METHOD AND SYSTEM FOR CALIBRATING A CONTROLLER OF AN ENGINE

Abstract

The invention relates to a method for the operational analysis of an engine and/or for calibrating a controller of the engine, in particular an internal combustion engine, wherein run-up occurs of test points defined by values of a plurality of predetermined operating parameters and selected from a multidimensional test space using a statistical experiment design, whereby at least one operating parameter is in each case changed from one test point to the next test point in a plurality of steps in the run-up of the test points, wherein operational measurements are performed at measurement points resulting from a respective increment and at the actual test points, whereby measurement data from the operational measurements for the analysis and calibration of the controller are output and continuously stored, as well as a corresponding system.

Claims

1. A method for the operational analysis of an engine and/or for calibrating a controller of the engine, in particular an internal combustion engine, wherein run-up occurs of test points defined by values of a plurality of predetermined operating parameters and selected from a multidimensional test space using a statistical experiment design, whereby a plurality of operating parameters are in each case changed from one test point to the next test point in a plurality of steps in the run-up of the test points by means of a slow dynamic slope method, wherein operational measurements are performed at measurement points resulting from a respective increment and at the actual test points so that the test points are linked by measurement ramps, whereby measurement data from the operational measurements for the analysis and calibration of the controller are output and continuously stored, wherein the measurement data is supplied to a modeling algorithm, whereby a model is adapted by means of the modeling algorithm during execution of the method, wherein test points which have not yet been measured and/or additional test points are positioned using the adapted model.

2. The method according to claim 1, wherein operational measurements are performed continuously, whereby preferably no separate stabilization phase in which the at least one operating parameter is kept constant is provided prior to a measurement phase in which operational measurements are made.

3. The method according to claim 1, wherein the temporal course of changing the at least one operating parameter in steps is ramped.

4. The method according to claim 1, wherein upon the exceeding of an operating limit value of an operating parameter observed by the operational measurement, run-up of the next test point is aborted and a previous test point, a predefined safe operating point or a test point subsequent to the next test point is obtained.

5. (canceled)

6. (canceled)

7. (canceled)

8. The method according to claim 1, wherein signal profiles of the measurement data are corrected by measurement channel-specific delay times during execution of the method.

9. The method according to claim 1, wherein a model formed or modified by means of the modeling algorithm is used to calculate an optimized calibration of the controller.

10. The method according to claim 9, wherein the model or the optimized calibration of the controller is verified using further operational measurements.

11. The method according to claim 1, wherein the measurement data is usable for different calibration tasks, whereby different input and/or output variables are selected from among the available recorded measurement data for each calibration task.

12. A computer program containing instructions which, when executed by a computer, prompts it to execute the steps of a method according to claim 1.

13. A computer-readable medium on which a computer program according to claim 12 is stored.

14. A system for the operational analysis of an engine and/or for calibrating a controller of the engine, in particular an internal combustion engine, comprising: a test bench for the run-up of test points defined by values of a plurality of predetermined operating parameters and selected from a multidimensional test space by means of a statistical experiment design, means for realizing run-up of the test points, configured to change a plurality of operating parameter in each case from one test point to the next test point in a plurality of steps using a slow dynamic slope method; sensors for realizing operational measurements at measurement points resulting from a respective increment and at the actual test points so that the test points are linked by measurement ramps; and a data interface for outputting measurement data from the operational measurements, on the basis of which the engine is analyzed and the controller calibrated; and a data storage configured for the continuous storage of the measurement data, wherein the system is configured to supply the measurement data to a statistical or mathematical evaluation method or a modeling algorithm, in particular an artificial neural network, adapt a model by means of the modeling algorithm and/or a test space during execution of the method, and position test points which have not yet been measured and/or additional test points using the adapted model with definable target corridors for output variables which are particularly optimally defined with regard to the desired optimization objective.

