection valves. The data-based evaluation model can correspond to or include a probabilistic regression model, a neural network, or a classification model.

[0016] To train the evaluation model, the time series of sensor signals can be detected. On a test bench, the actual opening and closing times of the respective injection valves can be detected for the combustion engine to be measured. This allows the generation of a training data set that maps a time series of a predetermined number of sampling times to a description of the opening and / or closing times.

[0017] However, this method of action is not robust and can result in high estimation uncertainties for the opening and / or closing times, particularly in the event of variations in the opening and / or closing movement of the injection valve or in the event of unusual pressure fluctuations during the opening period of the injection valve.

[0018] The resolution of approximately 5 μs to 20 μs is required to describe the opening and / or closing times in order to enable sufficiently precise monitoring of the injected fuel quantity. The sensor values ​​must be detected within a time window of between 200 μs and 1 ms, which includes the time periods for opening and closing the injection valve.

[0019] Typically, the sensor signal can be sampled at a significantly higher sampling rate, such as, for example, a sampling rate between 0.1 μs and 5 μs. For sufficiently robust evaluation of the profile of the sensor signal to determine the opening and closing times, a number of between 40 and 100 sampling values ​​(samples) is sufficient, each providing an evaluation point for a specific evaluation time of the evaluation point time series.

[0020] Furthermore, a plurality of evaluation point time series can be generated from each sensor signal in the following manner: time windows uniformly spaced in time are defined according to the time resolution of the evaluation point time series, each of which includes a plurality of consecutive sensor signal values, wherein for each time window one of the sensor signal values ​​therein is randomly selected as an evaluation point, so that for one sensor signal a plurality of different evaluation point time series are generated with evaluation points of consecutive time windows.

[0021] In particular, the random selection of the sensor signal values ​​may be performed according to a uniform distribution or a Gaussian distribution.

[0022] It is proposed here to increase the number of training data sets used to train the evaluation model. In order to create an evaluable evaluation point time series from the sensor signal, the sensor signal can be oversampled with respect to the time resolution required for the evaluation point time series. Due to this oversampling, it is possible to consider multiple sampling values ​​of the sensor signal for each evaluation moment of the evaluation point time series to be evaluated. This is used to generate multiple evaluation point time series with lower time resolution from the sensor signal time series with corresponding labels (that is, information about the opening and / or closing moments), and the multiple evaluation point time series are assigned to the corresponding labels. This can significantly increase the number of training data sets, thereby making it possible to train the data-based evaluation model in an improved manner. In addition, the training data sets generated in this way enable a more robust creation of the data-based evaluation model.

[0023] In particular, the evaluation moments define time windows that are continuous in time or evenly spaced in time. The sensor signal values ​​that are respectively in the time windows can be taken into account to determine the evaluation points assigned to the time windows or the corresponding evaluation moments. It is suggested that one of the sensor signal values ​​is randomly selected from each time window of the evaluation point. According to a uniform distribution, a Gaussian distribution or another predetermined distribution, a sensor signal value can be randomly selected from each time window in the time window defined by the evaluation moment. That is, the probability of selecting a sensor signal value within a time window is given by a distribution function. Since different evaluation point time series are always obtained by executing this action method multiple times, a large number of different evaluation point time series can be determined from the sensor signal time series in this way. By assigning the evaluation point time series to the basic labels assigned to the sensor signal time series, a training data set can be created respectively.

[0024] According to another aspect, a method for operating an injection system, in particular for operating an injection system in conjunction with determining an opening time and / or a closing time of an injection valve, is provided; the method comprises the following steps:

[0025] - sampling the sensor signal at a sampling rate to obtain a sensor signal time series having sensor signal values;

[0026] - determining an evaluation point time series from the sensor signal time series by dividing the sensor signal time series into evenly spaced time windows, and determining respective evaluation points of the evaluation point time series based on sensor signal values ​​within the respective time windows;

[0027] - determining an opening time and / or a closing time based on the determined time series of evaluation points and a data-based evaluation model, in particular an evaluation model trained according to one of the methods according to any one of claims 1 to 4, wherein the data-based evaluation model is trained in order to assign an opening time or a closing time to the time series of evaluation points;

[0028] - operating the injection system with the determined opening and / or closing times.

[0029] Furthermore, the injection valve can be operated in such a way that the opening duration of the injection valve, which is determined by the determined opening and / or closing time, is set to a predefined desired opening duration.

[0030] According to another aspect, a device is provided for carrying out the above-mentioned method. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Subsequently, embodiments are explained in more detail with reference to the attached drawings.

