Method and device for dynamic monitoring of gas sensors

Abstract

In a method for dynamic monitoring of gas sensors of an internal combustion engine, in the event of a change of the gas state variable to be measured, a dynamic diagnosis is carried out based on a comparison of a measured signal which is an actual value of an output signal of the gas sensor and a modeled signal which is a model value. The output signal of the gas sensor is filtered using a high-pass filter and higher-frequency signal components are analyzed.

Claims

1. A method for dynamic monitoring of at least one gas sensor of an internal combustion engine to determine a fault status of the gas sensor during operation of the internal combustion engine, the gas sensor having a low-pass behavior depending on one of geometry, measuring principle, aging, or soiling, the method comprising: measuring, using the gas sensor during the operation of the internal combustion engine in response to an operator torque request, a change of a gas state variable to produce a measured signal, which represents an actual value of the selected gas state variable output by the gas sensor; checking for a sufficient excitation of the measured signal representing the gas state variable by comparing a steepness of a transition of a signal of an excitation model, the signal of the excitation model based on the operator torque request, to a predetermined threshold value; upon detection of the sufficient excitation, performing a dynamic diagnosis of the gas sensor during the operation of the internal combustion engine, wherein the performing the dynamic diagnosis includes: modeling operation of the gas sensor to produce a modeled signal; placing a high-pass filter in series with the gas sensor, the high-pass filter having a limit frequency lower than a predetermined limit frequency of the low pass behavior of the gas sensor when the gas sensor is properly functioning, to filter the measured signal and the modeled signal modeling the gas sensor to produce higher-frequency components of the measured signal and higher-frequency components of the modeled signal; squaring, using a multiplier, and integrating, using an integrator, the higher-frequency components of the measured signal and higher-frequency components of the modeled signal to produce higher-frequency signal energy components of the measured signal and higher-frequency signal energy components of the modeled signal; comparing the higher-frequency signal energy components of the measured signal to the higher frequency signal energy components of the modeled signal; determining a time constant of the sensor based on the comparing; and determining the fault status of the gas sensor as indicating a faulty state of the gas sensor based on the comparing indicating that a low-pass limit frequency of the low-pass behavior of the gas sensor has decreased below the limit frequency of the high-pass filter as a result of aging of the sensor; and regulating operation of the internal combustion engine as a function of the determined time constant.

2. The method as recited in claim 1, wherein the dynamic diagnosis of the gas sensor is performed in the event of a change of an air-fuel ratio of an air-fuel mixture supplied to the internal combustion engine.

3. The method as recited in claim 1, wherein the dynamic behavior of the gas sensor is deduced on the basis of an energy ratio of the calculated higher-frequency energy components.

4. The method as recited in claim 3, wherein the calculated higher-frequency energy components are scaled using a scaling unit.

5. The method as recited in claim 3, wherein the integration of the higher-frequency signal components of the gas sensor and the integration of the higher-frequency signal components of the modeled signal are each carried out with the aid of an individual integration duration, and wherein the start of the integration being triggered in the event of a rising signal edge of the measured signal and in the event of a falling signal edge.

6. The method as recited in claim 5, wherein, the start of the integration begins after an occurrence of a specified signal deviation starting from a specific stationary point.

7. The method as recited in claim 3, wherein the energy ratio is compared to a specified threshold value which represents the dynamics of a marginal sensor, and wherein the threshold value and a model time constant selected in a model for the sensor are interdependent.

8. The method as recited in claim 7, wherein at least one of filter time constants and the threshold value is tracked.

9. The method as recited in claim 7, wherein an iterative identification of the time constant of the sensor is carried out, the model time constant being adapted in steps as a function of the energy ratio.

10. The method as recited in claim 9, wherein an identification of the time constant of the sensor is carried out with the aid of values stored in a characteristic map.

11. The method as recited in claim 9, wherein the gas sensor is one of a gas pressure sensor, a gas temperature sensor, a gas flow rate sensor, or a gas concentration sensor.

