GB2503246A

GB2503246A - Method and apparatus for estimating a dosing-error in a selective catalytic reduction system

Info

Publication number: GB2503246A
Application number: GB1210924.5A
Authority: GB
Inventors: Bastian Maass; Thomas Steffen; David Heaton; Michail Soumelidis
Original assignee: Perkins Engines Co Ltd
Current assignee: Perkins Engines Co Ltd
Priority date: 2012-06-20
Filing date: 2012-06-20
Publication date: 2013-12-25
Anticipated expiration: 2032-06-20
Also published as: US20150143884A1; CN104619962A; GB2503246B; GB201210924D0; CN104619962B; EP2864603A1; WO2013190315A1

Abstract

A method and apparatus for determining if a selective catalytic reduction (SCR) device is being mis-dosed with a reductant comprises (i) determining a NOx readout difference figure from the difference between an expected exhaust gas output from the SCR and a measured exhaust gas state from the SCR device; and (ii) determining a mis-dosing indication figure which indicates if the SCR device is being under, over or correctly dosed by cross-correlating the NOx readout difference with a NOx sensor sensitivity figure, wherein the NOx sensor sensitivity figure is an indication to how sensitive the measured exhaust gas output state is to dosing errors. Also shown is a method and apparatus that comprises determining a first cross correlation figure by cross-correlating a measured exhaust gas figure with a NOx sensor sensitivity figure; determining a second cross-correlation figure by cross correlating an expected exhaust gas figure with a NOx sensor sensitivity figure; and determining a mis-dosing indication figure from the difference between the first and second cross-correlation figures.

Description

Method and apparatus for estimating a dosing-error in a selective catalytic reduction system This disclosure relates to a method and apparatus for estimating a dosing error in a selective catalytic reduction system.

Background

Selective catalytic reduction (SCR) systems may be used to convert nitrous oxides (NOx) which may be produced, for example, by an internal combustion engine, intc less harmful emissions, such as nitrogen and water. The SOR system may comprise a catalyst which facilitates a reaction between the NOx, which may be present in a gas stream passing through the OCR system, and a reductant in crder substantially to remove the NOx from the gas stream.

The reduotant may be added to the gas stream and absorbed cnto the catalyst before it reacts with the NOx in the gas stream passing through the SCR system. Where the reductant used is ammonia, It may be added to the gas stream as, for example, anhydrcus ammcnia, agueous ammonia or urea which thermally decomposes into ammonia within the OCR system before being absorbed onto the catalyst.

When the OCR system is dosed with reductant correctly, the ammonia storage level on the catalyst may be maintained at an optimum level and the reaction between the ammonia and NOx may eliminate nearly all of the NOx and ammonia. If the SCR system is over dosed, there may be more ammonia within the SCR system than can be absorbed onto the catalyst, which may result in ammonia being emitted from the SCR system (commonly known as ammonia slip') . Ammonia emissions may be undesirable as they can be very harmful to the environment. If the system is under-dosed, there may be insufficient ammonia absorbed onto the catalyst to react with all of the NOx passing through the SCR system, which may result in unprocessed NOx being emitted from the SCR system. This reduces the conversion efficiency of the SCR system and is therefore also undesirable.

It may, therefore, be desirable to control the level of dosing so that the ammonia storage level is maintained at an optimum level. However, the amount of ammonia stored is not directly measurable. Consequently, it may be necessary to estimate the storage state within the catalyst.

The storage state of the catalyst may depend upon a number of factors, which might include the temperature of the catalyst and the amount of NOx passing into the SCR system.

These factors may be monitored using, for example, a downstream NOx sensor and an upstream NOx sensor, a mass flow sensor and a temperature sensor, the readings of which may be used to estimate the ammonia storage level within the catalyst. However, these sensors have a limited accuracy, and are subject to long term sensor drift, making it difficult accurately to determine the state of the gas stream over time. Furthermore, the dosing device itself may have limited accuracy, causing a discrepancy between the dosing level determined by the control system and the actual dosing level applied to the catalyst.

Consequently, even if the estimated storage state of the catalyst is initially accurate, over time the accuracy of the estimate, and therefore the accuracy of the urea dosing, may diminish due to inaccurate data readings. As a result of the conservation of mass (for ammonia and NOx) , even small errors may increase over time, and the actual storage level may differ significantly from the estimated storage level.

These inaccuracies may lead to a discrepancy in the order of 10% or more between the ideal dosage which should be applied to the OCR system in order to achieve maximum conversion efficiency with minimal ammonia slip, and the actual level of dosing which is applied to the SCR system. This may result in NOx or ammonia slip emissions as a consequence of under-dosing or over-dosing.

JS patent application 2010/0024389 describes a dosing control system for an OCR system wherein a catalyst ammonia storage model is used to estimate the current ammonia storage level within a catalyst and the theoretical ideal ammonia storage level. Dosing is then applied to the catalyst at a level which brings the estimated storage level towards the theoretical ideal storage level.

