Method for model-based control and regulation of an internal combustion engine

Abstract

The invention proposes a method for model-based control and regulation of an internal combustion engine, wherein in a first step, a pre-optimized quality measure (J(VO)) is calculated by an optimizer as a function of an operating situation (BS), wherein in the calculation of the pre-optimized quality measure (J(VO)), a discrete control parameter with discrete adjustment values ​​is interpreted as a continuous control parameter (SG(k)) with a continuous adjustment range, wherein in a second step, the continuous control parameter (SG(k)) is quantified and set as a new discrete control parameter (SG(new)) with discrete adjustment values, wherein in a third step, the optimizer calculates a re-optimized quality measure (J(NA)) as a function of the new discrete adjustment parameter (SG(new)) and the operating situation (BS) of the internal combustion engine (1), and the optimizer (21) sets the re-optimized quality measure (J(NA)) as decisive for the operating point of the internal combustion engine (1).

Description

Technical Field

[0001] The invention relates to a method for model-based control and regulation of an internal combustion engine, wherein a quality measure is calculated by an optimizer and is set so as to be decisive for an operating point of the internal combustion engine. Background Art

[0002] The characteristics of an internal combustion engine are determined decisively by the engine control unit depending on the performance expectations. For this purpose, corresponding characteristic lines and characteristic ranges are usually used in the software of the engine control unit. The adjustment parameters of the internal combustion engine, such as the injection start and the required rail pressure, are calculated from the performance expectations using the characteristic lines and characteristic ranges. The characteristic lines / characteristic ranges are provided with data on the test bench at the internal combustion engine manufacturer. However, the large number of characteristic lines / characteristic ranges and their interrelationships lead to a high level of tuning effort.

[0003] Therefore, in practice, attempts are made to reduce the tuning effort by using mathematical models. Thus, for example, DE 10 2006 004 516 B3 describes a Bayesian network with a probability table for determining the injection quantity, and US 2011 / 0172897 A1 describes a method for adapting the injection start and the injection quantity by means of a combustion model with the aid of a neural network. Since trained data are depicted here, they must first be learned during test bench operation.

[0004] A method for model-based control and regulation of an internal combustion engine is known from DE 10 2017 005 783 A1, wherein a theoretical value for an injection system adjustment link is calculated by a combustion model and a theoretical value for an air path adjustment link is calculated by a gas path model. Both the combustion model and the air path model are based on a Gaussian process model. The optimizer determines a quality measure from the theoretical value and predicts how the quality measure will develop within the prediction limit when the theoretical value changes. If the most feasible quality measure is calculated, the optimizer sets the injection system theoretical value and the air path theoretical value to be decisive for the operating point of the internal combustion engine.

[0005] Test bench tests have shown that the integration of control parameters with discrete switching states into the model-based methods described above is not yet satisfactory. Control parameters with discrete switching states are to be understood, for example, as: the switching on of the second exhaust gas turbocharger in the case of staged charging, cylinder edge closing, activation of pre-injection or re-injection, and the opening or closing position of various valves. The so-called branch-and-bound method for optimal solution with discrete control parameters is very computationally intensive, since in the worst case all possible combinations of discrete control parameters must be investigated. Its use in internal combustion engines quickly leads to a very complex structure, which cannot be presented on the engine control device. Summary of the invention

[0006] The invention is based on the object of improving the above-described model-based method with regard to the combination of adjustment parameters.

[0007] This object is achieved by the features of claim 1. Developments are presented in the dependent claims.

[0008] The method is implemented in three steps. In the first step, the optimizer calculates the quality scale optimized in advance depending on the operating conditions, wherein the discrete adjustment parameters with discrete adjustment values ​​are interpreted as continuous adjustment parameters with continuous adjustment ranges. The quality scale optimized in advance is a calculated parameter, that is, it is not connected to the internal combustion engine. Then, in the second step, the continuous adjustment parameter is digitized and set to a new discrete adjustment parameter with discrete adjustment values. The digitization is performed according to the switching threshold together with the hysteresis. Finally, in the third step, the optimizer then calculates the re-optimized quality scale depending on the new discrete adjustment parameters and the operating conditions of the internal combustion engine, and sets it to be decisive for the operating point of the internal combustion engine. However, when calculating the re-optimized quality scale, the new discrete adjustment parameters are assumed to be fixed and unchanged. In this regard, it no longer presents degrees of freedom for optimization within the predicted limits. The remaining continuous adjustment parameters are re-optimized as follows, so that the solution about the fixed new adjustment parameters is the most feasible.

