METHOD FOR ASCERTAINING AN EXHAUST GAS COMPOSITION OF AN EXHAUST GAS OF AN INTERNAL COMBUSTION ENGINE

Abstract

A method for ascertaining an exhaust gas composition of an exhaust gas of an internal combustion engine with regard to an ammonia fraction and a nitrogen oxides fraction in an exhaust gas system including an SCR catalytic converter. The method includes detecting, using a sensor, a first signal whose magnitude is a function of the nitrogen oxides fraction of the exhaust gas upstream from the SCR catalytic converter, detecting using a sensor a second signal whose magnitude is a function of the ammonia fraction and the nitrogen oxides fraction of the exhaust gas downstream from the SCR catalytic converter, storing the two signals over an observation period, and ascertaining the ammonia fraction and optionally the nitrogen oxides fraction of the exhaust gas downstream from the at least one SCR catalytic converter using a calculation rule that uses the two signals during the observation period as input variables.

Claims

1-10. (canceled)

11. A method for ascertaining an exhaust gas composition of an exhaust gas of an internal combustion engine with regard to an ammonia fraction and a nitrogen oxides fraction in an exhaust gas system including at least one SCR catalytic converter, the method comprising the following steps: detecting, using a sensor, a first signal whose magnitude is a function of the nitrogen oxides fraction of the exhaust gas upstream from the at least one SCR catalytic converter; detecting, using a sensor, a second signal whose magnitude is a function of the ammonia fraction and the nitrogen oxides fraction of the exhaust gas downstream from the at least one SCR catalytic converter; storing the first and second signals at least over an observation period; and ascertaining the ammonia fraction and the nitrogen oxides fraction of the exhaust gas downstream from the at least one SCR catalytic converter using a calculation rule that uses the first and second signals during the observation period as input variables, parameters of the calculation rule being adapted in such a way that a sum signal, calculated using the calculation rule, corresponds as exactly as possible to the detected second signal over the observation period, and the ammonia and the nitrogen oxides fractions being ascertained from the adapted parameters.

12. The method as recited in claim 11, wherein the calculation rule assumes an efficiency of the at least one SCR catalytic converter, with regard to a nitrogen oxides reduction, that is constant during the observation period.

13. The method as recited in claim 11, wherein the calculation rule ascertains the nitrogen oxides fraction downstream from the at least one SCR catalytic converter based on dynamics of the first signal.

14. The method as recited in claim 11, wherein the calculation rule assumes an ammonia fraction of the exhaust gas, downstream from the at least one SCR catalytic converter, that is constant during the observation period.

15. The method as recited in claim 11, wherein a most exact correspondence possible is reached by minimizing a sum of squared errors of a calculated sum signal and a detected signal, an ascertainment, based on the minimized sum of squared errors, of whether the adaptation of the parameters was successful.

16. The method as recited in claim 15, further comprising ascertaining a coefficient of determination, which indicates how strongly scattering of the calculated sum signal may be correlated with a scattering of the detected first and second signals, and an ascertainment, based on the coefficient of determination, of whether the adaptation of the parameters was successful.

17. The method as recited in claim 16, wherein the ammonia fraction is assumed to be zero, and the assumption is assessed as correct when the coefficient of determination exceeds a predeterminable threshold and/or when the minimized sum of squared errors falls below a predeterminable threshold value.

18. A processing unit configured to ascertain an exhaust gas composition of an exhaust gas of an internal combustion engine with regard to an ammonia fraction and a nitrogen oxides fraction in an exhaust gas system including at least one SCR catalytic converter, the processing unit configured to: detect, using a sensor, a first signal whose magnitude is a function of the nitrogen oxides fraction of the exhaust gas upstream from the at least one SCR catalytic converter; detect, using a sensor, a second signal whose magnitude is a function of the ammonia fraction and the nitrogen oxides fraction of the exhaust gas downstream from the at least one SCR catalytic converter; store the first and second signals at least over an observation period; and ascertain the ammonia fraction and the nitrogen oxides fraction of the exhaust gas downstream from the at least one SCR catalytic converter using a calculation rule that uses the first and second signals during the observation period as input variables, parameters of the calculation rule being adapted in such a way that a sum signal, calculated using the calculation rule, corresponds as exactly as possible to the detected second signal over the observation period, and the ammonia and the nitrogen oxides fractions being ascertained from the adapted parameters.