An analysis module estimates an expected NOx out value based upon the catalyst ammonia storage model and measurements of exhaust conditions, such as NOx in, exhaust temperature and exhaust flow rate. An error term is then found by determining the difference between the estimated NOx out and a NOx out value measured by a downstream NOx sensor. This difference signal is fed-back to the analysis module to improve dosing accuracy by mcdifying the catalyst ammonia storage model to correot the dosing error.

However, NOx sensors are cross sensitive to ammonia to a degree which may vary over time. Consequently, a high reading from the NOx out sensor may be oaused either by untreated NOx, caused by under-dosing, or by ammonia slip, caused by over-dosing.

Thus, an elevated reading frcm the downstream NOx sensor might indicate either: too much NOx being output, or ammonia being output. Since the first of these scenarios reguires an increase in dosing and the second requires a decrease in dosing, the dosing level of the SCR system may not be reliably improved.

One solution to address the issues relating to the cross sensitivity of NOx sensors with ammonia, is to use a downstream ammonia sensor in combination with a downstream NOx sensor. With such an arrangement it may be possible to compensate for the cross-sensitivity and determine if there is untreated NOx or ammonia slip being output from the SCR system and correct the dosing error accordingly.

However, in addition to not taking into consideration any of the other factors relating to dosage errors, such as sensor drift and variations in levels of ammonia which may be stored on the catalyst caused by catalyst temperature changes, ammonia sensors are also not as reliable as NOx sensors and such systems are therefore not very robust.

Furthermore, the additional sensor increases the system complexity and cost. Therefore, it may be undesirable to use ammonia sensors for controlling the dosing of an 5CR system.

Summary

The disclosure provides: a method for determining if a selective catalytic reduction (SCR) device which is dosed with a reductant is being mis-dosed, the method comprising the steps of: determining a NOx readout difference figure from the difference between an estimated exhaust gas output state from the 5CR device and a measured exhaust gas output state from the OCR device; and determining a mis-dosing indication figure which indicates if the OCR device is being under-dosed, correctly dosed or over-dosed by cross-correlating the NOx readout difference with a NOx sensor sensitivity figure, wherein the NOx sensor sensitivity figure is an indication of how sensitive the measured exhaust gas output state is to dosing errors.

The disclosure also provides: a method for determining if a selective catalytic reduction (OCR) device which is dosed with a reductant Is being mis-dosed, the method comprising the steps of: determining a first cross-correlation figure by cross-correlating a measured exhaust gas output state with a NOx sensor sensitivity figure, whereIn the NOx sensor sensitivity figure is an indication of how sensitive the measured exhaust gas output state is to dosing errors; determining a second cross-correlation figure by cross-correlatIng an estimated exhaust gas output state from the OCR device with the NOx sensor sensitivity figure; and determining a mis-dosing indication figure which indicates If the OCR device is being under-dosed, correctly dosed or over-dosed from the difference between the first cross-correlation figure and the second cross-correlation figure.

The disclosure also provides: a controller to determine if a selective catalytic reduction (3CR) device that is dosed with a reductant is being mis-dosed, the controller being configured to determine a NOx readout difference figure from the difference between an estimated exhaust gas output state from the OCR device and a measured exhaust gas output state from the OCR device; and determine a mis-dosing indication figure which indicates if the 3CR device is being under-dosed, correctly dosed or over-dosed by cross-correlating the NOx readout difference with a NOx sensor sensitivity figure, wherein the NOx senscr sensitivity figure is an indication of how sensitive the measured exhaust gas output state is to dosing errors.

The disclosure also provides: a controller to determine if a selective catalytic reduction (3CR) device that is dosed with a reductant is being mis-dosed, the controller being configured to determine a first cross-correlation figure by cross-correlating a measured exhaust gas output state with a NOx sensor sensitivity figure, wherein the NOx sensor sensitivity figure is an indication of how sensitive the measured exhaust gas output state is to dosing errors; determine a second cross-correlation figure by cross-correlating an estimated exhaust gas output state from the OCR device with the NOx sensor sensitivity figure; and determine a mis-dosing indication figure which indicates if the OCR device is being under-dosed, correctly dosed or over-dosed from the difference between the first cross-correlation figure and the second cross-correlation figure.

Figures Figure 1 shows a schematic drawing of an engine unit comprising an SCR device; Figure 2 shows a schematic drawing of a controller that may be used to control the dosing of the 5CR device shown in Figure 1; Figure 3 shows a graphical representation of the NOx output levels that may be estimated at a single point in time by the estimator unit within the controller shown in Figure 1 using a catalyst model, an over-dosing model and an under-dosing model; and Figure 4 shows an example vehicle within which the engine unit shown in figure 1 may be used.

Detailed description

An 5CR device may be used for a variety of applications where a reduction in NOx levels in a gas stream is desired.

Such applications may include, but are not exclusive to, boilers, gas turbines and internal combustion engines, for example diesel engines.

