Method for regulating the combustion air ratio of an internal combustion engine and control unit

Abstract

The invention relates to a method for controlling the air-fuel ratio of an internal combustion engine (100), comprising the following steps: - determining an instability value (160) that is indicative of an instability of the control; - evaluating a criterion based on the instability value (160); - if the criterion is met, determining a control response (121) to a control deviation (120) by means of a first controller (111), wherein the control deviation (120) corresponds to a deviation between an actual value (118) of a controlled variable (135) determined by means of a measuring element (117) and a setpoint value (119) of the controlled variable (135), and the controlled variable (135) is characteristic of the air-fuel ratio; - if the criterion is not met, determining the control response (121) to the control deviation (120) by means of a second controller (112); and - Implementing the control response (121) using a manipulated variable (122) as an input to a controlled system (113).Furthermore, the invention relates to a corresponding control unit (104) and computer program.

Description

TECHNICAL AREA

[0001] The present disclosure relates to methods for lambda control of an internal combustion engine, as well as corresponding control units and a computer program designed to carry out such methods. BACKGROUND OF THE INVENTION

[0002] Incomplete combustion of an air-fuel mixture in an internal combustion engine produces not only nitrogen, carbon dioxide, and water, but also other gaseous combustion products. There are legal restrictions regarding the emission of hydrocarbons, carbon monoxide, and nitrogen oxides, for example. When a three-way catalytic converter (TWC) is used in the exhaust system of an internal combustion engine, these emissions are converted into nitrogen, carbon dioxide, and water.

[0003] However, such conversion is only optimal at a stoichiometric operating point, which is characterized by a lambda-air ratio of one. To achieve this operating point, lambda control is performed in the engine control unit of the internal combustion engine. Lambda control is typically based on signals from lambda sensors upstream and downstream of the three-way catalytic converter. For upstream lambda control, also known as pre-catalyst control, the oxygen content of the exhaust gas upstream of the catalytic converter is measured using a linear lambda sensor. Depending on the measured value, the control unit corrects the injected fuel quantity from the engine's pilot control.

[0004] Accuracy can be further improved by using the signal from the second lambda sensor behind the catalytic converter.

[0005] The quality of the pre-catalyst control system plays a crucial role in the accuracy, speed, and stability of the lambda control. These control characteristics, in turn, directly influence emissions, because any deviation from a predefined lambda target value or lambda setpoint causes emissions. Therefore, optimal lambda control is of paramount importance, especially in light of increasingly stringent emissions legislation.

[0006] For pre-catalyst control, PI controllers have frequently been used, but these exhibit stability problems under certain operating conditions. Such stability problems can, for example, lead to a constant or systematic deviation from the lambda target value, which also worsens the combustion engine's pollutant emissions. Furthermore, engine running can become unstable, which can negatively affect the combustion engine's acoustics or the vehicle's driving characteristics. In the worst-case scenario, uncontrolled oscillations can cause significant damage to the drive system or the entire vehicle.

[0007] For the reasons mentioned, robust controllers with a slow response time are most commonly used for lambda control nowadays. However, this comes at the expense of an optimal response time and generally leads to higher emissions.

[0008] From DE 37 27 369 A1, a control system for processes that can be approximated by a delay element and a dead-time element is known, wherein the controller behavior switches based on the difference between the setpoint and the controlled variable when significant setpoint changes occur. Under normal operating conditions, the controller operates in proportional-integral mode, while upon a significant setpoint change, it switches to integral control. SUMMARY AND FORMS OF EXECUTION

[0009] It is therefore an objective of the present disclosure to provide a method for controlling the combustion air ratio of an internal combustion engine and a corresponding control device which enable fast, accurate, but also robust control, in particular to reduce emissions of the internal combustion engine.

[0010] This task is solved by a method, a control unit, and a computer program according to the independent patent claims. Advantageous embodiments and further developments are described in the dependent claims, the following description, and the drawings.

[0011] Thus, according to a first aspect, a method for controlling the air-fuel ratio of an internal combustion engine is provided. The method comprises the following steps: (a) determining an instability value, which is indicative of an instability in the control system; (b) evaluating a criterion based on the instability value; (c) if the criterion is met, determining a control response to a control deviation using a first controller, where the control deviation corresponds to a deviation between an actual value of a controlled variable determined by a measuring element and a setpoint value of the controlled variable, and the controlled variable is characteristic of the air-fuel ratio; (d) if the criterion is not met, determining the control response to the control deviation using a second controller; and (e) implementing the control response using a manipulated variable as an input to a controlled system.To determine the instability value, at least one value of the controlled variable modeled by a model is compared with a corresponding value of the controlled variable determined by the measuring element.