Description

[0049] Further advantages and features will become apparent from the following description of exemplary embodiments referencing the figures. Shown therein at least partially schematically:

[0050] FIG. 1 a block diagram of an exemplary embodiment of a method for calibrating a controller of an engine;

[0051] FIG. 2 a depiction of an exemplary embodiment of a system for calibrating a controller of an engine;

[0052] FIGS. 3a and 3b diagrams of the progression of respective input parameters/input variables and output parameters/output variables of operational measurements in a calibration method of the prior art;

[0053] FIG. 4 a further diagram of input parameters/input variables and output parameters/output variables of operational measurements performed by means of the method for calibrating a controller of an engine;

[0054] FIG. 5 various diagrams with measurement profiles recorded using the method for calibrating a controller of an engine; and

[0055] FIG. 6 a diagram pursuant to FIG. 4, wherein different delay times are shown for output parameter(s)/output variable(s) determined by means of a model and measured output parameter(s)/output variable(s).

[0056] FIGS. 7a and 7b diagrams of the temporal course of an example input parameter and the output variables resulting therefrom at an assumed delay/dead time as compensated by the method by the shifting of the faster input and output variables.

[0057] The following will describe the exemplary embodiments in respect of an internal combustion engine 1. It is however obvious to one skilled in the art that the described teaching is also applicable to other engines, in particular prime movers such as electric motors.

[0058] FIG. 1 shows an exemplary embodiment of a method 100 for calibrating a controller of an engine 1.

[0059] The workflow into which the method 100 for calibrating a controller is integrated is preferably as follows:

[0060] First, the respective test or experiment to be performed, particularly a test run, is defined. Preferably specified is what input parameters are to be set or changed and which output parameters are to be recorded. The environmental conditions under which the test is to be carried out are specified, e.g. the temperature of the coolant. Also preferably specified are the output parameters needing to be monitored in relation to operating limits Lim and what the limit values Lim are. Based on this, adjustment ranges for the input parameters P.sub.1 are specified and a space-filling design furthermore preferably generated using this key data.

[0061] Based on the test definition and the space-filling design, an experiment design is preferably created using a statistical design of experiments.

[0062] This test plan comprises a plurality of test points. Test points are thereby defined by a plurality of operating parameters P, or their values respectively, and the test space is also a multidimensional test space due to the plurality of operating parameters.

[0063] The test points are usually measured in test bench operation, for example on an engine test bench, a powertrain test bench or a roller test bench. Generally speaking, so-called stationary test benches are used to this end.

[0064] The test points to be approached are set. Should the test point be reached, a wait ensues until the operation of the internal combustion engine has stabilized and upon there being only minor or no changes in the operation of the internal combustion engine during this stabilization phase STAB, a measurement is taken in measurement phase MEAS. A general sequence of such is depicted in FIG. 3a.

[0065] The calibration method 100 preferably starts at this point.

[0066] In method 100, similar to calibration methods in the prior art, run-up of the test points 101 ensues. At least one operating parameter P.sub.1, which is a regulating parameter, is thereby changed in a plurality of steps from one test point T.sub.n to the next test point T.sub.n+1.

[0067] In practice, multiple operating parameters P.sub.1 are usually adjusted either simultaneously or successively. The adjustment is thereby preferably made so slowly and/or at such small increments that the internal combustion engine 1 is in a quasi-stationary operating mode.

[0068] Operational measurements are thereby performed both on those value constellations ensuing from the increment set for each operating parameter P after each step, hereinafter referred to as measurement points M.sub.n, as well as at the actual test points T.sub.n, T.sub.n+1, T.sub.n+2 ensuing from the selection via the statistical design of experiments.

[0069] Since the internal combustion engine 1 is kept in a quasi-stationary operating mode during the input parameter P.sub.1 adjustment process, stabilization periods STAB are preferably not provided either before or after measuring the measurement points M.sub.n nor before or after measuring the test points T.sub.n, T.sub.n+1, T.sub.n+2, as is the case in the prior art. The setting phase SET can therefore also be included in the measurement phase MEAS.

[0070] The measurement phase MEAS of the method 100 can thus extend over a considerably longer period of time than in the conventional calibration methods of the prior art. Preferably, the measurement phase MEAS lasts the entire duration of the method 100, further preferably without interruption.

[0071] Lastly, the measurement data from the operational measurements, which can further be used to analyze and calibrate the controller, are output and continuously stored 103.