19. A non-transitory machine-readable memory medium on which is stored a computer program for ascertaining an exhaust gas composition of an exhaust gas of an internal combustion engine with regard to an ammonia fraction and a nitrogen oxides fraction in an exhaust gas system including at least one SCR catalytic converter, the computer program, when executed by a processing unit, causing the processing unit to perform the following steps: detecting, using a sensor, a first signal whose magnitude is a function of the nitrogen oxides fraction of the exhaust gas upstream from the at least one SCR catalytic converter; detecting, using a sensor, a second signal whose magnitude is a function of the ammonia fraction and the nitrogen oxides fraction of the exhaust gas downstream from the at least one SCR catalytic converter; storing the first and second signals at least over an observation period; and ascertaining the ammonia fraction and the nitrogen oxides fraction of the exhaust gas downstream from the at least one SCR catalytic converter using a calculation rule that uses the first and second signals during the observation period as input variables, parameters of the calculation rule being adapted in such a way that a sum signal, calculated using the calculation rule, corresponds as exactly as possible to the detected second signal over the observation period, and the ammonia and the nitrogen oxides fractions being ascertained from the adapted parameters.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG. 1 shows an exhaust gas system, as may be used within the scope of the present invention, in a simplified schematic illustration,

[0017] FIG. 2 shows one advantageous embodiment of a method according to the present invention in the form of a greatly simplified flowchart.

[0018] FIG. 3 shows a diagram that illustrates one advantageous embodiment of a method according to the present invention.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

[0019] FIG. 1 schematically shows an arrangement that includes an exhaust gas system, as may be used within the scope of the present invention, denoted overall by reference numeral 100.

[0020] Arrangement 100 includes an internal combustion engine 1 and multiple catalytic converters 11, 12, 13 situated downstream from internal combustion engine 1. In the illustrated example, multiple sensors 17, 18, 19 are provided, in particular sensors that are configured to ascertain a composition of the exhaust gas in the exhaust gas system. Sensors 17, 18, 19 are each connected in a data-conveying manner to a processing unit 20, for example a control unit of a motor vehicle that includes arrangement 100.

[0021] In the illustrated example, processing unit 20 is also connected in a data-conveying manner to internal combustion engine 1 and to external devices 14, 15, 16, for example secondary air supply lines, exhaust gas burners, reducing agent metering units, electric heating elements, or the like, which are in each case associated with one of catalytic converters 11, 12, 13. In particular, electric heating elements may also be situated directly in the catalytic converter or within a housing of the catalytic converter.

[0022] An exhaust gas 10 that is generated by internal combustion engine 1 is supplied to catalytic converters 11, 12, 13 in succession in order to be cleaned or detoxified therein. Each of catalytic converters 11, 12, 13 may be provided for a specific detoxification or for multiple simultaneous detoxifications. For example, a first catalytic converter 11, which may be situated near internal combustion engine 1, may be designed as a three-way catalytic converter (TWC), while a second catalytic converter 12 and a third catalytic converter 13 may encompass other catalytic converters and/or cleaning components such as NO.sub.x storage catalytic converters, SCR catalytic converters, particulate filters, or the like. However, second and third catalytic converters 12, 13 may likewise include one or multiple further TWCs. In addition, first catalytic converter 11 may also include one or multiple other cleaning components, and does not necessarily have to be designed as a TWC.

[0023] It is understood that for exhaust gas systems of a diesel engine, correspondingly adapted catalytic converters may be used, in particular oxidation catalytic converters, lean NO.sub.x traps (LNTs), SCR catalytic converters, particulate filters, and the like. For better readability, the present invention is explained in greater detail here based on an example with a gasoline engine as internal combustion engine 1.