[0009] The operating situation of the internal combustion engine is to be understood not only as external framework conditions, in particular emission limits or performance expectations, but also as the current operating point. Both the pre-optimized quality measure and the re-optimized quality measure are determined by calculating injection system target values, such as target rail pressure, for controlling the injection system calibration by means of a combustion model, calculating gas circuit target values ​​for controlling the gas circuit calibration by means of a gas circuit model and subsequently varying the target values ​​by the optimizer with the goal of minimizing them.

[0010] The invention allows the solution of optimization tasks with partially continuous and partially discrete input parameters, given limited computing power for the optimization method used. Instead of a parallel calculation of the adjustment parameters, as is necessary for implementing the branch and bound method, the invention uses a serial algorithm. Firstly, the quality measure and the values ​​for the adjustment parameters derived therefrom can be calculated completely on the motor control unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The preferred embodiment is shown in the drawings.

[0012] Figure 1 Showing the system diagram,

[0013] Figure 2 Shows a model-based system diagram,

[0014] Figure 3 Showing the block diagram,

[0015] Figure 4 Show the program flow chart,

[0016] Figure 5 Show subroutine,

[0017] Figure 6 Show subroutine,

[0018] Figure 7 Show subroutine,

[0019] Figure 8 A timeline diagram is shown, and

[0020] Fig. 9 A timeline diagram is shown. DETAILED DESCRIPTION

[0021] Figure 1The system diagram of an electronically controlled internal combustion engine 1 with a common rail system is shown. The common rail system includes the following mechanical components: a low-pressure pump 3 for delivering fuel from a fuel tank 2, a variable intake throttle 4 for influencing the throughflowing fuel volume flow, a high-pressure pump 5 for delivering fuel under increased pressure, a rail 6 for storing fuel and an injector 7 for injecting fuel into the combustion chamber of the internal combustion engine 1. Optionally, the common rail system can also be implemented with a single reservoir, in which case, for example, a single reservoir 8 is integrated in the injector 7 as an additional buffer volume. The other functional assumptions of the common rail system are known. The air path shown includes not only an air intake but also an exhaust gas removal. Arranged in the air intake are: a compressor of an exhaust gas turbocharger 11, a charge air cooler 12, a throttle valve 13, an inlet 14 for combining charge air with the returned exhaust gas, and a variably controllable inlet valve 15. Arranged in the exhaust gas duct are a variably controllable outlet valve 16 , an EGR control element 17 , a turbine of the exhaust gas turbocharger 11 and a turbine bypass valve 18 .

[0022] The operating mode of the internal combustion engine 1 is determined by an electronic control unit 10 (ECU). The electronic control unit 10 contains the usual components of a microcomputer system, such as a microprocessor, I / O modules, buffers and storage modules (EEPROM, RAM). In the storage module, operating data relevant to the operation of the internal combustion engine 1 are used as a model. Using the model, the electronic control unit 10 calculates output parameters from input parameters. Figure 1 In the example, the following input parameters are shown: the setpoint torque M (theoretical) predefined by the operator, the actual rail pressure pCR measured by means of the rail pressure sensor 9, the engine speed nactual, the charge air pressure pLL, the charge air temperature TLL, the humidity phi of the charge air, the exhaust gas temperature Tex, the air-fuel ratio λ, the NOx actual value, the pressure pES of the optional individual storage 8 and the input parameter "In" (EIN). Further sensor signals (e.g. the coolant temperature) not shown are summarized as input parameters "In". In Figure 1 In , as output parameters of the electronic control unit 10, the following are shown: the signal PWM for controlling the intake throttle 4, the signal ve (injection start / injection end) for controlling the injector 7, the adjustment signal DK for controlling the throttle valve 13, the adjustment signal VVT for controlling the inlet valve or the outlet valve, the adjustment signal AGR for controlling the AGR adjustment element 17, the adjustment signal TBP for controlling the turbine bypass valve 18, and the output parameter "Out" (AUS). The output parameter "Out" represents representatively other adjustment signals for controlling and regulating the internal combustion engine 1, for example, a adjustment signal for activating the second exhaust gas turbocharger in the case of staged supercharging. In Figure 1In the illustration of , for example, the throttle flap 13, the EGR control element 17, the turbine bypass valve 18 or the intake throttle 4 can be controlled by means of a continuous control signal and can therefore be adjusted within a continuous value range. In contrast, a discrete control parameter would be the control signal for activating the second exhaust gas turbocharger, since it can only assume individual discrete values, i.e. there are no intermediate values.