[0032] Figure 1 shows a schematic diagram of an injection system for injecting fuel into cylinders of a combustion engine;

[0033] Figure 2 A graphical illustration of a method for determining a training data set for training a data-based estimation model for determining an injected fuel quantity is shown;

[0034] Figure 3 A schematic diagram showing the sampling instants of the sensor signals and the positions of the time windows with respect to the evaluation instants of the time series of evaluation points; and

[0035] Figure 4 A flow chart is shown for illustrating a method for determining an injected fuel quantity using a trained data-based estimation model. DETAILED DESCRIPTION

[0036] Figure 1 The layout of an injection system 1 for a combustion engine 2 of a motor vehicle is shown, for which combustion engine 2 one cylinder 3 (particularly of a plurality of cylinders) is shown as an example. The combustion engine 2 is preferably designed as a diesel engine with direct injection, but can also be provided as a gasoline engine.

[0037] The cylinder 3 has an inlet valve 4 and an outlet valve 5 for supplying fresh air and for discharging combustion exhaust gases.

[0038] Furthermore, to operate the combustion engine 2, fuel is injected into the combustion chamber 7 of the cylinder 3 via the injection valve 6. For this purpose, the fuel is supplied to the injection valve via a fuel supply device 8, via which the fuel is provided at a high fuel pressure in a manner known per se (e.g., a common rail).

[0039] The injection valve 6 has an electromagnetically or piezoelectrically actuated actuator unit 61, which is coupled to a valve needle 62. In the closed state of the injection valve 6, the valve needle 62 rests on a needle valve seat 63. By actuating the actuator unit 61, the valve needle 62 moves in the longitudinal direction and releases the valve opening in the needle valve seat 63. part so that the pressurized fuel is injected into the combustion chamber 7 of the cylinder 3.

[0040] Injection valve 6 further comprises a piezoelectric sensor 65 arranged in injection valve 6. Piezoelectric sensor 65 is deformed by pressure changes in the fuel conducted through injection valve 6 and generates a sensor signal in the form of a voltage signal.

[0041] The injection is controlled by the control unit 10, which predetermines the fuel quantity to be injected by energizing the actuator unit 61. The sensor signal is temporally sampled by means of an A / D converter 11 in the control unit 10, in particular at a sampling rate of 0.5 MHz to 5 MHz.

[0042] During operation of combustion engine 2, sensor signals are used to determine the correct opening and / or closing times of injection valve 6. To this end, the sensor signals are digitized by means of an A / D converter 11 into a sensor signal time series and evaluated using a suitable evaluation model. This allows the opening duration of injection valve 6 and, correspondingly, the injected fuel quantity to be determined based on the fuel pressure and other operating variables. In particular, the opening and closing times are required to determine the opening duration, so that the opening duration can be determined as the time difference between these variables.

[0043] From the observation of the sensor signal profile, the opening and / or closing time can be determined. In particular, the opening and / or closing time can be determined with the aid of a data-based evaluation model.

[0044] Data-based models require training datasets to create these models, where the robustness and accuracy of the models greatly depend on the quality of the training data.

[0045] The training data set is determined on a test bench, in which the operation of the injection valve 6 is measured in the operating combustion engine 2. Figure 2 A flowchart is used to describe the method for determining the training dataset.

[0046] In particular, in step S1, operating variables, the temporal course of the sensor signal of the piezoelectric sensor and the precise opening or closing time are detected as tags for one or more working cycles. The sensor signal is detected by sampling, wherein the sampling frequency is between 0.5 MHz and 5 MHz. The sampling of the sensor signal is preferably carried out in an evaluation time period in which the injection valve 6 is opened and closed once. The evaluation time period can be specified with respect to an actuation time window of the injection valve, which is defined by the start of the actuation of the actuator unit 61 and a specified duration, the specified duration corresponding to the valve opening predetermined by the actuation signal of the actuator unit 61. The control time window thus has a defined time basis, for which a time series of evaluation points is provided: the time series of evaluation points is the basis for further determination of the opening or closing time.

[0047] To determine a training dataset from this sensor signal time series, in step S2, the sensor signal time series is divided into time windows evenly spaced apart in time in a processing unit of the test bench. The time length of these time windows is greater than the time step corresponding to the sampling rate of the sensor signal. In particular, the time windows are separated in time by an evaluation time corresponding to the temporal resolution of the evaluation points processed as an evaluation point time series in the data-based evaluation model. The evaluation time can be between 5 μs and 20 μs.

[0048] The time windows can be immediately adjacent to one another or spaced apart in time. The time windows are selected to be large enough to always include a plurality of sampling instants of the sensor signal.

[0049] This is Figure 3 is shown by way of example in the diagram of FIG. A series of sampled values ​​A can be seen, each of which is assigned a sensor signal value W of the sensor signal S. The sampled values ​​A are evenly spaced in time, for example, at intervals of 1 μs, and form a sensor signal time series. Successive time windows F are provided, which in the illustrated embodiment each include four sampled values ​​of the sensor signal S. The time windows F have a corresponding interval of 8 μs, with the sampled values ​​being detected at intervals of 1 μs.