Figure 1 shows an internal combustion engine 10 with an SCR device 20 at the exhaust of the engine 10. The 5CR device in this arrangement may be dosed by injecting urea into the exhaust gas upstream of the SCR device 20 with an injector 40. However, any other suitable dosing agent, for example anhydrous or aqueous ammonia, may alternatively be used and added to the SCR device 20 using any suitable technigue known to the skilled person, or ammonia may be created in a separate part of the system, such an ammonia reactor -A first (or upstream) NOx sensor 42, mass flow sensor 44 and temperature sensor 46 may be arranged to measure the state of the exhaust gas upstream of the SCR device 20.

Additional or alternative sensors may be used to measure the state of the exhaust gas upstream of the SCR device 20.

Aiternatively, the state of the upstream exhaust gas may be estimated from measured engine parameters, for example engine speed, fuel injection quantity, altitude and ambient temperature.

The measured or estimated state of the exhaust gas upstream of the SOP. device 20 may include at least one of NOx concentration, mass flow rate and temperature.

A second (or downstream) NOx sensor 48 may be arranged to measure the NOx concentration of the exhaust gas downstream of the OCR device 20. Additional sensors may also be provided downstream of the OCR device 20 in order to measure other aspects of the exhaust gas state. Different sensor positions for the second NOx sensor 48 are possible, for example within the OCR device 20 in a mid-brick position.

It may be desirable to control the level of urea dosing in order to maintain an optimum ammonia storage level for minimising NOx emissions and ammonia slip in the exhaust gas output from the OCR device 20. A controller 30 may be used to control the urea injector 40 to this end.

Figure 2 shows details of the controller 30, which includes an estimator unit 32, an error calculation unit 34 and a dosing calculation unit 36.

Data relating to the exhaust gas state 31 of the exhaust gas flow into the SCR device 20 may be fed into the estimator unit 32 and the dosing calculation unit 36. In this example, the exhaust gas state 31 may include at least one of NOx concentration, mass flow rate and temperature, which may be read by the upstream NOx sensor 42, mass flow rate sensor 44 and temperature sensor 46. However, the exhaust gas state 31 may include different measurements of the exhaust gas input to the SCR device 20, or may instead be measured engine parameters, from which the estimator unit 32 may estimate an exhaust gas state.

In the example shown in figures 1 and 2, the catalyst temperature may be estimated from the exhaust gas temperature upstream of the SCR device 20. However, the temperature may alternatively be obtained from a temperature sensor within the catalyst, or estimated from the downstream exhaust gas temperature, the upstream and downstream exhaust gas temperatures, or any other direct or indirect temperature measurement or estimation technigues which would be known to the skilled person.

Within the estimator unit 32, the exhaust gas state 31, the measured or estimated catalyst temperature and the current dosing level 39 may be fed into a catalyst model which may determine an estimate of the ammonia storage level 33 on the catalyst.

-10 -The dosing calculation unit 3 may use feed forward control to determine what level of dosing should be applied to the SOR device 20. This may be determined using a number of different techniques, and by considering a number of different measurement signals. For example, at least one of the ourrent measured or estimated exhaust gas state 31, the oatalyst temperature and the estimated ammonia storage state 33 may be oonsidered in order to determine the dosing level.

The dosing calculation unit 3 may also perform feed forward control by determining the ideal or desired ammonia storage level of the catalyst for a given situation as determined by at least one of the measured or estimated exhaust gas state 31, the catalyst temperature, and the current or anticipated engine work load. This ideal or desired ammonia storage level may be the storage level at which NOx and ammonia slip output from the 5CR system 20 is expected to be at a minimum. By determining the ideal or desired ammonia storage state, the dosing calculation unit 36 may compare it with the estimated storage state 22 and determine the dosing level required in order to bring the estimated storage state 33 closer to the ideal or desired ammonia storage state.

Alternatively, an ideal or desired ammonia storage state may be input to the dosing calculation unit 3 from a different unit, either internal or external to the controller 30.

However, when using only feed forward control, the accuracy of the dosing level may not be very reliable, and factors such as sensor drift and changes in catalyst storage levels for a given temperature may cause the inaccuracies to increase over time.

-11 -The accuracy of the dosing level may be improved by determining whether mis-dosing is occurring, and if so what type of mis-dosing it is (i.e. over-dosing or under-dosing) Such a determination may then be used to perform feedback control of the dosing level.

In order to determine if mis-dosing is occurring, the estimator unit 32 may generate an expected NOx output level 37, which may be determined using the catalyst model with, for example, at least one of the exhaust gas state 31, the catalyst temperature and the current dosing level 39.

The error calculation unit 34 in the controller 30 may then find the difference between an expected NOx output level 37 and a measured NOx output level 38 from the downstream NOx sensor 48. The result of this difference calculation shall be referred herein as the MIx readout difference'.

The NOx readout difference may then be cross correlated with a MIx sensor sensitivity figure in order to determine whether under-dosing or over-dosing is taking place. The NOx sensor sensitivity figure may indicate the sensitivity of the downstream NOx sensor 48 to reduotant dosing errors.