12. The method as recited in claim 9, wherein the gas sensor is an exhaust gas sensor in the form of one of a broadband lambda sensor or a NO.sub.x sensor, using which an oxygen content in a gas mixture is determined.

13. A device for dynamic monitoring of at least one gas sensor in one of an exhaust duct of an internal combustion engine or an air supply channel of the internal combustion engine to determine a fault status of the gas sensor during operation of the internal combustion engine, the gas sensor having a low-pass behavior depending on one of geometry, measuring principle, aging, or soiling, the device comprising: a detection unit for measuring, using the gas sensor during the operation of the internal combustion engine in response to an operator torque request, a change of a gas state variable to produce a measured signal, which represents an actual value of the selected gas state variable output by the gas sensor; and a diagnostic unit for performing a dynamic diagnosis during the operation of the internal combustion engine, wherein the diagnostic unit: checks for a sufficient excitation of the measured signal representing the gas state variable by comparing a steepness of a transition of a signal of an excitation model, the signal of the excitation model based on the operator torque request to a predetermined threshold value; upon detection of the sufficient excitation, places a high-pass filter in series with the gas sensor, the high-pass filter having a limit frequency lower than a predetermined limit frequency of the low pass behavior of the gas sensor when the gas sensor is properly functioning, to filter the measured signal and a modeled signal modeling the gas sensor to produce higher-frequency components of the measured signal and higher-frequency components of the modeled signal; squares, using a multiplier, and integrates, using an integrator, the higher-frequency components of the measured signal and higher-frequency components of the modeled signal to produce higher-frequency signal energy components of the measured signal and higher-frequency signal energy components of the modeled signal; compares the higher-frequency signal energy components of the measured signal to the higher frequency signal components of the modeled signal; determines a time constant of the sensor based on the comparing; and determines the fault status of the gas sensor as indicating a faulty state of the gas sensor based on the comparing indicating that a low-pass limit frequency of the low-pass behavior of the gas sensor has decreased below the limit frequency of the high-pass filter as a result of aging or soiling of the sensor; wherein the device regulates operation of the internal combustion engine as a function of the determined time constant.

14. A method for dynamic monitoring of at least one gas sensor of an internal combustion engine, the method comprising: measuring, using the gas sensor during the operation of the internal combustion engine in response to an operator torque request, a change of a gas state variable to produce a measured signal, which represents an actual value of the selected gas state variable output by the gas sensor; checking for a sufficient excitation of the measured signal representing the gas state variable by comparing a steepness of a transition of a signal of an excitation model, the signal of the excitation model based on the operator torque request, to a predetermined threshold value; upon detection of the sufficient excitation, performing a dynamic diagnosis of the gas sensor during the operation of the internal combustion engine, wherein the performing the dynamic diagnosis includes: modeling operation of the gas sensor to produce a modeled signal; comparing higher-frequency signal energy components of the measured signal to higher frequency signal energy components of the modeled signal; and determining a time constant of the sensor based on the comparing; and regulating operation of the internal combustion engine as a function of the determined time constant.

15. The method of claim 14, wherein the performing the dynamic diagnosis further includes placing a high-pass filter in series with the gas sensor, the high-pass filter having a limit frequency lower than a predetermined limit frequency of a low pass behavior of the gas sensor when the gas sensor is properly functioning, to filter the measured signal and the modeled signal modeling the gas sensor to produce higher-frequency components of the measured signal and higher-frequency components of the modeled signal.

16. The method of claim 15, wherein the performing the dynamic diagnosis further includes squaring, using a multiplier, and integrating, using an integrator, the higher-frequency components of the measured signal and higher-frequency components of the modeled signal to produce higher-frequency signal energy components of the measured signal and higher-frequency signal energy components of the modeled signal.

17. The method of claim 15, wherein the performing the dynamic diagnosis further includes determining a fault status of the gas sensor as indicating a faulty state of the gas sensor based on the comparing indicating that a low-pass limit frequency of the low-pass behavior of the gas sensor has decreased below the limit frequency of the high-pass filter as a result of aging of the sensor.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 shows a schematic view of the technical environment in which the method according to the present invention may be applied.