[0023] Figure 2 A system diagram based on a model is shown. In the diagram, a combustion model 19, a gas path model 20 and an optimizer 21 are cited within the electronic control device 10. Both the combustion model 19 and the gas path model 20 describe the system characteristics of the internal combustion engine as mathematical equations, for example in the form of a Gaussian process model. The combustion model 19 statically describes the process during combustion. In contrast, the gas path model 20 describes the dynamic characteristics of the air guide and the exhaust gas guide. The combustion model 19 contains individual models, which are used, for example, for NOx and soot formation, for exhaust gas temperature, for exhaust gas mass flow and for injection pressure. The individual models are in turn dependent on the boundary conditions and injection parameters in the cylinder. The combustion model 19 is determined in a reference internal combustion engine in a test bench operation, i.e., the so-called DoE test bench operation (DoE: Design of Experiments). In the DoE test bench operation, the operating parameters and the adjustment parameters are systematically changed with the following goal, that is, depending on the engine parameters and the environmental boundary conditions to describe the overall characteristics of the internal combustion engine. The optimizer 21 evaluates the combustion model 19, more precisely with respect to the theoretical torque M(theoretical), emission boundary values, environmental boundary conditions (such as the humidity phi of the charge air) and the operating conditions of the internal combustion engine. The operating conditions are defined by the engine speed nactual, the charge air temperature TLL, the charge air pressure pLL, etc. Now, the function of the optimizer 21 is to estimate the injection system theoretical values ​​for controlling the injection system adjustment link and the gas path theoretical values ​​for controlling the gas path adjustment link. Here, the optimizer 21 selects the following solution, in the case of which the quality metric is minimized. The quality metric is calculated as the integral of the square of the theoretical-actual deviation within the prediction limit. For example, with the following formula:

[0024] .

[0025] In this case, w1, w2 and w3 are the corresponding weighting factors. It is known that nitrogen oxide emissions are derived from the humidity phi of the charge air, the charge air temperature TLL, the start of injection SB and the rail pressure pCR.

[0026] The optimizer 21 calculates the most feasible quality measure by minimizing the value in the following manner: a first quality measure is calculated at a first time point, the injection system setpoint value and the gas circuit setpoint value are changed, and a second quality measure is predicted based on them within the prediction limit. Then, the optimizer 21 determines the minimum quality measure based on the deviation of the two quality measures relative to each other and sets it as being decisive for the internal combustion engine. For the example shown in the figure, this is the setpoint rail pressure pCR (SL) for the injection system. The setpoint rail pressure pCR (SL) is a reference parameter for the next-level rail pressure control loop 22. The adjustment parameter of the rail pressure control loop 22 corresponds to the PWM signal for loading the suction throttle. For the gas circuit, the optimizer 21 indirectly determines the gas circuit setpoint value. In the example shown, this is the lambda setpoint value LAM (SL) and the AGR setpoint value AGR (SL) for presetting the two next-level control loops 23 and 24. The returned measured variable MESS is read in by the electronic control unit 10. The measured variable MESS is to be understood as meaning both directly measured physical variables and auxiliary variables calculated therefrom. In the example shown, the lambda actual value LAM (actual) and the AGR actual value AGR (actual) are read in. The reference symbol SG summarizes the control parameters of the internal combustion engine. Both continuous control parameters with a continuous adjustment range and discrete control parameters with discrete adjustment values ​​fall into this category. Continuous control parameters, such as the injection start and injection end, which control the injection of the injector ( Figure 1 : 7) Direct loading. Discrete control parameters with discrete adjustment values ​​can only be adjusted in stages to fixed values, such as cylinder deactivation.