[0012] According to another aspect, a control unit with a first controller and a second controller is provided, which is configured to carry out the procedure described above.

[0013] According to another aspect, a computer program is provided which includes instructions that, when executed by a computer, cause it to carry out the previously described procedure. In the context of the present disclosure, a computer is defined, for example, as a device that processes data using programmable computational instructions. Computers can be embedded in everyday devices, such as the control units of motor vehicles.

[0014] According to another aspect, a storage medium is provided with a computer program, wherein the computer program includes instructions which, when the computer program is executed by a computer, cause it to carry out the procedure described above.

[0015] In the context of this disclosure, a controlled system is defined, for example, as that part of a control loop that contains the system or process to be controlled. The controller(s) of the control loop can be configured to act on the controlled system via one or more manipulated variables. If the manipulated variable characterizes fuel injection into an internal combustion engine and the controlled variable characterizes an air-fuel ratio or lambda, the controlled system can, for example, describe the physical relationship between fuel injection into the internal combustion engine and the measurement of the air-fuel ratio in an exhaust system downstream of the internal combustion engine.

[0016] In the context of this disclosure, a manipulated variable is defined, for example, as the output of a controller and / or as the input of the controlled system. The manipulated variable can be suitable for influencing the controlled variable. A control response of the controller can be implemented using the manipulated variable. For example, the manipulated variable can characterize a fuel injection process.

[0017] In the context of this disclosure, a controlled variable is defined, for example, as an output variable of the controlled system. It can be a variable to be controlled, which, for example, is to be controlled such that it corresponds to a predetermined setpoint or target curve. The controlled variable can be, for example, an air-fuel ratio or lambda. The air-fuel ratio can be or characterize the ratio of air to fuel compared to a stoichiometric combustion mixture. The stoichiometric combustion mixture can be an optimal ratio between the reactants, for example, combustible materials and oxygen as an oxidizing agent.

[0018] In the context of this disclosure, a measuring element is defined, for example, as a measuring device at or after the output of the controlled system, which is specifically configured to determine the controlled variable directly or indirectly. The measuring element may be configured to determine measured values ​​from which corresponding values ​​of the controlled variable can be derived. The measuring device may be wholly or partially part of the controlled system or may be separate from it. For example, the measuring element may be a measuring device configured to determine the oxygen content in an exhaust gas from an internal combustion engine, from which the air-fuel ratio can be derived, at least approximately.

[0019] In the context of this disclosure, an instability value is defined, for example, as a value that is indicative of instability in the control system. For instance, the instability value may be higher the more unstable the control system is, particularly the narrower the range of operating conditions under which the control system is stable. The instability value may correlate with an uncertainty in the control system, which may be characterized, for example, by a deviation of the controlled variable from a modeled controlled variable.

[0020] The device and / or method described above can be advantageous in order to enable fast, accurate and robust control.

[0021] Particularly when disturbances affect the controlled system and / or the measuring element, such as an aging lambda sensor or its replacement, or inaccuracies in model parameters, a controller designed for fast or optimal control can exhibit stability problems. The greater the inaccuracy or disturbance, the less stable the control loop can be. Consequently, in the event of instability, the adaptation of the control loop can be so impaired that the actual control value shows a constant deviation from the target control value or target control curve. This, in turn, can lead to increased pollutant emissions.

[0022] To counteract such stability problems, two controllers are provided, which are used depending on an instability value. In this way, for example, a better, and especially faster, control adaptation can be ensured at a low instability value. At a high instability value, on the other hand, a more robust control adaptation can be implemented, in which the control response is slower and / or has a smaller amplitude.

[0023] In this way, the control system can be stabilized under a wide range of operating conditions while simultaneously improving the control response. This not only reduces pollutant emissions but also stabilizes engine operation, thereby improving, for example, driving acoustics or overall driving characteristics. Furthermore, harmful oscillations of the drive system or the vehicle can be avoided or at least reduced, thus protecting system components.

[0024] According to one embodiment, the first controller comprises a first plurality of controllers and / or the second controller comprises a second plurality of controllers. In other words, a third controller and possibly further controllers can be provided. Depending on the instability value, switching to the third controller or the further controllers can occur. Exactly one of the different controllers is always used for control. Such an embodiment can be advantageous for implementing multiple controllers of varying robustness and thus for finer control response than with just two controllers.