(2) FIGS. 2a and 2b show Bode diagrams for a fast gas sensor and a slow gas sensor, respectively.

(3) FIG. 3 shows a block diagram of a dynamic diagnostic circuit according to the present invention.

(4) FIG. 4 shows a curve diagram for different signal paths.

(5) FIG. 5 shows a flow chart for an iterative identification of a sensor time constant TS.

DETAILED DESCRIPTION OF THE INVENTION

(6) FIG. 1 schematically shows, in an example of a gasoline engine, the technical environment in which the method according to the present invention may be used for diagnosing an exhaust gas sensor 15. Air is supplied to an internal combustion engine 10 via an air supply 11 and its mass is determined using an air flow meter 12. Air flow meter 12 may be designed as a hot-film air flow meter. The exhaust gas of internal combustion engine 10 is discharged via an exhaust duct 18, an emission control system 16 being provided in the flow direction of the exhaust gas downstream from internal combustion engine 10. Emission control system 16 generally includes at least one catalytic converter.

(7) To control internal combustion engine 10, an engine controller 14 is provided, which supplies fuel to internal combustion engine 10 via a fuel metering unit 13, on the one hand, and to which the signals of air flow meter 12 and exhaust gas sensor 15 situated in exhaust duct 18 and an exhaust gas sensor 17 situated in exhaust gas line 18 are supplied, on the other hand. In the example shown, exhaust gas sensor 15 determines a lambda actual value of a fuel-air mixture supplied to internal combustion engine 10. It may be designed as a broadband lambda sensor or as a continuous lambda sensor. Exhaust gas sensor 17 determines the exhaust gas composition downstream from emission control system 16. Exhaust gas sensor 17 may be designed as a bistable sensor or binary sensor.

(8) With respect to improved dynamic monitoring of exhaust gas sensor 15, it is provided according to the present invention that high-pass and low-pass filters are used to check, in the event of a load change of internal combustion engine 10, whether the higher-frequency components of a concentration change are still recognized by exhaust gas sensor 15. Such gas sensors have a typical low-pass behavior, which is dependent, inter alia, on the geometry of their protective tube. In diesel engines, such a protective tube may additionally become sooted, whereby the bandwidth of the sensor decreases. In the time range, the decreasing limiting frequency is expressed in a greater rise time, i.e., the signal edges become flatter with unchanged excitation. Therefore, if a suitable high-pass filter is connected in series with the sensor, it may be recognized in the event of steep load changes on the output signal of the high-pass filter whether the limiting frequency of the low-pass filter is greater or less than the limiting frequency of the high-pass filter.

(9) FIGS. 2a and 2b schematically show, on the basis of Bode diagrams 20, the functional principle. An input spectrum 21 developed in simplified form as a line spectrum including two frequency components and a first-order high-pass filter 23, the transfer function of which may be described by the equation
G(j)=T.sub.Fj/(T.sub.Fj+1)(1)
where T.sub.F is the limiting frequency of the filter (filter limiting frequency 24). If the limiting frequency of exhaust gas sensor 15 exceeds limiting frequency T.sub.F of first-order high-pass filter 23, the series circuit behaves like a bandpass filter, i.e., the high frequencies of input spectrum 21 of exhaust gas sensor 15 are still transmitted and may be detected in output spectrum 22, as schematically shown in FIG. 2a. In contrast, if the limiting frequency of exhaust gas sensor 15 decreases, as a result of a dynamic loss, below limiting frequency T.sub.F of first-order high-pass filter 23 (filter limiting frequency 24), the series circuit blocks all frequencies, so that no frequency components of any type may be measured in output spectrum 22 (FIG. 2b). It is to be noted that this line spectrum is only used to explain the principle. The real frequency spectrum of an exhaust gas sensor 15 may be described by continuously extending frequency components.

(10) In principle, the present invention is not restricted to first-order high-pass filters. Rather, any arbitrary other high-pass filters may also be used. The monitoring method is also applicable if the low-pass filter including exhaust gas sensor 15 is itself parameterized differently, for example, using a limiting frequency instead of the time constant, or has a higher order.