[0024] Within the scope of the present invention, at least one of catalytic converters 11, 12, 13 is designed as an SCR catalytic converter that is supplied with ammonia or an ammonia-producing reducing agent, in particular urea solution, with the aid of a reducing agent metering unit (one of external devices 14, 15, 16). As explained at the outset, nitrogen oxides contained in the exhaust gas supplied to exhaust gas system 120 are reduced to nitrogen and water with the aid of the reducing agent, thus supplied, which may also be stored, at least in part, in the SCR catalytic converter. For better readability, in the following discussion it is assumed that first catalytic converter 11 is designed as a diesel oxidation catalytic converter (DOC), second catalytic converter 12 is designed as an SCR catalytic converter, and the third catalytic converter is likewise designed as an SCR catalytic converter. In such a case, sensors 17, 18, 19 are each designed as NO.sub.x sensors that output a signal whose intensity (for example, the magnitude of a signal voltage) is a function of a concentration of non-elemental nitrogen compounds in the particular analyzed exhaust gas. These types of sensors are typically not selective for nitrogen oxides, but instead have cross-sensitivity to other nitrogen compounds, in particular ammonia. External devices 15, 16, situated in each case upstream from SCR catalytic converters 12, 13, are considered here as metering devices for ammonia or urea solution.

[0025] FIG. 2 schematically illustrates one advantageous embodiment of a method according to the present invention in the form of a simplified flowchart, denoted overall by reference numeral 200.

[0026] When device components are referenced in the description of method 200, these relate in particular to components discussed in conjunction with FIG. 1. For the sake of simplicity and to avoid repetitions, method 200 is described with reference to first SCR catalytic converter 12, but may also be carried out for second SCR catalytic converter 13 or for both SCR catalytic converters 12, 13.

[0027] FIG. 3 shows a diagram illustrating the application of one advantageous embodiment of a method according to the present invention, for example method 200 from FIG. 2, and which is denoted overall by reference numeral 300. Time t is plotted on the abscissa of diagram 300, while the ordinate represents the magnitude of the signal of a sensor 17, 18, 19.

[0028] Method 200 described here is based on stored measured data of sensors 17, 18 for a predefined time period (also referred to as a register), which at the point in time of the evaluation is in the past, i.e., maps the history. Based on the history, parameters of a simple model of SCR catalytic converter 12 are optimized, and the fraction of nitrogen oxides and ammonia in the exhaust gas downstream from catalytic converter 12 is ascertained from the optimized parameters. Curve patterns r.sub.NOx.sup.Us and r.sub.NOx.sup.Ds illustrated in FIG. 3 are an example of such a history that is stored in the register.

[0029] The mentioned time period may be 3 s, a value being stored every 100 ms, for example. In the example, 31 measured values result for a signal, for example r.sub.NOx.sup.U (=measured NO.sub.x concentration prior to or upstream from SCR catalytic converter 12). The measured NO.sub.x concentrations upstream from SCR catalytic converter 12 (r.sub.NOx.sup.Us) and downstream from SCR catalytic converter 12 (r.sub.NOx.sup.DsSens) are stored as measuring channels in the described register.

[0030] At a present point in time, for the input data and output data stored in this way a simple model is calculated which maps output values r.sub.NOx.sup.DsSens (1 . . . n) based on input values (r.sub.NOx.sup.Us (1 . . . n)):

r.sub.NOxSens.sup.DsMod(t)=k*r.sub.NOx.sup.Us(t−Δt)+r.sub.NH3.sup.DsMod  (1)

[0031] where the parameters have the following meanings:

[0032] k: 1—catalytic converter efficiency, i.e., factor for NO.sub.x slip (constant over the observation period)

[0033] Δt: gas travel time (constant over the observation period)

[0034] r.sub.NH3.sup.DsMod: modeled NH.sub.3 concentration downstream from the SCR catalytic converter (constant over the observation period) and the output value

[0035] r.sub.NOxSens.sup.DsMod modeled value at the location of the NO.sub.x sensor downstream from SCR catalytic converter 12 at any point in time.

[0036] The values for the parameters are determined at each evaluation point in time based on the stored data, using an optimization method. The objective of the optimization method is to find the parameters for which the sum of squared errors SSE is at a minimum.

