DE102012200032B4 - Method and device for dynamic diagnosis of sensors - Google Patents

Abstract

Method for the dynamic diagnosis of sensors of an internal combustion engine (10), wherein the sensors have a low-pass behavior depending on the geometry and due to aging or contamination, a dynamic diagnosis being carried out when an input variable changes on the basis of a modeled and a measured signal, and wherein the measured signal is an actual value of an output signal of the sensor and the modeled signal is a model value, characterized in that a maximum slope (23) of a step response of a closed control loop for the sensor to a sudden change in the input variable for all sensible combinations of controller and system parameters (22.1, 22.2, 21.1 and 21.2) of the model are determined based on simulations and a time constant (24) for the sensor is derived from this and the dynamics diagnosis of the sensor is carried out based on the determined time constant (24), with the parameters of the simulation and an inverse reference model is trained based on the resulting gradient.

Description

State of the art

The invention relates to a method for the dynamic diagnosis of sensors of an internal combustion engine, wherein the sensors have a low-pass behavior depending on the geometry and due to aging or contamination, with a dynamic diagnosis being carried out when an input variable changes on the basis of a modeled and a measured signal and where the measured signal is an actual value of an output signal of the sensor and the modeled signal is a model value.

The invention further relates to a device for carrying out the method.

To reduce emissions in cars with gasoline engines, 3-way catalytic converters are usually used as exhaust gas purification systems, which only convert sufficient exhaust gases if the air-fuel ratio λ is regulated with high precision. For this purpose, the air-fuel ratio λ is measured using an exhaust gas probe upstream of the exhaust gas purification system, often in the form of a continuous lambda probe.

The oxygen storage capacity of such an exhaust gas purification system is used to absorb oxygen in lean phases and release it again in rich phases. This ensures that oxidizable harmful gas components of the exhaust gas can be converted. An exhaust gas probe connected downstream of the exhaust gas purification system is used to monitor the oxygen storage capacity of the exhaust gas purification system. The oxygen storage capacity must be monitored as part of the on-board diagnosis, as it represents a measure of the conversion ability of the exhaust gas purification system. To determine the oxygen storage capacity, either the exhaust gas purification system is first filled with oxygen in a lean phase and then emptied in a rich phase with a lambda value known in the exhaust gas, taking into account the amount of exhaust gas passing through, or the exhaust gas purification system is first emptied of oxygen in a rich phase and then in a lean phase filled with a lambda value known in the exhaust gas, taking into account the amount of exhaust gas passing through. The lean phase is ended when the exhaust gas probe connected downstream of the exhaust gas purification system detects the oxygen that can no longer be stored by the exhaust gas purification system. Likewise, a rich phase is ended when the exhaust gas probe detects the passage of rich exhaust gas. The oxygen storage capacity of the exhaust gas purification system corresponds to the amount of reducing agent supplied for emptying during the rich phase or the amount of oxygen supplied for filling during the lean phase. The exact quantities are calculated from the signal from the upstream exhaust gas sensor and the exhaust gas mass flow determined from other sensor signals.

If the dynamics of the upstream exhaust gas sensor decreases, e.g. due to contamination or aging, the air-fuel ratio can no longer be regulated with the required precision, so that the conversion performance of the exhaust gas purification system decreases. Furthermore, deviations can arise in the diagnosis of the exhaust gas purification system, which can lead to an exhaust gas purification system that is actually working correctly being incorrectly assessed as not functioning.

The law requires a diagnosis of the probe properties during driving to ensure that the required air-fuel ratio can still be set with sufficient precision, that emissions do not exceed the permissible limits and that the exhaust gas purification system is monitored correctly. Among other things, a deterioration in the probe dynamics must be recognized, which can become noticeable through an increased time constant and/or dead time. An additional distinction must be made between directional dependence (rich to lean or lean to rich). Since the lambda controller is designed to be error-free, a deterioration in the sensor dynamics can cause the lambda controller and thus the air-fuel ratio to oscillate, which increases emissions and can lead to problems when driving.

From the DE 10 2008 042 549 A1 is a method and a device for diagnosing the rate of rise and the dead time of an exhaust gas probe, which is arranged in an exhaust duct of an internal combustion engine, the diagnosis being based on a comparison of a modeled and a measured signal after a predetermined change in a fuel-air ratio air-fuel mixture supplied to the internal combustion engine is carried out and the signal is an output signal of the exhaust gas probe or a modeled or measured signal derived from the output signal. It is provided that a first extreme value is determined in the course of the modeled signal and that a first point in time and a first starting value are determined when the modeled signal deviates from the first extreme value by a predetermined amount, that a second extreme value in the course of the measured signal is determined. Furthermore, it is provided that a second time and a second starting value are determined if that measured signal deviates from the second extreme value by the predetermined amount, that a first integral is formed over a predetermined period of time, starting at the first time, via the difference between the first starting value and the modeled signal and that a second integral over a second period of time, starting at the second time, via the difference between the second starting value and the measured signal, it is formed that the second period is equal to the predetermined period or that the end of the second period is based on the time of the change in the fuel-air ratio or related is set to the first point in time and that a quantitative comparison value is formed from a quantitative comparison between the first integral and the second integral, from which a conclusion is drawn about the rate of rise and / or the dead time of the exhaust gas probe.