(11) In order that a differentiation may be made between a slow exhaust gas sensor 15 and an inadequate excitation, the change speed of the exhaust gas composition must be judged, which may be carried out, for example, in the case of a broadband lambda sensor on the basis of air mass and fuel mass change. This may be carried out using a similar series circuit of filters. In the case of a broadband lambda sensor, for this purpose, only the above-mentioned masses must be converted into an O.sub.2 concentration and delayed using a low-pass filter, which corresponds to a functional exhaust gas sensor. This low-pass filter is then to be connected in series to a high-pass filter, which has the same transmission function as the real sensor. By comparing the two high-pass filter outputs, the operational reliability of the real sensor may be deduced. In the case of a different gas component, it may be necessary to use an additional untreated emission model.

(12) FIG. 3 shows a block diagram 30 of the functionality of the above-described principle in one preferred method variant. A path for an oxygen concentration 31 measured using exhaust gas sensor 15 is shown in the top part. As a result of a real gas run time and sensor delay 32, which may be described by a reaction time T.sub.I or a first-order low-pass filter having a sensor time constant T.sub.S, an oxygen sensor signal 32.1 results from real oxygen concentration 31. The transmission function of sensor and gas run time 32 results from the following equation, K.sub.S representing an amplification factor for the sensor:
G(j)=K.sub.Sexp(T.sub.tj/(T.sub.Sj+1)(2)

(13) K.sub.S generally corresponds to the multiplicative error or slope error of the sensor, which originates from production variation and aging. However, if the oxygen concentration is not used as the sensor signal, but rather a variable proportional thereto, K.sub.S is a corresponding transmission coefficient for converting the sensor signal into an oxygen concentration and may then also be dimension-afflicted. Subsequently, oxygen sensor signal 32.1 is filtered using a high-pass filter 33, whose transmission function corresponds to that of first-order high-pass filter 23 from FIG. 2a or 2b, and squared using a multiplier 34, which provides a signal which corresponds to a signal power. This signal is subsequently integrated with the aid of an integrator 35, so that a signal energy 35.1 of the higher-frequency energy components of the measured oxygen content is obtained. In a scaling and division unit 36 connected downstream, an energy ratio 36.1 E, which is a proportion measure for the higher-frequency energy components, results from a comparison to a correspondingly prepared signal for a value determined by model.

(14) The preparation of the energy value determined by model is shown in the lower part of block diagram 30. A quotient is formed from an air mass 37 m.sub.L and a setpoint fuel mass 38 m.sub.K for fuel metering 13 after stoichiometric correction in a division unit 39 and a lambda value is calculated. Fuel mass 38 may result from the intended torque, which the driver specifies and which is converted into a fuel quantity. In a conversion unit 40, a calculated oxygen content 40.1 is determined from the lambda value. According to a model 41, using the transmission function
G(j)=K.sub.Sexp(T.sub.IMj/(T.sub.Mj+1)(3)
a modeled oxygen content 41.1 is calculated, T.sub.IM representing a model reaction time and T.sub.M representing a model time constant. Subsequently, modeled oxygen content 41.1 is filtered using a further high-pass filter 33, whose transmission function corresponds to that of first-order high-pass filter 23 from FIG. 2a or 2b, and squared using a further multiplier 34, which provides a signal which corresponds to a signal power. This signal is subsequently integrated with the aid of a further integrator 35, so that a signal energy 35.1 for the higher-frequency energy components of the modeled oxygen content is obtained.

(15) The integration of the squared signal delivers a measure of the energy of the higher-frequency components of particular O.sub.2 signal 32.1, 41.1. The smaller the area under the squared output signal of high-pass filter 33, the slower the sensor or the excitation. Alternatively to the signal energies, variables may also be formed and set in a ratio, which are closely related to the signal energies. For example, instead of the signal energy, the root of the signal energy may also be used or the absolute value of the high-pass filter output signal may be integrated. The ratio of such replacement variables must be 1 if the energy ratio is 1, i.e., if the dynamic behavior of gas sensor and model is identical.