[00001]$\begin{array}{cc}\mathrm{SSE}=\frac{1}{2}.Math.{\left(r_{NOx}^{DsSens}-\left(k*r_{NOx}^{Us}\left(t-\Delta t\right)+r_{NH3}^{DsMod}\right)\right)}^2& \left(2\right)\end{array}$

[0037] Conventional optimization methods such as the gradient method may be used for this purpose. Since this involves a greatly simplified model for a short observation period, it is also possible to find the optimal parameters on the control unit in a sufficiently short time, despite limited resources. The end of the optimization is reached (as is customary in the gradient method) when optimization criterion SSE is no longer greatly improved to a sufficient degree (ΔSSE<threshold), or the maximum number of iterations have been carried out. It is understood that other optimization algorithms may also be used.

[0038] In addition, by use of equation (3) below, the change in Δ(t) or its direction with respect to the optimization is ascertained in order to quickly arrive at an appropriate value for the gas travel time.

[00002]$\begin{array}{cc}\Delta \left(\Delta t\right)=-{\gamma }_2.Math.\mathrm{\Sigma e}.Math.\frac{k\left(d{r}_{N{O}_x}^{Us}\right)}{dt}& \left(3\right)\end{array}$

[0039] where e=r.sub.NO.sub.x.sup.DsSens−r.sub.NO.sub.x.sup.DsMod

[0040] Depending on the operating situation and as a function of the recorded measured data, it is possible, with more or less good results, to achieve the optimization task. To establish at the end of the optimization whether the optimization was sufficiently good, optimization criterion SSE may be checked for a threshold (SSE<SSE threshold value=>optimization was sufficiently good). For a 3 s observation time, a value of 500 approximately corresponds, for example, to an average deviation of 4 ppm, and is a good guideline value.

[0041] Alternatively, the coefficient of determination may also be calculated and used according to formula (4):

[00003]$\begin{array}{cc}{R}^2=1-.Math.\frac{{\left(r_{NOx}^{DsSens}-r_{NOxSens}^{DsMod}\right)}^2}{{\Sigma \left(r_{NOx}^{DsSens}-\stackrel{\_}{r_{NOx}^{DsSens}}\right)}^2}& \left(4\right)\end{array}$

[0042] where r.sub.NOx.sup.DsSens is the average value of the sensor values downstream from the SCR.

[0043] Coefficient of determination R.sup.2 indicates how strongly the scattering of NO.sub.x model value r.sub.NOxSens.sup.DsMod may be correlated with the scattering of sensor values r.sub.NOx.sup.DsMod. The value is 1 when this is the case 100% of the time, i.e., the signals are thus strongly correlated. R.sup.2 assumes smaller values the more poorly the model depicts reality. If the coefficient of determination is above a threshold (0.5, for example) after the optimization, it may be assumed that the optimization was sufficiently good, and the parameters in question are highly likely to be very close to the actual NO.sub.x and NH.sub.3 values. The NH.sub.3 value may be read off directly) The NO.sub.x fraction may be calculated from term k *r.sub.NOx.sup.Us (t−Δt).

[0044] In principle, the described method can function only when the two signal components (NO.sub.x and NH.sub.3) have a different characteristic or different dynamics. Since it is assumed that the NH.sub.3 signal does not change during the short observation period, the NO.sub.x signal (if >0) must have certain dynamics. This is checked by determining the gradients (i.e., difference quotients (cf. formula 5)) of NO.sub.x sensor signal r.sub.NOx.sup.Us upstream from the catalytic converter during the stored time period. These gradients are compared to a threshold value (25 ppm/s, for example). If the proportion of the gradient values higher than this threshold is above a threshold value (80%, for example), the dynamics of the input signal are sufficiently high, if the SCR catalytic converter allows NO.sub.x to pass through, that output signal r.sub.NOx.sup.DsSens likewise has sufficiently high dynamics to distinguish from NH.sub.3.