To that end, the NOx sensor sensitivity figure may be an indication of how much the downstream MIx sensor reading is expected to change with the dosing error. The figure may be positive or negative, indicating a positive or negative sensitivity of the NOx out measurement to dosing errors, and may have a magnitude indicating the degree of sensitivity.

-12 -The NOx sensor sensitivity figure may be obtained using a number of different techniques, either before the control method is run, or during the running of the control method (run-time) For example, the NOx sensor sensitivity figure might be estimated and set before running the control method, and that estimate be used throughout the running of the control method. This technigue might simplify computation during operation, but it may not be able to react to any changes in the sensitivity of the downstream NOx sensor 48.

Alternatively, during run-time, a NOx sensor sensitivity figure might be determined using a model within the estimator unit which considers the NOx output measurement 38, as well as at least one of the catalyst temperature, the dosing level 39 and the exhaust state 31 in order to generate the NOx sensor sensitivity figure. The model might also consider historical NOx output measurements.

A further run-time technique might be to differentiate the expected NOx output 37 by a measure of the level of dosing error applied to the 8CR device 20. The measure of the dosing error may be the dosing error e, the determination of which is explained later. When the dosing error e is used, the NCx sensor sensitivity figure may initially be calculated by assuming a notional dosing error of, for example, 0, which indicates that there is no dosing error.

The NOx sensor sensitivity figure may then be gradually improved during run time when progressively more accurate figures for the dosing error e are determined. This differentiation may be performed in different ways, such as -13 -manual, computer aided or automatic differentiation, and may generate an accurate figure for the NOx senscr sensitivity, which may react to changes in the sensitivity of the NOx sensor 48. This technique may produce quite an accurate NOx sensor sensitivity figure, but may be quite difficult to implement -The NOx sensor sensitivity figure may also be obtained during run-time from the difference between the expected NOx out 37 obtained using the oatalyst model and a NOx out estimate obtained using a modified catalyst model. In this instance, the estimator unit 32 may further comprise a model bank which comprises the catalyst model and the modified catalyst model, wherein each of the models may be run in parallel.

The modified catalyst model may be used to estimate the NOx output expected for an over-dosed or under-dosed SCR device 20. For example, the modified catalyst model may be used to obtain an estimate of NOx out for a 1% over-dosing condition, compared with the dosing level 39. It should be noted, however, that rather than 1% overdosing, the modified catalyst model may alternatively use a different over-dosing or under-dosing percentage, for example 5% under-dosing.

The NOx sensor sensitivity figure may then be obtained by dividing the NOx readout difference by the difference between the expected NOx out 37 obtained using the catalyst model and the NOx out estimate obtained using the modified catalyst model. This technique may be easier to implement than the differentiation technique explained above.

-14 -Alternatively, the NOx sensor sensitivity figure may be obtained from the difference between a NOx out estimate obtained using an over-dosing catalyst model and a NOx out estimate obtained using an under-dosing catalyst model. In this instanoe, the estimator unit 32 may include a model bank that comprises the oatalyst model, the over-dosing model and the under-dosing model, wherein eaoh of the models may be run in parallel.

The under-dosing model may, for example, be used to estimate the NOx cut expected for a 1% under-dosing condition oompared with the dosing level 39, and the over-dosing model may, for example, be used to estimate the NOx out expected for a 1% over-dosing conditicn compared with the dosing level 39.

It should be noted, however, that the over-dosing and under- dosing models do not have to use equal and opposite mis-dosing percentages with respect to the dosing level 39 and may, for example, use 3% under-dosing and 5% over-dosing.

The NOx sensor sensitivity figure may then be obtained by dividing the NOx readout difference by the difference between the expected MIx out obtained using the over-dosing model and the NOx out estimate obtained using the under-dosing model.

This technique may be computationally more intensive than the two model technique explained above. However, it may provide a more accurate figure as a consequence of considering both over-dosing and under-dosing, whilst the two model technique may be lopsided' because it considers only over-dosing or under-dosing.

-15 -It should be noted, however, that good results may still be obtained when the two mis-dosing models (i.e., the over-dosing and the under-dosing models), both estimate NOx output levels for different levels of over-dosing, or both estimate NOx output levels for different levels of under-dosing. For example, one model might use 3% over-dosing and the other model use 5% over-dosing, or alternatively one model might use 1% under-dosing and the other model use 2% under-dosing.

Figure 3 shows an example of an expeoted NOx out generated by the catalyst model, and NOx out estimates generated by the over-dosing model and the under-dosing model, at a single point in time. In this example, the actual NOx out reading from NOx sensor 48 is somewhere between the expected NOx out generated using the catalyst model and the estimated NOx out generated using the over-dosing model. The difference between the NOx out estimate obtained using the over-dosing model and the NOx out estimate obtained using the under-dosing model is labelled diff2. The difference between the expected NOx out generated by the catalyst model and the NOx out measurement from the downstream NOx sensor 48 is labelled diffl.