(16) In order that multiplicative errors of the gas sensor and/or the model or its input signals do not corrupt the signal comparison, they are largely eliminated by scaling as follows. Energy ratio 36.1 E may accordingly be represented as follows:

(17) $\begin{array}{cc}E=\frac{{\left(\mathrm{Um}\left(t1+t\right)-\mathrm{Um}\left(t1\right)\right)}^2{}_{t0}^{t0+t}{\left(\mathrm{Ys}\left(t\right)\right)}^2\mathrm{dt}}{{\left(\mathrm{Us}\left(t0-t\right)-\mathrm{Us}\left(\mathrm{t0}\right)\right)}^2{}_{t1}^{t1+t}{\left(\mathrm{Ym}\left(t\right)\right)}^2\mathrm{dt}}& \left(4\right)\end{array}$
therein, U.sub.S and Y.sub.S represent the high-pass input and output, respectively, in the signal path of the sensor and U.sub.M and Y.sub.M represent their counterparts in the model path. This means that U.sub.S corresponds to oxygen sensor signal 32.1 and U.sub.M corresponds to calculated oxygen content 40.1 from the model. The influence of the signal deviation of the excitation is also eliminated by the ratio formation.

(18) The integration in the two signal paths does not necessarily have to be started simultaneously, but rather may be started in such a way that different reaction times do not influence the diagnosis result.

(19) FIG. 4 shows, in a curve diagram 50, points in time 55, 56 t.sub.1, t.sub.0 as examples for the start of the integration in the event of a falling edge. The signal level is shown as a function of time 52 for a signal curve sensor 53 and for a signal curve model 54. Proceeding from a specific stationary point, in the event of the beginning of a signal edge, a certain signal deviation is first covered before the integration is started. The individual integration duration may be of different lengths for the two paths. Furthermore, it may be provided that in the event of a rising signal edge, the integration is started, the procedure then continuing accordingly. A short-term stationary operation prior to the monitoring ensures that the signal edges of sensor and model also originate from the same excitation in the event of reaction time differences. This is important in particular if the installation site of exhaust gas sensor 15 was manipulated.

(20) If rising and falling edges of the sensor signal do not have to be monitored separately, i.e., if so-called pinpointing is not necessary, the method according to the present invention may be simplified per se. It is alternatively possible to start the integration in both paths at an arbitrary point in time t.sub.0 and to carry it out for duration t. The only condition is a sufficient excitation by change of the gas concentration or state variable to be measured, for example, as a result of changing engine operating points. Time interval t may thus include multiple falling and rising edges. The particular maximum and minimum during t is to be detected at both high-pass inputs for the scaling. Energy ratio 36.1 E for the diagnosis and/or identification then reads:

(21) $\begin{array}{cc}E=\frac{{\left(\mathrm{Um},\max -\mathrm{Um},\min \right)}^2{}_{t0}^{t0+t}{\left(\mathrm{Ys}\left(t\right)\right)}^2\mathrm{dt}}{{\left(\mathrm{Us},\max -\mathrm{Us},\min \right)}^2{}_{t0}^{t0+t}{\left(\mathrm{Ym}\left(t\right)\right)}^2\mathrm{dt}}& \left(5\right)\end{array}$ where U.sub.s/m=O.sub.2 concentration measured using the sensor/determined from the model Y.sub.s/m=high-pass filter output signal sensor/model U.sub.s/m,max=maximum of the high-pass filter input signal in interval t U.sub.s/m,min=minimum of the high-pass filter input signal in interval t

(22) t>>T.sub.IM, T.sub.t, i.e., integration duration>>model reaction time T.sub.IM or sensor reaction time T.sub.t

(23) A dynamic error of exhaust gas sensor 15 may be deduced from the value of energy ratio 36.1 E as follows: If the above-mentioned division of signal energies 35.1 results in a value greater than one, the real sensor is faster than the model. If the fraction is less than one, the real sensor is slower than the model.