[00004]$\begin{array}{cc}\frac{d{r}_{NOx}^{Us}}{dt}{|}_i=\frac{r_{NOx}^{Us}{|}_i-r_{NOx}^{Us}{|}_{i-1}}{dt}& \left(5\right)\end{array}$

[0045] where i is the index of a loop through the stored register, and dt is the sampling rate.

[0046] Since the tolerances of the NO.sub.x sensor have a relatively strong effect for very small NO.sub.x sensor values r.sub.NoX.sup.DsSens, the optimization and signal assessment may be limited to values above a threshold value (10 ppm, for example). In this case, it is meaningful to start the assessment only when a sufficiently high proportion of NO.sub.R sensor values (at least 70%, for example) above the threshold value are present.

[0047] In one alternative embodiment, the method is carried out without an offset, i.e., r.sub.NH3.sup.DsMod. In this variant, the optimization is successful only when primarily NO.sub.R is involved and the NH.sub.3 fraction is very small. The success may be read off at the coefficient of determination (cf. formula 4). This means that if the coefficient of determination is above a threshold (0.8, for example), the signal involves primarily NO.sub.R. If the coefficient of determination is below a threshold, it is highly likely that a fairly large NH.sub.3 fraction is also present. Thus, in contrast to the method described above, this embodiment results in a binary statement. This information may also be useful for the metering strategy.

[0048] The parameters to be determined, as described above, may be determined by an optimization. As an alternative or in addition to a determination of the starting values for the optimization, the parameters may be calculated based on the stored measured data (cf. FIG. 3):

[0049] Delay time Δt: The delay time is calculated based on the time interval of the high points or low points of the NO.sub.R sensor signal upstream and downstream from the SCR catalytic converter (r.sub.NOx.sup.Us and r.sub.NOx.sup.DsSens) For recognizing the high points or low points in the recorded measured data, the points in time of the changes in algebraic sign of the gradients/difference quotients are determined (positive gradient followed by negative gradient=>high point, and vice versa). Based on these points in time, the delay time may be calculated separately for each high point and each low point. This represents step 201 of method 200 as illustrated in FIG. 2.

[0050] For the subsequent optimization, multiple gas travel times for the calculation, for example three different gas travel times Δt.sub.1, Δt.sub.2, Δt.sub.3 that differ slightly from one another, are established in a step 202 based on the ascertained delay time(s).

[0051] One of the gas travel times established in step 202 is selected in a step 203 for the further calculation.

[0052] Efficiency factor k: The efficiency factor may be calculated based on the maximum changes in signals r.sub.NOx.sup.Us and r.sub.NOx.sup.DsSens during the observation period. This takes place in a step 204.

[00005]$\begin{array}{cc}k=\frac{\Delta r_{NOx}^{DsSens}}{\Delta r_{NOx}^{Us}}& \left(6\right)\end{array}$

[0053] where Δr.sub.NOx.sup.DsSens=r.sub.NOx.sup.DsSens_Max−r.sub.NOx.sup.DsSens_Min and Δr.sub.NOx.sup.Us=r.sub.NOx.sup.Us_max−r.sub.NOx.sup.Us_Min, as illustrated in FIG. 3.

[0054] Based on calculated delay time Δt and determined efficiency factor k, optimal offset r.sub.NH3.sup.DsMod may be determined as the average value of all differences of the sensor value and of modeled NO.sub.x fraction r.sub.NOx.sup.DsSens−k*r.sub.NOx.sup.Us (t−Δt). In the illustrated example, this likewise takes place in step 204.

[00006]$\begin{array}{cc}{r}_{\mathrm{NH}3}^{DsMod}=\stackrel{\_}{r_{NOx}^{DsSens}{|}_{1.Math.n}-k*\stackrel{.fwdarw.}{r_{NOx}^{Us}}{|}_{1.Math.n}}& \left(7\right)\end{array}$

[0055] where {right arrow over (r.sub.NOx.sup.Us )} is the NO.sub.x signal, shifted by Δt, upstream from the catalytic converter (=r.sub.NOx.sup.Us (t−Δt)).

[0056] The values determined in this way may be used directly in the model or used as initialization values for a subsequent optimization. In the latter case, which corresponds to the example illustrated in FIG. 2, sum of squared errors SSE is also calculated in step 204, based on the initialization values.