[0027] Figure 3 A block diagram is shown with the operating condition BS of the internal combustion engine as an input variable and a quality measure, here referred to as a reoptimized quality measure J(NA), as an output variable. Within the block diagram, a preoptimization 25, a digitization 26 and a reoptimization 27 are shown. In a first step, a preoptimized quality measure J(VO) is calculated by the preoptimization 25, wherein a discrete control parameter with discrete adjustment values ​​is interpreted as a continuous control parameter with a continuous adjustment range.

[0028] An example of a discrete tuning parameter is pre-injection, which can only be activated or deactivated. By using pre-injection, the peak pressure of the combustion can be significantly reduced. In addition, but in the case of activated pre-injection, all other combustion parameters such as, for example, NOx emissions or the number of particles also change. The internal combustion engine is measured once with activated pre-injection and once with deactivated pre-injection. This results in two independent combustion models. When calculating the pre-optimized quality measure J(VO), the optimizer interpolates intermediate values, which means that the state of activated or deactivated pre-injection is artificially converted into a continuous input parameter by interpolating between the two combustion models. These parameters are then applied continuously in the pre-optimization 25. Figure 3 In the embodiment of the present invention, the continuous adjustment parameter is referred to as SG(k). The pre-optimized quality measure J(VO) is a purely internal calculation parameter, which does not have access to the adjustment link of the internal combustion engine. In other words: the pre-optimized quality measure J(VO) is not accessible and is not connected to the internal combustion engine. In the second step, a new discrete adjustment parameter SG(new) is calculated from the continuous adjustment parameter SG(k) by digitization 26. Therefore, for the pre-injection, a fixed and unchanged assignment to the activated pre-injection or the deactivated pre-injection is performed in the digitization. The digitization 26 provides the following advantages, that is, for example, the variable valve control is determined with three discrete values ​​(i.e., minimum value, intermediate value and maximum value), such as crankshaft angles of 450°, 495° and 540°. This greatly reduces the computational cost when subsequently determining the re-optimized quality measure. In addition, in the digitization 26, the calculated value is stabilized by an optional hysteresis band. In a third step, the new discrete tuning parameter SG(new) is combined with the operating situation and the reoptimized quality measure J(NA) is calculated by the optimizer. When calculating the reoptimized quality measure J(NA), the new discrete tuning parameter SG(new) does not change. In this sense, it is not a degree of freedom when calculating the reoptimized quality measure J(NA). In the reoptimization, the continuous tuning parameter is actually adapted to the trend preset from the digitization, for example, of the pre-injection. In other words: in the reoptimization, the tuning parameter changes, which is also actually described by the continuous tuning parameter. The reoptimized quality measure J(NA) corresponds to the minimum quality measure J(min), which is set by the optimizer to be decisive for the operating point of the internal combustion engine (1), that is, connected to the internal combustion engine.

[0029] exist Figure 4In the method, a program flow chart is used. After initialization in S1, it is checked in S2 whether the start process is completed. If the start process is still running, the query result S2: No, then the branch returns to point A. If the start process is completed, the operating conditions of the internal combustion engine are detected in S3. The operating conditions are defined by the motor speed nactual, the charge air temperature TLL, the charge air pressure pLL, etc. In S4, the subroutine optimizer is called, and initial values, such as injection start, are generated in S5. In steps S6 to S8, the subroutines pre-optimization, digitization and re-optimization are called successively. The subroutines are combined Figures 5 to 7 The reoptimized quality measure calculated in the reoptimization subroutine is set as the minimum quality measure J (min), which determines the operating point of the internal combustion engine. Next, it is checked in S10 whether the motor stop has been initiated. If this is not the case, the query result S10: No, then the branch returns to point B. Otherwise, the program flow chart ends.