(24) Above-mentioned energy ratio 36.1 E is compared to a threshold value, which represents the dynamic of a marginal sensor. This threshold value is dependent on how model time constant T.sub.M was selected. If T.sub.M already corresponds to a marginal sensor dynamic, the matching threshold value is at 1, so that its application is omitted. If model time constant T.sub.M corresponds to a sufficiently fast gas sensor, in contrast, for example, a nominal sensor, the matching threshold value is less than 1. For the case in which energy ratio 36.1 Ethe threshold value, exhaust gas sensor 15 is to be considered to be properly functioning; in the case in which energy ratio 36.1 Ethe threshold value, exhaust gas sensor 15 is to be considered to be faulty.

(25) In the event of a lacking excitation, the case may occur in principle that numerator and/or denominator of E in equation (4) will go to zero. The division by zero may be intercepted in that a test for sufficient excitation is carried out. For this purpose, for example, the model path may optionally additionally be implemented for a nominal sensor and it may be checked whether calculated oxygen content 40.1 or model oxygen content 41.1 has a sufficient edge steepness. In one preferred variant of this test, it is established whether the associated high-pass filter output, in the event of positive edges for calculated oxygen content 40.1 or for modeled oxygen content 41.1, exceeds a specific positive threshold value or, in the event of negative edges, falls below a specific negative value.

(26) The present invention may be expanded, so that an iterative identification of sensor time constant T.sub.S is also possible therewith. FIG. 5 shows the functional sequence in a flow chart 60.

(27) For this purpose, in an initialization unit 61, model time constant T.sub.M is initialized using a starting value T.sub.init and then corrected after each integration as a function of energy ratio 36.1 E. For this purpose, a value E.sub.k is determined in function unit 62 by measuring and analysis. The following equation applies:
k=1, T.sub.M,k=T.sub.init(6)

(28) If T.sub.M is greater than sought value T.sub.S, the sensor is faster than the model and energy ratio 36.1 E is greater than one. In this case, T.sub.M is decreased in the next step, in that it is divided by E, which is carried out in calculation unit 63. In the case in which T.sub.M<T.sub.S is E<1 and the following division T.sub.M/E raises new model time constant T.sub.M,k+1, the iteration rule therefore reads:
T.sub.M,1=T.sub.init(7a)
T.sub.M,k+1=T.sub.M,k/E.sub.k(7b)

(29) Where k denotes the kth iteration step. In the course of the integration, E.sub.k converges toward 1 and T.sub.M,k converges toward T.sub.S. It is unimportant whether T.sub.M,k moves strictly monotonously toward T.sub.S or oscillates around T.sub.S with decreasing amplitude. This applies correspondingly for E.sub.k. The iteration is ended when
1<E.sub.k<1+ or |E.sub.k1|<(8).

(30) This check is performed with the aid of query 64. If the above-mentioned condition is met, iteration T.sub.S=T.sub.M, k+1 is output as result 66. If the condition is not yet met, a counter 65 for k is increased by one.

(31) The iterative identification may be carried out online, another excitation occurring in each iteration step. The iteration may also be carried out off-line, the same measurement being analyzed multiple times. If the present invention is used for the identification in this way, for the sensor diagnosis, the final value of T.sub.M,k must be compared to a marginal time constant. A comparison of E to 1 is then obviously no longer reasonable.

(32) The convergence of the identification method may be accelerated or also slowed in that E.sub.k itself is not used for the calculation of T.sub.M,k+1, but rather a variable dependent thereon. The iterative identification method searches for the solution T.sub.M=T.sub.S of the equation E (T.sub.M)=1. Alternatively, other numeric methods for solving so-called fixed point equations may be used for the iterative identification.

(33) Alternatively, as a simple approach for a sensor identification, a characteristic map for estimated time constant T.sub.S may also be spanned. The following combinations of variables come into consideration as inputs for this characteristic map: T.sub.M and E or numerator and denominator of E (cf. (4) and (5)). T.sub.S or the ratio T.sub.S/T.sub.M may be stored in the characteristic map.