[0050] To determine the training data set, in step S3, sample values ​​A are randomly selected from a time window F: these sample values ​​A correspond to sensor signal values ​​W and to evaluation points of the evaluation point time series to be created. The random selection of sample values ​​A within the time window F can be performed based on a uniform distribution of the sample values ​​A contained in the time window F, or based on a Gaussian distribution or other distribution. In other words, the random selection is performed based on the probability of the distribution of the sample values ​​within the time window, which can correspond to a uniform distribution, a Gaussian distribution, or other distribution. In the case of a uniform distribution, this means that each sample value within the time window is selected with equal probability. In the case of a Gaussian distribution, this means that sample values ​​in the middle of the time window are selected with a higher probability than sample values ​​at the edges of the time window.

[0051] In this way, a time series of evaluation moments that are different from one another is obtained.

[0052] In step S4, all evaluation point time series created from the sensor signal time series can be assigned the same label to form a training data set. These labels can correspond to the injection duration of the fuel measured on the test bench, and can correspond to the opening time and / or the closing time.

[0053] Therefore, in step S5 , a data-based estimation model is created based on the training data set and implemented in the control unit 10 of the combustion engine 2 .

[0054] Figure 4 A flow chart is shown for illustrating a method for operating the injection system 1 , in particular for operating the injection system 1 in conjunction with determining an injection duration, an opening time, and / or a closing time of an injection valve.

[0055] In step S11 , the sensor signal of the piezoelectric sensor 65 is sampled for this purpose at a high sampling rate.

[0056] In step S12 , the sensor signal is divided into time windows F, which each have a spacing from one another that corresponds to a predefined evaluation rate.

[0057] In step S13 , the sampled values ​​of the sensor signal within the time window (sensor signal time series) are now averaged or weighted according to a predefined distribution and averaged in order to obtain an evaluation point.

[0058] In step S14 , the evaluation points are supplied as a time series of evaluation points to a previously trained data-based evaluation model in order to determine the opening and / or closing times therefrom.

[0059] In step S15 , the opening and / or closing time is now used to determine the opening duration of the injection valve and, correspondingly, the injected fuel quantity.

[0060] In particular, the determined injection quantity can be used in a manner known per se to regulate the engine torque, to improve the smooth running of the combustion engine, to improve its efficiency and to carry out a method for exhaust gas aftertreatment.

Claims

1. A method for training a data-based evaluation model for determining an opening or closing time of an injection valve (6) based on a sensor signal (S), the method comprising the following steps: - measuring (S1) the operation of the injection valve (6) in order to determine at least one sensor signal (S) and the associated opening or closing time; - sampling the sensor signal (S) at a sampling rate in order to obtain a sensor signal time series having sensor signal values; - determining a plurality of training data sets by assigning a plurality of evaluation point time series generated from a sensor signal time series to opening or closing moments associated with the sensor signal, wherein the evaluation point time series has a lower time resolution than the sensor signal time series; - Based on the determined training data set, training (S5) the data-based evaluation model.

2. The method according to claim 1, wherein The multiple evaluation point time series are generated from each sensor signal (S) in the following manner: time windows (F) that are uniformly spaced in time are defined according to the time resolution of the evaluation point time series, and the time windows (F) respectively include multiple continuous sensor signal values, wherein for each time window (F), one of the sensor signal values ​​therein is randomly selected as an evaluation point, so that multiple different evaluation point time series of the evaluation points with continuous time windows (F) are generated for one sensor signal (S).

3. The method according to claim 2, wherein: The random selection of the sensor signal values ​​is performed according to a uniform distribution or a Gaussian distribution.

4. The method according to any one of claims 1 to 3, wherein The data-based evaluation model corresponds to or includes a probabilistic regression model, a neural network or a classification model.

5. The method according to any one of claims 1 to 3, wherein The injection valve (6) is operated such that an opening duration of the injection valve (6) determined by the determined opening and / or closing time is set to a predetermined desired opening duration.

6. A method for operating an injection system in conjunction with determining an opening or closing time of an injection valve, the method comprising the following steps: - sampling (S11) the sensor signal (S) at a sampling rate in order to obtain a sensor signal time series having sensor signal values; - determining an evaluation point time series from the sensor signal time series by dividing the sensor signal time series into evenly spaced time windows (F) and determining respective evaluation points of the evaluation point time series based on the sensor signal values ​​within the respective time windows; - determining the opening time or the closing time based on the determined evaluation point time sequence and a data-based evaluation model, wherein the data-based evaluation model is trained to assign the opening time or the closing time to the evaluation point time sequence; - operating ( S15 ) the injection system with the determined opening time or closing time.

7. The method according to claim 6, wherein: The evaluation model is trained according to the method according to any one of claims 1 to 4.

8. The method according to claim 6 or 7, wherein: The injection valve (6) is operated such that an opening duration of the injection valve (6) determined by the determined opening and / or closing time is set to a predetermined desired opening duration.

9. An apparatus for performing the method according to any one of claims 1 to 8. 10 . A computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method according to claim 1 . 11 . A machine-readable storage medium comprising instructions, which, when executed by a computer, cause the computer to execute the steps of the method according to claim 1 .

Images