[0030] exist Figure 5 In the example of FIG. 1 , the subroutine pre-optimization is shown as a program flow chart. In S1, the pre-optimized first quality measure J1 (VO) is calculated according to equation (1). The essential feature is that, in the calculation of the first quality measure J1 (VO), in addition to continuous adjustment parameters with a continuous adjustment range, discrete adjustment parameters with discrete values ​​are interpreted as continuous adjustment parameters by interpolation. In S2, the operating variable i is set to zero. Then, in S3, the initial value is changed and calculated as a new theoretical value for the adjustment parameter. In S4, the operating variable i is increased by 1. Based on the new theoretical value, the pre-optimized second quality measure J2 (VO) is then predicted in S5 within the prediction limits, for example for the next 8 seconds. In S6, the first quality measure J1 (VO) is subtracted from the second quality measure J2 (VO) and compared with the limit value GW. The further development of the quality measure is checked by subtracting the two quality measures. Alternatively, based on the comparison of the operating variable i with the limit value iGW, it is checked how many times the optimization has been carried out. In this regard, the two limit value observations are stop criteria for further optimization. If further optimization is possible, i.e. the result of the query S6: No, the branch returns to point A. Otherwise, in S7 the optimizer outputs the second quality metric J2(VO) as the pre-optimized quality metric J(VO) together with the calibration parameters calculated therein and returns it to Figure 4 The pre-optimized quality measure J(VO) is a purely calculated parameter, that is, the calculated injection system theoretical value, the calculated gas path theoretical value and the calculated adjustment parameters are not connected to the internal combustion engine by the optimizer.

[0031] exist Figure 6, the subroutine digitization is shown. In S1, the pre-optimized quality measure J(VO) with the associated calibration parameters is read in. Subsequently, the calibration parameters are discretized with the initial discrete adjustment values. This is done in S2 with a hysteresis band according to the corresponding threshold value. The hysteresis band avoids floating calculated values. Instead of the hysteresis band, other logics that prevent rapid changes, such as time control, can also be used. Afterwards, the new discrete calibration parameter SG(new) is output in S3 and the process returns to Figure 4 in the main program.

[0032] exist Figure 7 In the example of FIG. 1 , the subroutine reoptimization is shown as a program flow chart. By means of the subroutine reoptimization, a reoptimized quality measure is determined from the operating conditions of the internal combustion engine and the new discrete tuning parameters SG(new). When calculating the reoptimized quality measure, the new discrete tuning parameters are not updated (nachgeführt). In S1, the reoptimized first quality measure J1 (NA) is calculated according to equation (1). In S2, the operating variable i is set to zero. Thereafter, in S3, the initial value is changed and calculated as a new theoretical value for the tuning parameter. In S4, the operating variable i is increased by 1. Then, in S5, the reoptimized second quality measure J2 (NA) is predicted within the prediction limits, for example for the next 8 seconds, according to the new theoretical value. In S6, the first quality measure J1 (VO) is subtracted from the second quality measure J2 (VO) and compared with the limit value GW. By subtracting the two quality measures, the further progress of the quality measure is checked. Alternatively, it is checked how many times the optimization has been performed based on a comparison of the operating variable i with the limit value iGW. In this respect, the two limit value observations are stop criteria for further optimization. If further optimization is possible, i.e. the result of the query S6: No, the branch returns to point A. Otherwise, in S7, the optimizer outputs the second quality measure J2 (VO) as the minimum quality measure J (min) and returns to Figure 4 in the main program.