[0025] According to one embodiment, the deviation between the actual value of the controlled variable and the setpoint value of the controlled variable corresponds to a difference. Other functional dependencies are also possible.

[0026] According to one embodiment, the control response of the first controller is reduced compared to the control response of the second controller at the same control deviation; in particular, it is slower and / or has a smaller amplitude. In other words, the second controller aims for "optimal," especially fast, control, while the first controller is designed for "robust," especially stable control under many operating conditions.

[0027] According to one embodiment, the second controller is configured to consider a rise time, overshoot, settling time, and a state error or steady-state error in the control response. The rise time, particularly a short rise time, may be preferred over the other parameters. The overshoot may be preferred over the remaining parameters, i.e., apart from the rise time. Such an embodiment can characterize an "optimal" control system.

[0028] According to one embodiment, the control response of the first controller, compared to the control response of the second controller at the same control deviation, is characterized by the fact that the controlled variable exhibits at least one of the following properties: an increased rise time; a reduced overshoot; a reduced state error or steady-state error. These differences can occur, for example, in response to a sudden change in the setpoint. Such an embodiment can be advantageous because the aforementioned measures all contribute to the robustness of the control system.

[0029] According to one embodiment, the control deviation, and in particular its magnitude, is taken into account when determining the instability value. The control deviation can be a current deviation, for example, at the end of an observation window. Such an embodiment can be advantageous because control deviations can be indicative of possible instabilities in the control system, especially when such deviations occur together with model deviations.

[0030] According to one embodiment, to determine the instability value, a model deviation is determined which characterizes a deviation of the controlled variable from a model, and this model deviation is weighted with the control deviation to obtain the instability value.

[0031] According to the invention, to determine the instability value, at least one modeled value of the controlled variable is compared with a corresponding value of the controlled variable determined by the measuring element. The two values ​​can refer to the same time. The comparison can include calculating a difference between the two values, in particular the magnitude of the difference. A difference normalized using the modeled value can also be calculated, in particular the magnitude of the normalized difference.

[0032] The model deviations determined through comparison can indicate potential instabilities in the control system. If such model deviations are identified, the control response can be modified, for example, by switching from the second controller to the first controller after evaluating the criterion, particularly towards a more robust control system. The control response can be slowed down and / or its amplitude reduced. This can increase the robustness of the control system when stability problems occur.

[0033] According to one embodiment, to determine the instability value, several modeled values ​​of the controlled variable are compared with corresponding values ​​of the controlled variable determined by the measuring element at different times during an observation window. The compared measured and modeled values ​​can each refer to the same time. The comparison can include calculating a difference between the measured and the modeled value, in particular the magnitude of the difference. A normalized difference can also be calculated, in particular the magnitude of the normalized difference. The differences or the normalized differences can be summed to characterize model deviations during the observation window.

[0034] Such an embodiment can be advantageous because model deviations can indicate potential instabilities in the control system. If such model deviations are detected, the control response can be modified, for example, by switching from the second controller to the first controller after evaluating the criterion, particularly towards a more robust control system. The control response can be slowed down and / or its amplitude reduced. This can increase the robustness of the control system when stability problems occur.

[0035] According to one embodiment, the observation window shifts over time. This shifting can be implemented using a ring buffer and / or a FIFO (first in, first out) principle. For example, an observation window of constant duration can be used. A predefined number of time points or samples within the observation window can be used for comparison.

[0036] According to one embodiment, model parameters are determined at least partially by diagnosing the lambda sensor, i.e., the lambda sensor actually used in the exhaust system, or a nominal lambda sensor. A production tolerance can be taken into account for the nominal lambda sensor. Such calibration of the model can be advantageous to avoid costly re-diagnostics for model adaptation.

[0037] According to one embodiment, the model comprises a model of the controlled system and / or the measuring element. Only aspects of the controlled system and / or the measuring element relevant to the controlled variable may be modeled. It can therefore be an abstract model, comprising, for example, only a few mathematical formulas, such as nth-order polynomial functions, and / or one or more tables. The model can be an nth-order system with dead time, in particular a first-order system with dead time. According to one embodiment, the dead time corresponds to or correlates with the transit time from the injection of the fuel until the measurement of the corresponding lambda value by the measuring element.