The type of dosage error (i.e., over-dosing or under-dosing) indicated by the NOx out measurement shown in Figure 3 may be found by cross-correlating the NOx readout difference (the difference between the expected NOx output level 37 and a measured NOx output level 38 from the downstream NOx sensor 48) with the NOx senscr sensitivity figure. The cross-correlation operation finds the product between the -16 -NOx readout difference and the NOx sensor sensitivity figure at the same point in time. The outcome from the cross-correlation shall be referred to herein as the mis-dosing indication figure, which indicates if the 6CR device 20 is being under-dosed, correctly dosed or over-dosed.

The mis-dosing indication figure is a dimensioniess number indicating the similarity between the NOx readout difference and the NOx sensor sensitivity figure. A positive mis-dosing indication figure may indicate that the 6CR device 20 is being over-dosed, a negative mis-dosing indication figure may indicate that the 6CR device 20 is being under-dosed and a mis-dosing indication figure of zero may indicate that the 6CR device 20 is being correctly dosed. :is

The mis-dosing indication figure is not subject to the cross-sensitivity of the downstream NOx sensor 48 to ammonia. That is to say, the mis-dosing indication figure reliably indicates what type of dosing error may be occurring with independence from the ammonia cross-sensitivity.

This is a consequence of the inherent non-linear nature of cross-correlation. Linear control approaches, for example those that consider only the difference between expected NOx out and measured NOx out, may not differentiate between an elevated NOx out reading caused by an increase in NOx out and an elevated reading caused by ammonia slip. In the present disclosure, however, the NOx sensor sensitivity figure enables the cross-correlation to interpret the difference between the actual NOx out reading 38 and the -17 -expected NOx out and accurately indicate if under-dosing or over-dosing is oausing the difference.

As a consequence, the mis-dosing indication figure determined by the error oalculation unit 34 may be used in olosed loop control in order to indicate what type of dosing error may be taking place. Closed loop feedback of this type might enable dosing levels applied by the injector 40 to be changed more accurately in order to correct dosing errors.

For example, the mis-dosing indication figure may be fed back to the dosing calculation unit 36 in order to adjust the dosing level 39 to correct for any mis-dosing identified by the mis-dosing indication figure. The mis-dosing indication figure may also, or alternatively, be fed back to the estimator unit 32 in order to adjust the catalyst model.

By adjusting the catalyst model on the basis of any mis-dosing identified by the mis-dosing indication figure, the catalyst model may more accurately reflect the actual 5CR device 20 and therefore better estimate the catalyst storage state 33, which might enable the dosing calculation unit 36 to generate a more suitable dosing level 39.

In contrast to other dosing control techniques currently in use, the mis-dosing indioaticn figure may be improved by transient events in the NOx cut measurement 38 as a consequence of the cross-correlation operation.

Transient events may be caused, for example, by a sudden, short-lived increase in NOx cutput from the engine 10, caused by a short, rapid increase in engine load, which may -18 -cause a transient spike in Mix output from the 3CR device 20. This may be an inevitable consequence of the dosing control unit 30 being unable to instantaneously increase the amount of ammonia stored on the catalyst before the spike in NOx out from the engine 10 has reached the catalyst.

The transient spike of NOx output from the SCR device 20 may be predicted to an extent by the catalyst model in the estimator unit 32, and therefore the expected NOx out 37 may include the transient spike and the NOx readout difference may remain steady. However, there may be a disagreement between the state of the 3CR device 20 and the state of the catalyst model, in which case the NOx readout difference may also spike, which may reveal valuable information about the differences between catalyst model and catalyst itself. The mis-dosing indication figure may pick up this information by virtue of the cross-correlation operation, which may result in a more accurate mis-dosing indication figure and, therefore, more accuracy in the closed-loop feedback described above.

The mis-dosing indication figure may be calculated multiple times by taking multiple upstream exhaust gas state 31 readings and readings from the downstream NOx sensor 48 and performing the above steps for each set of readings. For example, readings may be taken every second, or more preferably every lOOms. The sensitivity estimate used to determine the mis-dosing indication figure may be determined once, either before running the control method or at the start of running the control method, or may be determined a number of times during run-time, for example at the same -19 -time as each NOx cut and input exhaust gas state reading, cr more or less frequently than those measurements.

The plurality of mis-dosing indication figures may be accumulated over time and the average of those accumulated figures found. As a consequence of accumulation and averaging, the influence of short term errors and noise may average nut.

The accumulation of mis-dosing indication figures may take place for only a short period of time, for example two or three sample periods, or may last indefinitely, with the averaging function continually improving and refining the cross-correlation figure. Preferably, the period over which the cross-correlation takes place may be at least one hour, after which time short term errors and noise may more reliably average out. However, significantly shorter time periods may still generate useful results.