[0057] For two values k.sub.pos and k.sub.neg that deviate in various directions from the initial value of efficiency factor k by the same small amount, associated offsets r.sub.NH3pos and r.sub.NH3neg and respective SSE.sub.pos or SSE.sub.neg are calculated in a step 205.

[0058] The algebraic sign for the further optimization of efficiency factor k is subsequently established in a step 206 by selecting the algebraic sign for which the smaller sum of squared errors resulted from SSE.sub.pos and SSE.sub.neg in step 205. In addition, the magnitude of further change Δk of efficiency factor k is established in step 206, based on the ratio of the change in sum of squared errors ΔSSE to the change in efficiency factor Δk.

[0059] This change in efficiency factor k is implemented, and in turn offset r.sub.NH3 and SSE are calculated, in step 207.

[0060] It is subsequently checked in a step 208 whether SSE has once again increased and/or whether a maximum number of iteration steps has been reached. If neither is the case (output “n”), method 200 returns to step 206, the most recently calculated SSE being defined as a new reference (step 209). In contrast, if it is established in step 208 that the SSE has once again increased and/or that the maximum number of iterations has been reached (output “y”), the lowest of ascertained sum of squared errors SSE is identified, and associated values Δt, k, and r.sub.NH3 are stored as (preliminary) optimal values, in a step 210.

[0061] A check is then made in a step 211 as to whether all gas travel times defined in step 202 have been used as an initial value for the calculation (step 204). If this is the case (output “y”), the method ends in a step 212 (and outputs the ascertained optimal values). Otherwise, (output “n”), the method returns to step 203 and selects a further of the multiple gas travel times, not yet used, for the further calculation.

[0062] As described above, there are particularly suitable operating conditions (in particular, high dynamics upstream from the SCR catalytic converter) under which the determination of the NO.sub.x/NH.sub.3 fraction functions particularly well. At the same time, it is not always guaranteed that the model is sufficiently accurate (SSE<threshold or coefficient of determination>threshold). As a result, a valid NH.sub.3 and NO.sub.x model signal may be calculated only occasionally. Depending on the application, it may be meaningful to “freeze” the NH.sub.3 signal (i.e., assume it to be constant) at times for which no new signal is calculated, and to calculate the NO.sub.x value based on the difference between NO.sub.x sensor value r.sub.NOx.sup.DsSens and frozen NH.sub.3 value r.sub.NH3.sup.Freeze. The maximum frozen value may be as great as the present sensor value (since no more NH.sub.3 can ever be present than the signal indicated by the NO.sub.x sensor downstream from the SCR catalytic converter). The frozen value thus decreases along with the sensor value as soon as the latter falls below the previously frozen value. Only when a new valid value is available, is this value adopted.

[0063] Alternatively, offset r.sub.NH3.sup.DsMod may be filtered prior to the freezing. In particular, conventional exponentially weighted moving average (EWMA) filters are suited for this purpose, although other filters may also be used.

[0064] In a further embodiment, instead of the simple model for the observation period, a more complex model, for example a kinetic model, is calculated. In this case, the starting filling level may be used as a variable parameter. Either only the NO.sub.x fraction is included in the calculation (and the NH.sub.3 fraction continues to be held constant), or the NH.sub.3 fraction is also included in the calculation. If the optimization is successful, in this case the filling level may be directly corrected via a larger or smaller metered quantity.

[0065] In a further embodiment, the optimizer is started using the values of the preceding optimization run. As a further alternative initialization value for Δt, the volume flow of the exhaust gas may also be calculated from the geometry, mass flow, pressure, and temperature, and the theoretical delay may be calculated from the volume flow. In many catalytic converter models this variable is present as the gas travel time.

[0066] It is understood that some of the mentioned steps of method 200 may also be carried out in another, for example reverse, order. In addition, the method is not necessarily limited to being carried out in a stepwise manner, so that the described “steps” of method 200 may possibly also be carried out, in whole or in part, in parallel to one another or simultaneously and continuously.