[0038] According to one embodiment, the instability value is indicative of a model inaccuracy, in particular an inaccuracy of one or more model parameters, an inaccuracy of one or more modeling assumptions, and / or one or more disturbances of the controlled system that are not fully accounted for by the model. An inaccurate model parameter can, for example, result from aging or replacement of the measuring element, such as a lambda sensor. The model inaccuracy can be weighted according to the control deviation, in particular with a greater weighting the larger the control deviation. The model can include a model of the controlled system and / or the measuring element. It can be suitable for modeling the controlled variable, in particular as a function of the manipulated variable.

[0039] In the context of this disclosure, a disturbance variable is defined, for example, as a quantity acting on the controlled system and / or the measuring element that is capable of influencing the controlled variable. A disturbance variable could, for example, be external damage to the measuring element that causes a pulsating signal from the lambda sensor.

[0040] According to one embodiment, the instability value is compared to a threshold value to evaluate the criterion. For example, the criterion may be met if the instability value is above the threshold, and the criterion may not be met if the instability value is below the threshold. The threshold value may be a constant. Alternatively, it may depend on a previous instability value. One or more additional threshold values ​​may be specified, which trigger a switch to a third controller or additional controllers.

[0041] According to one embodiment, the criterion incorporates a hysteresis whereby the criterion depends on a previous instability value. For example, the criterion can depend on whether the system is currently controlled by the first controller or the second controller. For instance, a threshold value used for the criterion can be chosen differently, particularly smaller, when the system is currently controlled by the first controller, compared to the threshold value when the system is controlled by the second controller. Such an embodiment can be advantageous to avoid rapid switching between the first and second controllers.

[0042] According to one embodiment, the instability value is compared to a first threshold when the instability value is increasing, and to a second threshold when the instability value is decreasing, to evaluate the criterion. The second threshold, in particular a deactivation threshold for the first controller, can be lower than the second threshold, in particular an activation threshold for the first controller. The second threshold can, for example, be more than 3 percent lower, in particular more than 5 percent lower, in particular more than 8 percent lower, than the first threshold. The activation threshold for the first controller can be a deactivation threshold for the second controller and / or the deactivation threshold for the first controller can be an activation threshold for the second controller.Such an embodiment can be advantageous because, in case of suspected instability of the system, a stable or robust control is prioritized, and because rapid switching between controllers is reduced.

[0043] According to one embodiment, the measuring element is a lambda sensor that measures the oxygen content in the exhaust gas of the internal combustion engine. Such lambda sensors for the indirect measurement of the air-fuel ratio are typically used in lambda control systems.

[0044] According to one embodiment, the lambda sensor is arranged in the exhaust stream of the internal combustion engine upstream of a catalytic converter, in particular a three-way catalytic converter. The lambda sensor is thus located between the internal combustion engine and the catalytic converter. In other words, the associated lambda control is a so-called pre-catalyst control. This can be advantageous for low-emission operation of the catalytic converter.

[0045] According to one embodiment, the control variable is characteristic of a fuel quantity injected into the internal combustion engine. Such an embodiment can be advantageous because the injected fuel quantity is a typical control variable when regulating the air-fuel ratio. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Further advantages and beneficial designs and further developments of the method and the control unit result from the following exemplary embodiments shown in conjunction with the figures.

[0047] They show: Fig. 1 an internal combustion engine with an exhaust system and a control unit according to an embodiment of the present disclosure; Fig. 2 a control loop for a method for controlling a combustion air ratio according to an embodiment of the present disclosure; Fig. 3 a flowchart for a method for controlling a combustion air ratio according to an embodiment of the present disclosure; Fig. 4 the course of an instability value as a function of measured and modeled control variable for evaluation in a method for controlling a combustion air ratio according to an embodiment of the present disclosure; Fig. 5 different aspects of the behavior of a controlled variable that are relevant for the comparison between optimal and robust control in a method for controlling a combustion air ratio according to an embodiment of the present disclosure.