This method uses sudden adjustments of the air-fuel ratio, which are used to evaluate the dynamics of the exhaust gas sensor, whereby a directional dependency, i.e. from rich to lean or from lean to rich, is also distinguished. For this purpose, the area under the lambda signal of the exhaust gas probe is integrated for a certain period of time after the jump and put in relation to an analog calculated area of a lambda signal modeled in the control unit. If the calculated ratio is smaller than an applicable threshold, the exhaust gas probe no longer meets the required dynamic behavior.

Another method - here called the slope method - also uses sudden adjustments of the air-fuel ratio and determines the maximum slope of the measured air-fuel ratio within a certain period of time after the jump. The time period for evaluation is determined by an applicable stroke by which the measured air-fuel ratio can change after the jump. This period of time is interpreted as dead time, corrected by the difference between the time at the end of the evaluation and the time at which a straight line with the previously determined maximum slope through the value of the measured air-fuel ratio at the end of the evaluation reaches the value of the minimum/maximum measured air-fuel ratio during the evaluation. A time constant error is recognized by a maximum slope that is too small in magnitude.

Furthermore, in the subsequently published DE 10 2011 085 115 A1 describes a method to explicitly calculate a time constant from the maximum slope of the step response, which can be used to adapt the controller and as a diagnostic result. In this invention, it is proposed to carry out a step-by-step adaptation of the lambda controller parameters for dynamic detection in exhaust gas sensors by comparing the maximum slope in terms of magnitude of the measured air-fuel ratio with an expected maximum slope in terms of magnitude. Furthermore, it is proposed to calculate a time constant from the maximum slope of the measured air-fuel ratio, which is used for diagnosis and for control adaptation. However, the calculation of the time constants has a systematic error because an open control loop is assumed when deriving the method, but the lambda control loop is closed when the method is used in the vehicle. The control intervention in the case of a slowed down probe leads to a maximum slope that is systematically larger than expected based on the derivation. This means that the time constant is systematically identified as too small.

The same procedures can also be carried out analogously with inverted lambda signals.

Further procedures are from the DE 198 44 994 A1 and the JP 2004 - 68 602 A known.

The object of the invention is to provide an improved dynamic diagnosis in which systematic deviations of the identified time constants from the real value can be at least partially eliminated.

It is a further object of the invention to provide a corresponding device for carrying out the method.

Disclosure of the invention

The task relating to the method is solved by determining a maximum slope of a step response of a closed control loop for the sensor to a sudden change in the input variable for all sensible combinations of controller and system parameters of the model using simulations and deriving a time constant for the sensor from this and the dynamic diagnosis of the sensor is carried out on the basis of the determined time constant, an inverse reference model being trained with the parameters of the simulation and the resulting gradient. The influence of the closed control loop is taken into account when deriving the calculation rule as to how the time constant follows from the maximum gradient. This almost eliminates the systematic deviation of the identified time constant from the real value. The controller can therefore be better adapted to what is happening partly reflected in a more precise diagnostic result of the dynamics of the sensor.

In a preferred method variant, it is provided that a nonlinear static approximator is used as an inverse reference model, by means of which the relationships between all controller and system parameters on the input side and the time constant on the output side are mapped, with training of the reference model and/or the simulations being carried out offline. In this case, offline means that the calculations do not take place during operation in the control unit, i.e. are carried out online, but rather in the application phase on the PC. A further advantage arises if the controller has only a few parameters, for example if the control parameters largely correspond to the system parameters, since the input dimension of the inverse reference model is then relatively small. In this case, there is a low computational effort and an increase in the approximation quality of the model. The step responses are generally not compared directly. The modeled signal is only calculated offline and then used to train the inverse reference model. A comparison only takes place implicitly during the calculation of the time constants via the reference model.

In an advantageous method variant, it can be provided that the inverse reference model is implemented in the form of a map structure, a polynomial calculation structure or as a neural network in a calculation unit, which is part of a diagnostic unit. Ultimately, any nonlinear, static approximator that can be trained with measurement data can be used here, although the ones mentioned are only to be understood as examples. The procedures are generally known from the literature. Nonlinear static approximators are given, for example, in the book “Nonlinear System Identification” by Oliver Nelles, Springer Verlag, 2000.