By continually monitoring the mis-dosing indication figure over time, any sensor drift, changes in the dynamics of the catalyst, changes in the accuracy of the dosing pump 40 and changes in any other factor affecting the accuracy of the catalyst model, may be picked up by the average mis-dosing Indication figure. Consequently, the above explained closed-loop feedback of the mis-dosing lndioatlon figure to the estimator unit 32 and/or the dosing control unit 36 may be performed throughout the life of the 5CR device 20, ensuring that the emission of NOx or ammonia slip in the exhaust gas output stream may be minimised for the lifetime of the 5CR device 20.

-20 -If the mis-dosing indication figure is monitored over time, it may be filtered in order further to improve the reliability of the mis-dosing indication figure.

The accumulated mis-dosing indication figure may be, for example, low pass filtered in order to remove high frequency events, such as noise in sensor readings or singular events.

By performing low pass filtering, long term, persistent causes of dosing error, such as sensor drift, may be represented in the mis-dosing indication figure, but short lived errors caused by noise and/or singular events may be removed, thus allowing the mis-dosing indication figure to converge gradually on a more reliable figure. This may enable the control unit 30 to achieve improved long term dosing in the future.

The mis-dosing indication figure may be accumulated and averaged over time using any one of a large number of techniques that would be well known to the person skilled in the art. For example, the mis-dosing indication figures may be stored in memory over a period of time and the mean mis-dosing indication figure determined using those stored figures.

Alternatively, an exponentially weighted mis-dosing indication figure may be determined by, for example, passing the mis-dosing indication figure through a low pass filter.

This particular technique might be preferable because less memory is required and, therefore, computation may be more straightforward. Furthermore, by using a low pass filter to generate an exponentially weighted average mis-dosing -21 -indication figure, both low-pass filtering and averaging may be performed by a single process.

An even further advantage of determining the exponentially weighted average mis-dosing indication figure might be that more reoent mis-dosing indication figures are given a greater importance than older figures. Thus, if the control method has been running for a long period of time, very old information, which might have been obtained when the SCR device 20 was operating under very different conditions, might effectively be discarded, and newer information, which might be more relevant to how the OCR device 20 is currently operating, might have more influence over the control of the OCR system 20. This may result in the determination of an average mis-dosing indication figure that is of more use for feed-back control.

There are a number of other techniques bywhich older information might be discarded, which will be well known to the skilled person. For example, if the average mis-dosing indication figure is determined from the mean of a number of stored mis-dosing indication figures, the mean might only be determined from a particular number of the most recent figures. For example, the 100 most recent figures stored in memory, and all older figures might be discarded.

It may be possible to improve the accuracy of the mis-dosing indication figure even further by using a time delay in the estimator unit 32. It may take exhaust gases some time to pass through the OCR device 20. Consequently, a volume of gas which is sensed by the sensors 42, 44 and 46 upstream of the OCR device 20 may take some time, for example, 2 -22 -seconds, to travel through the 5CR system 20 and be sensed by the downstream NOx sensor 48. This time delay may be referred to as the transport delay' By adding a time delay approximately equal to the transport delay of the 5CR devioe 20 into the estimator unit 32, it may be possible for the estimator unit 32 to generate an expected NOx out level 37 which takes the transport delay into account. For example, a volume of gas which is sensed upstream of the 5CR device 20 may experience a transport delay of 2 seconds. The estimator unit 32 may consider the readings from the upstream sensors and generate a NOx out estimate which it may expect to occur 2 seconds later. That NOx out estimate may then be sent to the error calculation unit 34 after the 2 second time delay so that the NOx out measurement from the downstream NOx sensor 48 may be directly compared with the delayed expected NOx out level 37. Alternatively, the error calculation unit 34 may include the time delay such that it stores the NOx out estimate from the estimator unit 32 for a period of time equal to the time delay, and then compares it to the NOx out measurement 38.

Thus, introducing use of a time delay may improve the accuracy of the mis-dosing indication figure determined by the error calculation unit 34. A time delay may be used regardless of whether or not the mis-dosing indication figure is accumulated over time.

The scale of the mis-dosing indication figure depends on many parameters, and so the magnitude of the figure by -23 -itself may not provide a useful indication of the degree of mis-dosing taking place.

In order to identify the magnitude of a dosing error which is taking place, the error calculation unit 34 might determine a dosing error e figure frcm the mis-dosing indicaticn figure.

The dosing error e is a dimensional number which indicates not only if the SCR device 20 is being over or under dosed, but also the extent of the dosage error. The dosing error e might indicate the dosage error in a number of different units; for example, it might indicate the mis-dosing as volume of reductant (e.g., 10mg over-dosing), as a rate of dosing of reductant (e.g., 5mg/sec under-dosing), as a concentration of active ingredient within the reductant (e.g., 15 ppm overdosing) or a percentage of mis-dosing (e.g., 2% under-dosing). For example, a dosing error e of 0 might indicate correct dosing; a dosing error e of 0.005 might indicate 0.5% overdosing; a dosing error e of 0.01 might indicate 1% overdosing; a dosing error e of -0.005 might indicate 0.5% under-dosing; and a dosing error e of - 0.01 might indicate 1% under-dosing.