[0048] Identical, similar, or similarly effective elements are marked with the same reference symbols in the figures. In some figures, individual reference symbols have been omitted for clarity. The figures and the relative sizes of the elements depicted within them are not to be considered to scale. Rather, individual elements may be exaggerated for better representation and / or comprehensibility. DETAILED DESCRIPTION OF EXAMPLES OF EXECUTION

[0049] Fig. Figure 1 shows an internal combustion engine 100 and an exhaust system 101 coupled to the internal combustion engine 100. The exhaust system 101 comprises a three-way catalytic converter 103 and a lambda sensor 102 both downstream and upstream of the catalytic converter 103. The lambda sensor 102, located between the internal combustion engine 100 and the three-way catalytic converter 103, is a linear lambda sensor designed to measure the oxygen content of the exhaust gas upstream of the three-way catalytic converter 103. Based on this measurement, a control unit 104 performs the lambda control upstream of the three-way catalytic converter 103, i.e., the so-called pre-catalyst control. Depending on the measurement, the control unit 104 corrects the fuel quantity control from the pilot control of the internal combustion engine 100. For higher accuracy, the signal from the second lambda sensor 102 downstream of the three-way catalyst 103 is also taken into account.

[0050] Fig. Figure 2 shows a circuit diagram or control loop 110 of a method that is derived from the one in Fig. The control unit 104 shown in section 1 is used for pre-catalyst control. Fig. Figure 2 shows the pre-catalyst control as a switchable control system with two parts or two controllers 111, 112. A switch 124 is configured to toggle between a first, robust controller 111 and a second, optimal controller 112. Depending on the switch position, the control response 121 of either the first or the second controller 111, 112 is used as the input to the controlled system 113. A controlled variable 118 at the output of the controlled system 113 is compared with a reference variable 119, whereby the deviation 120 of the controlled variable 135 from the reference variable 119 serves as the input to the two controllers 111, 112.

[0051] The optimal controller 112 refers to a controller that meets defined criteria regarding control quality. This could be, for example, an IMC (Internal Model Regulator), MPC (Model Predictive Regulator), LQR (Linear Quadratic Regulator), or a PI controller, especially an optimally calibrated PI controller. Control quality relates, among other things, to the signal metrics rise time, percentage overshoot, settling time, and steady-state error, which characterize a deviation from the lambda target value and can thus also influence emissions.

[0052] The robust controller 111 refers to a controller that meets defined stability criteria under all environmental, system, and operating conditions. With regard to robustness, control quality relates to the signal metrics: percentage overshoot and steady-state error.

[0053] The controlled system 113 can be described as an nth-order system with dead time, specifically as a first-order system with dead time. The controlled system 113 is modeled using a model 114, which determines a modeled value 118' of the controlled variable based on the manipulated variable 122. An inaccuracy in the parameters of the model 144 or an unknown disturbance caused by a disturbance variable 123 in the controlled system 113 can be referred to as an indeterminacy of the system.

[0054] The aim is to recognize that the control loop 110 behaves unstably when the system exhibits a certain degree of uncertainty. In such a case, the control loop 110 switches from the optimal controller 112 to the robust controller 111, thereby stabilizing the control loop. When the uncertainty disappears, the control loop switches back from the robust controller 111 to the optimal controller 112. The behavior of the control loop 110 is thus changed from optimality to robustness or vice versa.

[0055] A stabilizer 115 in the control loop 110 is configured to detect instabilities. For this purpose, the stabilizer 115 compares a value 118' of the controlled variable, modeled by a model 114, with an actual value 118 of the controlled variable determined by a measuring element 117. The stabilizer 115 determines an instability value 160, for which the control deviation 120 is also taken into account, and evaluates a criterion based on the instability value 160. The stabilizer 115 is further configured to send a switching signal to the switch 124 to toggle between the optimal controller 112 and the robust controller 111. A hysteresis element 116 arranged between the stabilizer 115 and the switch 124 prevents excessively rapid switching between the two controllers 115 and 116.

[0056] The two controllers 111 and 112, as well as the stabilizer 115, each receive an actual value 118 of the controlled variable, determined by the measuring element 117, and a setpoint 119 of the reference variable. A control deviation 120 is calculated by comparing the actual value 118 with the setpoint 119 and is used as input for the two controllers 111 and 112 and the stabilizer 115. In response to the control deviation 120, the two controllers 111 and 112 determine respective control responses 121. Depending on the position of the switch 124, one of these control responses 121 is converted into an input for the controlled system 123 by means of a manipulated variable 122, in order to obtain an actual value 118 of the controlled variable as the output of the controlled system 123. This actual value is measured by the measuring element 117. The combination of the first robust controller 111 and the second optimal controller enables precise and stable control of the actual value 118 to the setpoint 119.