The method with its previously described method variants can preferably be used for a dynamic diagnosis of sensors which are designed as exhaust gas probes in the exhaust gas duct of the internal combustion engine as part of an exhaust gas monitoring and exhaust gas reduction system or as gas concentration sensors in a supply air duct of the internal combustion engine, the exhaust gas probes being in the form of Broadband lambda sensors or NO _x sensors can be designed, with which an oxygen content in a gas mixture can be determined. For an exhaust gas probe designed as a broadband lambda probe or continuous lambda probe, an actual lambda value is compared with a modeled lambda model value according to the method variants described above for diagnosis. For a nitrogen oxide sensor, the output signal of the nitrogen oxide sensor is evaluated as the actual value, with the model value being determined from a modeled NO _x value. This diagnosis can therefore be used particularly advantageously in gasoline engines or lean-burn engines whose exhaust gas purification system has a catalytic converter and/or devices for nitrogen oxide reduction.

In the aforementioned cases, the input variable is an air-fuel ratio of an air-fuel mixture supplied to the internal combustion engine. Based on a variation of the air-fuel mixture, the previously described method can be used, in particular, to carry out the dynamics diagnosis of exhaust gas sensors.

The diagnostic method, as described above, can also be used advantageously in processes with at least one sensor, in which the process includes any suitable controller and controlled system, with only one system parameter being unknown. Particularly with complex structures of the controller, the transfer function of the closed control loop becomes so complicated that a back transformation into the time domain is practically no longer possible. In these cases, a sufficiently accurate approximation of the unknown route parameter can be found using the previously described method using simulations.

The object relating to the device is achieved in that a diagnostic unit is provided for carrying out the method according to the invention, which has a calculation unit in the form of a non-linear static approximator as an inverse reference model for carrying out the dynamic diagnosis, as described above. On the input side, controller and route parameters as well as a measured maximum slope of the step response of a closed control loop for the sensor to a sudden change in the air-fuel ratio are connected to this. On the output side, as a result of a simulation carried out offline, a time constant is available for the sensor, which can be used for the dynamics diagnosis of the sensor. The functionality of the diagnostic unit can be at least partially software-based, which can be provided as a separate unit or as part of a higher-level engine control system.

• 1 in schematischer Darstellung das technische Umfeld, in dem das erfindungsgemäße Verfahren angewendet werden kann und
• 2 eine Berechnungseinheit zur Durchführung einer Dynamikdiagnose.

The invention is explained in more detail below using an exemplary embodiment shown in the figures. Show it:

• 1 a schematic representation of the technical environment in which the method according to the invention can be used and
• 2 a calculation unit for carrying out a dynamics diagnosis.

1 shows schematically, using an example of a gasoline engine, the technical environment in which the method according to the invention for diagnosing an exhaust gas probe 15 can be used. Air is supplied to an internal combustion engine 10 via an air supply 11 and its mass is determined with an air mass meter 12. The air mass meter 12 can be designed as a hot film air mass meter. The exhaust gas from the internal combustion engine 10 is discharged via an exhaust gas duct 18, with an exhaust gas purification system 16 being provided behind the internal combustion engine 10 in the direction of flow of the exhaust gas. The exhaust gas purification system 16 usually includes at least one catalytic converter.

To control the internal combustion engine 10, an engine control 14 is provided, which, on the one hand, supplies fuel to the internal combustion engine 10 via a fuel metering device 13 and, on the other hand, the signals from the air mass meter 12 and the exhaust gas probe 15 arranged in the exhaust gas duct 18 and an exhaust gas probe 17 arranged in the exhaust gas duct 18 be supplied. In the example shown, the exhaust gas probe 15 determines an actual lambda value of a fuel-air mixture supplied to the internal combustion engine 10. It can be designed as a broadband lambda probe or a continuous lambda probe. The exhaust gas probe 17 determines the exhaust gas composition after the exhaust gas purification system 16. The exhaust gas probe 17 can be designed as a jump probe or binary probe.

For the dynamic diagnosis of the exhaust gas probe 15, the air-fuel ratio (LKV) in the combustion chamber is adjusted suddenly and the maximum slope of the measured air-fuel ratio is determined within a certain period of time after the jump.

To model the air-fuel ratio from the injection to the probe, a first-order filter with a time constant T and a gain K as well as a dead time model with the dead time Tt is used in the error-free case. The first order filter can therefore be described in Laplace space as follows: $$\text{G}\left(\text{s}\right)=\text{K exp}\left(-{\text{T}}_{\text{t}}\text{s}\right)/\left(\text{Ts}+1\right)$$

However, in the event of a fault, a dynamically slow probe acts like an additional first-order filter, so that the entire system can be modeled by a dead time Tt and a second-order filter with two time constants T and Ts for the probe: $$\text{G}\left(\text{s}\right)=\text{K exp}\left(-{\text{T}}_{\text{t}}\text{s}\right)/\left(\left(\text{Ts}+1\right)\left({\text{T}}_{\text{s}}\text{s}+1\right)\right)$$

Based on this transfer function of the controlled system and the transfer function of the controller, the transfer function of the closed control loop can be calculated. Theoretically, the step response and its maximum slope can be determined from the transfer function of the closed control loop by first carrying out a back transformation from the Laplacian to the time domain.