The dosing error e may be calculated using a number of different techniques depending upon the desired form of dosing error e. For example, a percentage dosing error e may be determined by dividing the mis-dosing indication figure (or accumulated average mis-dosing indication figure) by the auto-correlation of the NOx sensor sensitivity figure.

-24 -The auto-correlation of the NOx sensor sensitivity figure may be the square of the NOx sensor sensitivity figure, and it may represent the expected mis-dosing indication figure at a notional mis-dosing condition.

If the dosing error e is determined, it may be fed into the dosing calculation unit 36 and/or the estimator unit 32 instead of the mis-dosing indication figure to perform closed-loop control cf the dosing level 39. An advantage of feeding back the dosing error e, rather than the mis-dosing indication figure, is that the dosing error e may provide more useful information regarding the magnitude of the dosing error, rather than just what type of dosing error is taking place. 3-5

For example, the dosing control unit 36 may modify the dosing level that is determined by the feed forward controller on the basis of the dosing error e, such that the dosing level 39 applied to the injector 40 is corrected to overcome the dosing error e. In the case of a percentage dosing error e, the figure determined using the technique above may be added to 1 (such that a dosing error e=1 indicates perfect dosing, e=l.01 indicates 1% overdosing and e=0.995 indicate 0.5% under-dosing) and then the dosage level determined by feed forward control can be divided by the dosing error e. As a result, feed-back control may be achieved, which more accurately adjusts the level of dosing applied to the SOR device 20 so that ammonia slip and NOx output from the SC?. device 20 may be reduced.

Additionally, or alternatively, the dosing error e may be used by the estimator unit 32 to adjust the catalyst model.

-25 -Because the dosing error e is an indication of the level of mis-dosing caused by, for example, sensor inaccuracies and drift, inaccuracies in the injector 40 and changes in catalyst storage dynamics, adjusting the catalyst model on the basis of the dosing error e may result in the catalyst model itself incorporating, and compensating for, the causes of the mis-dosing. This again turns the dosing level feed forward control scheme into a feed back control scheme.

Consequently, the catalyst mcdel may generate a more accurate estimate of the ammonia storage level 33, which may result in the generation of a dosing control signal 39, which more accurately reflects the level of dosing required by the SCR device 20.

Thus, by improving the catalyst model in this way, the level of urea dosing actually applied to the SCR device 20 may be improved such that the level of untreated NOx or ammonia slip in the output of the SCR device 20 is reduced.

Various modifications of the method described above which fall within the scope of the present disclosure are contemplated.

For example, as will be readily apparent to the skilled person, a number of the steps which are performed by the controller 30 may be performed in a different order to that described above. This may lead to mathematically identical results, or it may lead to variations in results, which may still create a functional system.

-26 -One such example cf this is cbtaining the mis-dosing indication figure by performing cross-correlation first and then finding a difference value. In this example, the error calculation unit 32 may cross-correlate the NOx out measurement 38 with the NOx sensor sensitivity figure and cross-correlate the expected NOx out 37 with the NOx sensor sensitivity figure and then obtain the mis-dosing indication figure by finding the difference between those two cross-correlation figures.

Furthermore, the filtering operation described above in respect of the accumulated mis-dosing indication figure may instead be formed at any stage of the method. For example, it may be performed on the NOx out measurement 38 and/or the expected NOx out 37, or it may be performed on the NOx readout difference, or on one or both of the oross-oorrelation figures described in the above paragraph.

Rather than considering a the expected and measured the NOx output from the 3CR devioe 20, different measures of the exhaust gas state may alternatively be oonsidered, for example amount of ammonia.

Figures 1 and 2 show a controller 30 in aooordanoe with an

embodiment of the present disclosure.

The oontroller 30 may be configured to carry out the method

steps described in the present disclosure.

The oontroller 30 may have a number of inputs and outputs which may be used by the estimator unit 32, the error calculation unit 34 and the dosing calculation unit 36 in -27 -order to perform the above described method steps. For example, the inputs might include, but are not exclusive tc: the measured NOx cut level 38 and the measured state 31 cf the exhaust gas input to the 8CR system 20, which may include the NOx in level, the exhaust gas temperature and the mass flow rate of the exhaust gas. The ccntrcller 30 may also have a number of outputs, including, but not exclusive to, the dosing level control signal 39.

The controller 30 may be implemented in an engine control unit, for example the Caterpillar A4:E4 or A5:E2, or as a standalone control unit.

Figure 1 also shows an 8CR system comprising an 8CR device 20 and the controller 30, which is arranged to determine if the OCR device 20 is being mis-dosed. Furthermore, Figure 1 also shows an engine unit comprising an internal combustion engine 10 and the OCR system.

Figure 4 shows a vehicle within which the engine unit shown in Figure 1 could be used.