[0072] Due to the plurality of measurements, or dense measurement data respectively, generated by the method 100, the measurement data can be used not only for the calibration task for which it was collected. It can in fact also be used for other calibration tasks, provided that the operating variables required thereto have also been measured.

[0073] The stored measurement data is preferably fed to statistical and/or mathematical evaluation methods 104. In particular, a modeling ensues using these methods. Preferably being a model of the internal combustion engine, the powertrain or the entire vehicle with which the respectively modeled component can be simulated. Preferably, the model is a so-called artificial neural network which is trained on the basis of the measurement data. However, other machine learning methods can also be used, as can polynomial models or Gaussian models, etc. Further preferably, signal delay time up until the measurement is also factored into the modeling. This is explained further below with reference to FIG. 6. The models can be preferential or selected from a library of existing example models in which relationships between the input parameters P.sub.1 and output parameters P.sub.2 are roughly established.

[0074] Preferably, these models are continually or continuously verified and are adapted 105 while the method 100 is being carried out.

[0075] On the basis of these models, a numerical optimization can be made in order to improve the calibration of the controller. Preferably, such an optimized calibration can already be regenerated while the inventive method 100 is being performed by repeating the statistical design of experiments or by already factoring the knowledge of the optimized calibration into the statistical design of experiments 106 respectively.

[0076] Further preferably, an optimized calibration is lastly verified via further operational measurements 107.

[0077] FIG. 2 shows an exemplary embodiment of a system for calibrating a controller of an internal combustion engine 1.

[0078] The internal combustion engine 1 is thereby preferably arranged on a test bench 11 and further preferably non-rotatably connected to a dynamometer 3 via a shaft 4 which is part of the internal combustion engine 1 or the test bench 11.

[0079] The dynamometer 3 is preferably configured to apply a load to the internal combustion engine 1. Furthermore, the test bench 11 comprises sensors 13a, 13b, 13c to record operating variables of the internal combustion engine 1. In the present case, for example, sensor 13a could record the throttle valve position, sensor 13b a torque applied to the shaft 4 and thus to the internal combustion engine 1, and sensor 13c the power expended to brake the shaft 4 and thus the internal combustion engine 1 via the dynamometer 3.

[0080] Measured values are preferably output via a data interface 14 of the system 10 for outputting measurement data from the operational measurements to a data storage 15 or directly to means 12 for the run-up of the test points T.sub.n, T.sub.n+1, T.sub.n+2.

[0081] The means 12 for the run-up of the test points calculates the values of the input parameters P.sub.1 on the basis of a statistical experiment design and/or the measurement data already generated and relays them to a controller 2 of the internal combustion engine 1. The controller sets the input parameters P.sub.1 as internal combustion engine 1 setpoint values.

[0082] FIGS. 3a and 3b show two diagrams of the measurement profile of a conventional method for calibration.

[0083] The diagram of FIG. 3a thereby depicts the progression of the input parameter P.sub.1 of the controller of an internal combustion engine; i.e. the set variable, and the associated progression of the output parameter P.sub.2; i.e. the observed variable, as a function of time. FIG. 3b again shows a diagram of the progression of the output parameter P.sub.2 as a function of the input parameter P.sub.1, or the set variable respectively, in that region in which a limit violation of the output parameter P.sub.2 occurs.

[0084] In FIG. 3a, after test point T.sub.n has been measured, the input parameter P.sub.1 is incrementally brought to the value of subsequent test point T.sub.n+1. This phase of the measurement process is a setting phase SET in which the input parameter P.sub.1 or multiple input parameters respectively are set to the next test point T.sub.n+1 as determined using a statistical experiment design.

[0085] In the illustrated case, a limit violation of the operating limit Lim of output parameter P.sub.2 occurs at t.sub.1. As a result, the input parameter P.sub.1 cannot be set to controlled test point T.sub.n+1.

[0086] The input parameter P.sub.1 is consequently reset until the limit violation of output parameter P.sub.2 is eliminated. Due to hysteresis or an output parameter lag time, this takes some time, as is evident from FIG. 3a, and the input parameter P.sub.1 has to drop back far below the value at which the limit violation occurred.