[0057] Fig. Figure 3 shows the flowchart of a procedure for evaluating a switching criterion between the two controllers 111 and 112. Such a procedure can, for example, be carried out by the stabilizer 115. In the procedure, the controlled variable 135, here the lambda signal, is observed in a window 162 (see Figure 3). Fig. 4) modeled and measured.

[0058] In a first step, the observation window 162 is shifted. The observation window 162 has a defined size and moves over time. The window 162 can be implemented as a ring buffer. In each computation step, a new sample is added and an old sample is removed, for example, according to a FIFO (first in, first out) principle.

[0059] In a second step 151, an instability value 160 is determined based on the samples in the observation window 162. For this purpose, the relative deviation between the modeled and measured lambda sensor value 118', 118 is multiplied by the control deviation 120 for different time points during the observation window, i.e., for different samples, and then summed: δ=∑k=0N−1(|y(k)−ym(k)ym(k)|∗|e|)

[0060] Here, y is the measured lambda value 118 (output of the control loop 113), y m the modeled lambda value 118' (model output 115) and e the control deviation 120; k denotes the different time points or samples during the observation window.

[0061] In a third step 152, the switching criterion is evaluated by examining whether the sum δ, i.e., the instability value 160, exceeds an instability threshold 161. If this is the case, the switch 124 receives a trigger to switch from the optimal controller 112 to the robust controller 111. Thus, based on the comparison between the instability value 160 and the instability threshold 161, either the switch is changed or the switch position is maintained, i.e., the switch is made to the first controller 153 or to the second controller 154.

[0062] The window 162 is then shifted back in time by one sample (150), and the algorithm is repeated. The calculation runs continuously as long as the lambda control is active.

[0063] When evaluating the switching criterion 152, digital hysteresis can be taken into account using the hysteresis element 116. If an input, i.e., the instability value 160, reaches the activation threshold in the downward direction, the output of the evaluation is set to "True", i.e., the first controller 153 is switched on. If the input, i.e., the instability value 160, reaches the deactivation threshold in the downward direction, the output of the evaluation is set to "False", i.e., the second controller 154 is switched on. In other cases, the output of the hysteresis does not change, and the previous state is maintained.

[0064] The activation threshold can be set equal to the instability threshold 161. The deactivation threshold can be defined as instability threshold 161 minus an offset. The offset, for example, approximately 10% of instability threshold 161, defines the hysteresis band. The motivation for this is to prioritize a more robust control regime over a more optimal one when instability is suspected, and to maintain the more robust regime for a longer period.

[0065] Fig. Figure 4 shows exemplary progressions of modeled values ​​y in the upper part. m 118' of the controlled variable and measured values ​​y 118 of the controlled variable 135 over time 130. The observation window 162 determines which values ​​are considered for the switching criterion according to the one associated with Fig. 3 explained formula.

[0066] In the lower part of Fig. Figure 4 shows the corresponding curve of the instability value δ 160 calculated using this formula over time 130, as well as the instability threshold 161. Based on a comparison of the instability value 160 and the instability threshold 161, the switch 124 is used to toggle between the two controllers 111 and 112.

[0067] This method can ensure control stability under all environmental, system, and driving conditions. The control strategy is adjusted for optimality or robustness based on the uncertainty of the control process. Advantageously, the controllers can have very different topologies and therefore be optimally designed for various control criteria.

[0068] Fig.Figure 5 illustrates various control criteria. It depicts the response over time 130 of the actual value 118 of the controlled variable 135 to a sudden change in the setpoint 119 of the reference variable from a smaller constant value to a larger constant value. After the change in the setpoint 119, the actual value 118 changes in the direction of the changed setpoint 119. The time until the changed setpoint is reached is characterized by the rise time 131. The rise time 131 can be defined based on a tangent to the curve of the actual value 118 at the intersection point with the changed setpoint 119. An overshoot of the actual value 118 after the intersection point with the changed setpoint 119 is characterized by the overshoot amplitude 133. The overshoot 133 can be defined as the largest amplitude of a deviation from the changed setpoint 119 or from a steady-state value of the actual value 118 after the intersection point.After one or more oscillations of the actual value 118 around the changed setpoint 119, the actual value 118 approaches the setpoint 119. The total time of this approach, starting from the time of the change in the setpoint 119, is characterized by the settling time 132. Possible deviations of the actual value 118 from the setpoint 119 after the settling time 132 are referred to as steady-state errors 134 or state errors. In the case of a sudden change in the setpoint 119, optimal control is characterized, for example, by a fast rise time, while robust control is characterized by a slow rise time, but small overshoot and small steady-state errors.