[0033] These two Figure 8 and 9 The course of selected parameters over time in seconds is shown in a comparative manner. The parameters shown are: variable valve control VVT in crankshaft rotation degrees, injection start SB in degrees before top dead center (OT), combustion pressure pZYL in the cylinder and engine speed nMOT. For the combustion pressure pZYL, the maximum permissible combustion pressure pMAX is additionally drawn as a dotted line. The left half of the drawing page shows these parameters when applying a current optimization, while the right half of the drawing page shows these parameters when applying the present invention. The step-like rising theoretical torque as an input parameter is based on Figure 8 and Fig. 9First, according to Figure 8 to describe these parameters. In a first step, the optimizer calculates a pre-optimized quality measure by pre-optimization depending on the operating situation. In the calculation, discrete adjustment parameters with discrete adjustment values ​​are interpreted as continuous adjustment parameters with a continuous adjustment range. For the variable valve control VVT, a continuous trend with arbitrary intermediate values ​​is obtained over the entire time range. However, for a VVT adjustment link for controlling a variable valve with three defined adjuster positions, such a trend cannot be presented. The calculated injection start SB and the corresponding cylinder pressure pZYL correspond to the pre-optimized quality measure. With regard to the cylinder pressure pZYL, the maximum value pMAX is followed. The motor speed nMOT that increases in the observation time interval is obtained from the adjustment variable. Subsequently, Fig. 9 The VVT ​​trend shown corresponds to the trend after digitization. Figure 8 Unlike the illustration of the VVT ​​curve, only three discrete values ​​are shown, namely crankshaft angles of 450°, 495° and 540°. Advantageously, the VVT ​​adjustment element can be controlled with only three values, thereby significantly reducing the computing effort. Depending on the operating conditions of the internal combustion engine, a reoptimized quality measure is calculated from the VVT ​​curve. The curve of the injection start SB and the cylinder pressure pZYL correspond to the reoptimized quality measure, which in this case also remains below the maximum value pMAX.

[0034] Reference numerals list

[0035] 1 Internal combustion engine

[0036] 2 Fuel tanks

[0037] 3 Low pressure pump

[0038] 4 Suction throttle

[0039] 5 High pressure pump

[0040] 6 Tracks

[0041] 7 Injector

[0042] 8 Single memory

[0043] 9 Rail pressure sensor

[0044] 10 Electronic control instruments

[0045] 11 Exhaust gas turbocharger

[0046] 12 Charge air cooler

[0047] 13 Throttle valve

[0048] 14 Access site

[0049] 15 Inlet valve, variably controllable

[0050] 16 Outlet valve, variably controllable

[0051] 17 AGR adjustment link (AGR: exhaust gas recirculation)

[0052] 18 Turbine bypass valve

[0053] 19 Combustion Model

[0054] 20 Gas path model

[0055] 21 Optimizer

[0056] 22 Rail pressure regulation circuit

[0057] 23 λ regulation loop

[0058] 24 AGR regulation loop

[0059] 25 Pre-optimization

[0060] 26 Numericalization

[0061] 27 Optimize again.

Claims

1. A method for model-based control and regulation of an internal combustion engine (1), wherein: The optimizer (21) calculates a pre-optimized quality measure (J(VO)) in a first step as a function of an operating situation (BS), wherein in calculating the pre-optimized quality measure (J(VO)), a discrete control parameter with discrete adjustment values ​​is interpreted as a continuous control parameter (SG(k)) with a continuous adjustment range, wherein in a second step, the continuous control parameter (SG(k)) is quantified and set as a new discrete control parameter (SG(new)) with discrete adjustment values, wherein in a third step, the optimizer (21) calculates a re-optimized quality measure (J(NA)) as a function of the new discrete control parameter (SG(new)) and the operating situation (BS) of the internal combustion engine (1), and the optimizer (21) sets the re-optimized quality measure (J(NA)) as decisive for the operating point of the internal combustion engine (1), The pre-optimized quality measure (J(VO)) is determined by calculating an injection system theoretical value for controlling an injection system adjustment link by means of a combustion model (19), calculating a gas path theoretical value for controlling a gas path adjustment link by means of a gas path model (20), and calculating the continuous adjustment parameter (SG(k)) from discrete adjustment values ​​of the discrete adjustment parameter by means of interpolation, The re-optimized quality measure (J(NA)) is determined in the third step by calculating an injection system theoretical value for controlling an injection system adjustment link by a combustion model (19), calculating an air path theoretical value for controlling an air path adjustment link by an air path model (20), and by the optimizer (21) by changing the injection system theoretical value and the air path theoretical value under constant new discrete adjustment parameters (SG(new)) by minimizing the value within a prediction limit.

2. The method according to claim 1, characterized in that The pre-optimized quality measure (J(VO)) is determined in such a way that it is not accessible for a tuning session of the internal combustion engine.

3. The method according to claim 2, characterized in that In the second step, the continuous correction variable (SG(k)) is quantified via a switching threshold including hysteresis.

Images