[0087] In a next step of the method, the input parameter P.sub.1 is again adjusted in each direction in which the limit violation occurred, albeit at a reduced increment. If no limit violation then occurs, a stabilization phase STAB follows at time t.sub.2 at which point there is a wait until the engine is in a stationary mode of operation and only a minor or even no change in the output parameter P.sub.2 at all can be observed. Once this is the case, a phase of the measurement process in which a measurement MEAS is made follows from time t.sub.3 to time t.sub.4. An auxiliary test point T.sub.n+1′ is thereby measured.

[0088] As shown in FIG. 3b, only one single measurement is made of an auxiliary test point T.sub.n+1′ in a region where the output parameter P.sub.2 exhibits strong fluctuation, or a sharp gradient respectively, in relation to a change in the input parameter P1 . Yet a plurality of measurement points would be advantageous particularly in a region of major changes so as to be able to precisely analyze the dependency of the output parameter P.sub.2 on the input parameter P1 and be able to accordingly factor that into the modeling. This is also of particular importance because optimality when calibrating internal combustion engines often lies close to operating limits Lim.

[0089] FIG. 4 shows a diagram of a measurement profile of calibration method 100 for comparison.

[0090] Here as well, after measuring a test point T.sub.n selected using a statistical experiment design, a subsequent test point T.sub.n+1 is approached by adjusting the at least one input parameter P.sub.1. In contrast to the conventional measurement procedure, however, smaller increments are realized in method 100 and/or the run-up of subsequent test point T.sub.n+1 occurs at a lower input parameter P.sub.1 adjustment speed. The internal combustion engine 1 is thereby kept in a quasi-stationary operating mode.

[0091] Operational measurements can thus not only be performed at test points T.sub.n, T.sub.n+1 in the method 100 but also at measurement points M.sub.n resulting from an increment of the input parameter P.sub.1 adjustment.

[0092] In FIG. 4, these measurement points M.sub.n are depicted by crosses on the line connecting test point T.sub.n to the next test point T.sub.n+1. Preferably, a measurement point M.sub.n is thereby recorded after each adjustment step of input parameter P.sub.1. It is however also possible to perform operational measurements only after a number of steps.

[0093] As in the case of the conventional measurement method according to FIG. 3a, the output parameter P.sub.2 reaches an operating limit Lim at time t.sub.1 in the depicted case. Contrary to the conventional measurement method, however, no attempt is made in method 100 to reverse the input parameter P.sub.1 and then re-approach the operating limit Lim of the output parameter P.sub.2 in the original adjustment direction of the input parameter P.sub.1. Instead, the method 100 moves to the next test point T.sub.n+2, which is likewise determined via the statistical experiment design. Here, too, operational measurements are made at measurement points M.sub.n after each adjustment step of the input parameter P.sub.1 or after multiple adjustment steps. Alternatively, after the limit value violation at t.sub.1, run-up can occur to the previous test point T.sub.n or even a predefined safe operating point, i.e. an operating point known to be within the obtainable range of the test space.

[0094] The totality of the recorded measurement points M.sub.n in FIG. 4 is identified using curly brackets. As likewise evident from FIG. 4 is that the measurement period MEAS, which preferably lasts the entire measurement process, is considerably longer than in the conventional approach in FIG. 3a.

[0095] Substantially, measurements can be performed during the entire input parameter P.sub.1 adjustment process in method 100. This results in considerably higher information density in measurement phase MEAS. The measurement phase MEAS preferably lasts from the point of leaving a test point T.sub.n to reaching the operating limit Lim and in turn approaching a subsequent test point T.sub.n+1, T.sub.n+2 and, as can be seen from comparing the diagrams of FIG.

[0096] 3a and FIG. 4, is thereby even shorter than a single measurement of the alternate test point T.sub.n+1′in FIG. 3a.

[0097] FIG. 5 shows a series of diagrams, each depicting the two-dimensional test space respectively defined by the operating variables of throttle valve position, waste gate position, intake camshaft position and exhaust camshaft position indicated as input parameters along the abscissa and the identical operating variables of exhaust camshaft position, intake camshaft position, waste gate position and throttle valve position indicated as output parameters along the ordinate. Here as well, the measurements were conducted as part of method 100.