[0069] The invention is not limited to the exemplary embodiments described therein. Rather, the invention encompasses every new feature as well as every combination of features, which in particular includes every combination of features in the exemplary embodiments and claims. REFERENCE MARK 100 internal combustion engine 101 Exhaust system 102 Lambda sensor 103 Catalyst 104 Control unit 110 Control loop 111 first regulator 112 second controller 113 Control section 114 Model 115 Stabilizer 116 Hysteresis 117 Measuring element / Lambda probe 118 Actual value of the controlled variable 118' modeled actual value of the controlled variable 119 Setpoint of the controlled variable / reference variable 120 rule deviation 121 Rule Answer 122 Control variable 123 Disturbance variable 124 switches 130 Time 131 Ascent time 132 Control period 133 Overshoot 134 stationary errors 135 Control variable 150 Shifting the observation window 151 Determining the instability value 152 Evaluating the switching criterion 153 Switching to the second controller 154 Switching to first regulator 160 instability value 161 Instability threshold 162 observation windows

Claims

[1] Method for controlling the air-fuel ratio of an internal combustion engine (100), comprising the following steps: - Determining an instability value (160) that is indicative of instability in the control system; - Evaluating a criterion based on the instability value (160); - if the criterion is met, determining a control response (121) to a control deviation (120) by means of a first controller (111), wherein the control deviation (120) corresponds to a deviation between an actual value (118) of a controlled variable (135) determined by means of a measuring element (117) and a setpoint value (119) of the controlled variable (135) and the controlled variable (135) is characteristic of the combustion air ratio; - if the criterion is not met, determine the control response (121) to the control deviation (120) using a second controller (112); and - Implementing the control response (121) by means of a manipulated variable (122) as an input to a controlled system (113), wherein, for the determination of the instability value (160), at least one value (118') of the controlled variable (135) modeled by means of a model (114) is compared with a corresponding value (118) of the controlled variable (135) determined by means of the measuring element (117). [2] Method according to the preceding claim, wherein the control response (121) of the first controller (111) is reduced compared to the control response (121) of the second controller (112) at the same control deviation (120), in particular is slowed down and / or has a smaller amplitude. [3] Method according to one of the preceding claims, wherein the control response (121) of the first controller (111) is compared to the control response (121) of the second controller (112) at the same control deviation (120) characterized byis that a control variable (135) has at least one of the following properties: an increased rise time (131); a reduced overshoot (133); a reduced state error or stationary error (134). [4] Method according to one of the preceding claims, wherein the control deviation (120), in particular the amount of the control deviation (120), is taken into account for determining the instability value (160). [5] Method according to one of the preceding claims, wherein for determining the instability value (160) several values ​​(118') of the controlled variable (135) modeled by means of the model (114) are compared with corresponding values ​​(118) of the controlled variable (135) determined by means of the measuring element (117) at different times during an observation window (162), wherein the observation window (162) shifts as time progresses. [6] Method according to one of the preceding claims, wherein model parameters of the model (114) are determined at least partially by a diagnosis of the lambda probe (117) or a nominal lambda probe. [7] Method according to one of the preceding claims, wherein the model (114) comprises a model of the controlled system (113) and / or the measuring element (117). [8] Method according to one of the preceding claims, wherein the instability value (160) is indicative of a model inaccuracy which is weighted depending on the control deviation (120). [9] Method according to one of the preceding claims, wherein the instability value (160) is compared with a threshold value (161) for the evaluation of the criterion. [10] Method according to one of the preceding claims, wherein the criterion takes into account a hysteresis (116) according to which the criterion depends on a previous course of the instability value (160). [11] Method according to one of the preceding claims, wherein the measuring element (117) is a lambda probe (117) which measures an oxygen content in an exhaust gas of the internal combustion engine (100), wherein the lambda probe (117) is arranged in an exhaust gas stream (101) of the internal combustion engine (100) upstream of a catalyst (103), in particular a three-way catalyst. [12] Method according to one of the preceding claims, wherein the control variable (122) is characteristic of a quantity of fuel injected into the internal combustion engine (100). [13] Control unit (104) comprising a first controller (111) and a second controller (112) configured to perform the method according to one of the preceding claims. [14] Computer program comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any one of claims 1 to 12.

Images