Industrial Applicability

The present disclosure finds application in determining if an OCR device is being mis-dcsed with reductant, which may lead to improvements in controlling the dosing of an OCR device.

Claims

-28 -Claims 1. A method for determining if a seleotive catalytic reduction (8CR) device that is dosed with a reductant is being mis-dosed, the method comprising the steps of: determining a NOx readout differenoe figure from the difference between an expected exhaust gas output state from the SCR device and a measured exhaust gas output state from the OCR device; and determining a mis-dosing indication figure which indicates if the SCR device is being under-dosed, correctly dosed or over-dosed by cross-correlating the NOx readout difference with a NOx sensor sensitivity figure, wherein the NOx sensor sensitivity figure Is an indication of how sensitive the measured exhaust gas output state is to dosing errors.
2. A method for determining If a selective catalytic reduction (8CR) device that is dosed with a reductant is being mis-dosed, the method comprising the steps of: determining a first cross-correlation figure by cross-correlating a measured exhaust gas output state with a NOx sensor sensitivity figure, wherein the NOx sensor sensitivity figure is an indication of how sensitive the measured exhaust gas output state is to dosing errors; determining a second crcss-correlation figure by oross-correlating an expected exhaust gas output state from the 8CR device with the NOx senscr sensitivity figure; and determining a mis-dosing indication figure which indicates if the OCR device is being under-dosed, correctly dosed or over-dosed from the difference between the first -29 -cross-correlation figure and the second cross-correlation figure.
3. The method of either of claims 1 or 2, wherein a quantitative dosing error figure is determined by dividing mis-dosing indication figure by an auto-correlation of the expected exhaust gas output state.
4. The method of claim 3, wherein the NOx sensor sensitivity figure is determined by differentiating the expected exhaust gas output state by the quantitative dosing error figure.
5. The method of any one of claims 1 to 3, wherein the NOx sensor sensitivity figure is determined by dividing the difference between the expected exhaust gas output state and the measured exhaust gas output state by the difference between the expected exhaust gas output state and an estimated exhaust gas output state determined by a mis-dosing model which estimates the exhaust gas output state based upon a notional under-dosing or over-dosing condition.
6. The method of any one of claims 1 to 3, wherein the NOx sensor sensitivity figure is determined by dividing the difference between the expected exhaust gas output state and the measured exhaust gas output state by the difference between an exhaust gas output state estimated by a notional over-dosing model and an exhaust gas output state estimated by a notional under-dosing mcdel.
7. The method of any preceding claim, wherein the mis-dosing indication figure is determined a plurality of times -30 -over a period of time and the average of the plurality of mis-dosing indication figures is found.
8. The method of claim 7, wherein the plurality of mis-dosing indication figures is low-pass filtered in order to remove high-frequency noise.
9. The method of any preceding claim, wherein the expected exhaust gas output state from the OCR device is determined from at least one of a measured state of the exhaust gas input to the OCR device, the level of reductant dosing applied to the OCR device and an estimate, or measurement, of the catalyst temperature.
10. The method of any preceding claim, wherein the measured state of the exhaust gas input to the OCR device comprises at least one of the amount of NOx in the exhaust gas, the temperature of the exhaust gas and the mass flow rate of the exhaust gas.
11. The method of any preceding claim, wherein the measured state of the exhaust gas output from the SCR device comprises a measurement of the amount of NOx in the exhaust gas.
12. A controller to determine if a selective catalytic reduction (OCR) device that is dosed with a reductant is being mis-dosed, the controller being configured to: determine a NOx readout difference figure from the difference between an estimated exhaust gas output state from the OCR device and a measured exhaust gas output state from the SCR device; and -31 -determine a mis-dosing indication figure which indicates if the SCR device is being under-dosed, correctly dosed or over-dosed by cross-correlating the NOx readout difference with a NOx sensor sensitivity figure, wherein the NOx sensor sensitivity figure is an indication of how sensitive the measured exhaust gas output state is to dosing errors.
13. A controller to determine if a selective catalytic reduction (SOR) device that is dosed with a reductant is being mis-dosed, the controller being configured to: determine a first cross-correlation figure by cross-oorrelating a measured exhaust gas output state with a NOx sensor sensitivity figure, wherein the NOx sensor sensitivity figure is an indication of how sensitive the measured exhaust gas output state is to dosing errors; determine a second cross-correlation figure by oross-correlating an estimated exhaust gas output state from the 5CR device with the NOx senscr sensitivity figure; and determine a mis-dosing indication figure which indicates if the OCR device is being under-dosed, correctly dosed or over-dosed from the difference between the first cross-correlation figure and the second cross-correlation figure.
14. An 5CR system comprising: an 5CR device that is dosed with a reductant, and the controller defined in either of claims 12 or 13, the controller being arranged to determine if the 5CR device is being mis-dosed.

-32 -
15. An internal combustion engine comprising the SOR system defined in claim 14.
16. A vehicle comprising the internal combustion engine defined in claim 15.