[0098] It is clear that the measurements provide very good coverage of the test spaces. The measurement density is considerably higher here than had only the few test points selected via a statistical experiment design been measured.

[0099] Because of the density of the operational measurements, the pairs of measured operational parameters can be used not only for a single calibration task but the recorded operational measurements can also be used for other calibration tasks in which the test spaces are preferably more or less the same as the test spaces shown in FIG. 5.

[0100] A consideration of a signal lag time in the modeling is described with reference to FIG. 6. This relates to dealing with time-delayed signals in method 100. As is known, many real measurement signals have a delay time, e.g., due to line lengths to the measuring sensor for a measuring medium (such as with emission measuring devices) or due to time lags (such as with temperature measurement points). Particularly real output and emission measurement signals are measured with a certain delay. When this measurement data is to be treated as (quasi) stationary, the respective time delay relative to these signals needs to be factored into the modeling.

[0101] Typical delay times for e.g., emission measuring devices are in the range of a few seconds.

[0102] Within the scope of the exemplary embodiment of method 100 as described, preferably the modeling algorithm decides, particularly in post-processing, which lag time is the most appropriate. Various output parameter models, e.g., emissions having an identical model structure but different transmission signal lag times, are thus generated via a modeling algorithm.

[0103] To that end, each vector is preferably shifted by three different lag times of x, y and z seconds and the empirical model algorithms then test which delay is the most appropriate and therefore exhibits the best model quality R.sup.2. This is then used for the final data evaluation and any potential optimization necessary. That means that the model having a simulated lag time which best matches the measured lag time, i.e., exhibits the best model quality R.sup.2, is selected as the model for the optimization.

[0104] In FIG. 6, this is the model incorporating a 4s lag time for the output parameter.

[0105] In a further advantageous embodiment of the method 100, measurement data signal profiles over a test run during which the method 100 is executed are corrected by measurement channel-specific delay times. These delay times can result in particular from signal propagation delays or control paths. Such measurement channel-specific delay times in collected raw data are depicted in FIG. 7a for an input, an output A and an output B. All the signals are thereby synchronized to the signal with the longest delay time. Such a correction for the input and output A is depicted in FIG. 7b. In the illustrated case, the input, in particular a signal profile of an input variable, is shifted by 4.6s and the output A, in particular a signal profile of an output variable A, is shifted by 4s. This is indicated by the arrows. The temporal relationships between the signals can thereby be restored. A change in the value of the input in the form of a theta function induces a signal response (step response) of outputs A and B. This thereby further improves the quality and efficiency of the modeling algorithms, particularly the continuous model optimization.

[0106] In a further advantageous embodiment of the method 100, the delay times of the measurement channels are determined individually prior to the start of the measuring program by adjusting the input parameter(s) and analyzing the signals, particularly of the output parameters, associated with the delay, for instance by evaluating step responses as shown in FIG. 7a and FIG. 7b, for example. This can ensue, preferably in automated fashion, either in advance or in the course of the method 100.

[0107] The above-described exemplary embodiments are merely examples which are in no way to be limiting of the scope of protection, the application or the configuration. Rather, the preceding description affords one skilled in the art a guideline for the implementation of at least one exemplary embodiment, whereby various modifications can be made, in particular with regard to the function and arrangement of the described components, without departing from the protective scope resulting from the claims and equivalent combinations of features. In particular, individual exemplary embodiments may be combined with one another.

LIST OF REFERENCE NUMERALS

[0108] 1 engine

[0109] 2 controller

[0110] 3 dynamometer

[0111] 4 shaft

[0112] 10 system

[0113] 11 test bench

[0114] 12 means for test point run-up

[0115] 13a, 13b, 13c sensors

[0116] 14 data interface

[0117] 15 data storage

[0118] P.sub.1 input parameter, input variable

[0119] P.sub.2 output parameter, output variable

[0120] T.sub.n, T.sub.n+1, T.sub.n+2 test point

[0121] T.sub.n+1′auxiliary test point

[0122] t.sub.1, t.sub.2, t.sub.3, t.sub.4 time

[0123] M.sub.n measurement point

[0124] Lim operating limit