However, depending on the structure of the controller, the transfer function of the closed control loop becomes so complex that a back transformation into the time domain is practically no longer possible. Instead, like this 2 shows schematically, the maximum slope of the step response of the closed control loop for all sensible combinations of controller and system parameters 22.1, 22.2 or 21.1, 21.2 is determined based on simulations. A non-linear static approximator is then trained as an inverse reference model using the controller and system parameters 22.1, 22.2 or 21.1, 21.2 and a maximum gradient 23 simulated therefrom, with the simulated maximum gradient 23 as well as all controller and system parameters 22.1, 22.2 or 21.1, 21.2 in addition to the time constant 24 of the probe (Ts) to be identified can be used as the input of the reference model. In 2 This nonlinear static approximator is shown as a calculation unit 20, which can be designed as a map structure, polynomial calculation or as a neural network. The output of the reference model is the time constant 24 to be identified for the probe ( _TS ).

This inverse reference model as part of a diagnostic unit can be implemented in the engine control 14 and calculates the time constant 24 for the probe (T _S ). Except for the evaluation of the inverse reference model in the control unit, all work steps and calculations (simulation, training) can be carried out offline. Route parameters are, for example, a time constant and a dead time. In the case of a model-based controller, these can also serve as controller parameters as well as an additional parameter for determining the controller speed.

Claims (7)

Method for the dynamic diagnosis of sensors of an internal combustion engine (10), the sensors having a low-pass behavior depending on the geometry and due to aging or contamination have, wherein when an input variable changes on the basis of a modeled and a measured signal, a dynamic diagnosis is carried out and wherein the measured signal is an actual value of an output signal of the sensor and the modeled signal is a model value, characterized in that a maximum slope ( 23) a step response of a closed control loop for the sensor to a sudden change in the input variable for all sensible combinations of controller and system parameters (22.1, 22.2, 21.1 and 21.2) of the model is determined based on simulations and from this a time constant (24) for the sensor derived and the dynamic diagnosis of the sensor is carried out based on the determined time constant (24), an inverse reference model being trained with the parameters of the simulation and the resulting slope. Procedure according to Claim 1 , characterized in that a nonlinear static approximator is used as an inverse reference model, by means of which the relationships between all controller and system parameters (22.1, 22.2, 21.1 and 21.2) on the input side and the time constant (24) on the output side are mapped, with training of the reference model and/or the simulations are carried out offline. Procedure according to one of the Claims 1 or 2 , characterized in that the inverse reference model is implemented in the form of a map structure, a polynomial calculation structure or as a neural network in a calculation unit (20), which is part of a diagnostic unit. Procedure according to one of the Claims 1 until 3 , characterized in that exhaust gas probes (15) in the exhaust duct (18) of the internal combustion engine (10) as part of an exhaust gas monitoring and exhaust gas reduction system or gas concentration sensors in a supply air duct (11) of the internal combustion engine (10) are used as sensors, the exhaust gas probes (15) are designed in the form of broadband lambda sensors or NO _x sensors, with which an oxygen content in a gas mixture can be determined. Method according to one of the preceding claims, characterized in that the input variable is an air-fuel ratio of an air-fuel mixture supplied to the internal combustion engine. Application of the procedure according to one of the Claims 1 until 3 for processes with at least one sensor, in which the process includes any suitable controller and controlled system, with only one system parameter being unknown. Device for dynamic monitoring of sensors, wherein the sensors have a low-pass behavior depending on the geometry and due to aging or contamination, with a change in an air-fuel ratio of an air-fuel mixture supplied to the internal combustion engine (10) based on a modeled and a measured signal Dynamic diagnosis can be carried out in a diagnostic unit and wherein the measured signal is an actual value of an output signal of the sensor and the modeled signal is a model value, characterized in that the diagnostic unit has a calculation unit (20) in the form of a non-linear static approximator as an inverse reference model for carrying out the Dynamics diagnosis according to the Claims 1 until 5 has, the controller and route parameters (22.1, 22.2, 21.1 and 21.2) as well as a measured maximum slope (23) of the step response of a closed control loop for the sensor to a sudden change in the air-fuel ratio are connected on the input side and on the output side as a result of a During the simulation carried out offline, a time constant (24) is available for the sensor, which can be used for the dynamics diagnosis of the sensor.

Images