The invention relates to a method of identifying engine gas composition.
The demands for lower fuel consumption and tough emissions reduction targets has led to the requirement of after treatment systems. However, the systems required for diesel engines are costly and therefore in order to delay their introduction, there is much focus on finding new ways of reducing engine-out emissions. It is well established that combustion duration within a cylinder of an engine correlates significantly with the charge content of said cylinder. EGR (exhaust gas recirculation) is conventionally employed to control the temperature and rate of combustion in order to achieve non-conventional combustion modes.
In general, the higher the amount of inert gas (EGR), the slower and more controlled the rate of combustion and therefore the less NOx out and the cooler the engine. However, the distribution of EGR, air and O2 between individual engine cylinders is becoming more significant. Taking the example of a diesel engine, all cylinders would normally receive the same amount of fuel (adversely reducing torque) in order to control the smoke output, but the overall performance of an engine is often limited to the ‘culprit’ cylinder that contains either the least or most of one of the species for the required transient or steady-state conditions.
Because of recent developments in the individual cylinder control of fuelling and valve actuation for example by an ECU (engine control unit) the estimation of the composition of the gas within each cylinder is becoming more of a practical requirement.
One known approach is to obtain the rate of heat release from cylinder pressure signals and use this to estimate the AFR (Air/Fuel Ratio) and EGR through empirical look-up tables. However, this is prone to error at light loads or with more complex multi-injection fuelling systems.
Other known approaches are based on individual cylinder pressure sensor data but suffer from problems in obtaining sufficiently accurate data for passing to the ECU for engine control purposes. In US648694, cylinder pressure sensor drift is corrected according to detected manifold pressure. This is a well known practice on test beds, however is not so ideal in real-world engines, such as those in production vehicles, where cylinder to cylinder interaction and signal noise will exist. This is due to lower quality sensors and the need for transient control. WO02/095191 estimates polytropic index based on three pressure sensor samples which suffers from the problem of inaccuracy and noise. For cylinder charge estimation, JP2001-15293 describes using cylinder pressure to estimate the total gas composition within a cylinder, however it does not consider the individual species. The air or O2 content is important for controlling smoke emissions on a diesel engine. U.S. Pat. No. 5,611,311 discloses TDC (Top Dead Centre) estimation and correction where the cylinder pressure is observed at maximum in over-ran (zero-fuelling) without considering thermal loss in the system which can lead to inaccuracies. This is particularly relevant for strategies based on cylinder pressure feedback control that rely on calculations involving both instantaneous pressure and volume.
The invention is set out in the claims.
Embodiments of the invention will now be described, with reference to the drawings, of which:
The invention makes use of the observation that the polytropic index (Npoly) of an enclosed gas is closely related to its heat loss and constituent species concentration. For a fully warmed-up engine, this heat loss correlates closely with intake manifold temperature. The steady-state test-bed results of
In a calibration phase, therefore, test-bed results are obtained for each species concentration and plotted against polytropic index and intake manifold temperature.
Z
X
=f
X(NPoly,TInt) (1)
where:
ZX=Concentration of species X (air, EGR, O2, etc) (0-1) as a proportion of total mass M
Npoly=Polytropic index at compression (−)
TInt=Intake temperature (K)
It will be seen that the concentration ZO2, ZEGR and so forth can all be obtained in the calibration phase and stored in respective look-up tables. These concentrations can be based on any appropriate parameter such as but not limited to volume or mass. When an engine is running under real-world conditions, and it is desired to obtain Zx, calculations take place in two stages. By applying energy balance to the fixed mass of air, fuel and inert gas in the cylinder during the compression stroke before ignition, the derivation of the pressure signal offset and polytropic index is possible. In a first stage, Npoly is estimated and Tint is sampled, preferably local to a cylinder to provide a rough estimation of species concentration Zx from the 2-D look-up table derived in the calibration phase as represented by (1). In a second stage, real-time pressure measurements (sensed pressure and a calculated offset) enable the further correction of Zx which in turn is used to derive the mass of the particular species present in each cylinder. This information is then fed back to the ECU for subsequent use in controlling variables such as but not limited to the ignition, EGR feedback or fuelling of each individual cylinder.
In the diesel engine shown in
Intake manifold sensors 420 (pressure) and 422 (temperature) and in-cylinder pressure sensor 424 are arranged to sample data sufficient for the monitoring of charge content per cylinder and hence provide the means with which the ECU obtains Tint, estimates Npoly, obtains Zx, further refines Zx, and therefore controls EGR valve 428 in order to alter the bulk charge proportion of EGR within the intake manifold 412, inlet valve 418 and exhaust valve 428 in order to alter the individual cylinder charge content, and fuel injector 430 in order to achieve an optimised trade-off between performance, emissions and fuel economy.
Data acquired as set out above is manipulated in real-time to constantly monitor the charge content per cylinder. Stage 1 of this process comprises estimating the polytropic index for a single cylinder:
Applying the polytropic gas law PVN=Const to the cylinder charge gives:
(PSens+POffset)VCylN
where:
PSens=Cylinder pressure measurement (Pa)
POffset=Sensor offset due to drift (Pa)
VCly=Cylinder volume (m3)
NPoly=Polytropic index (−)
KPoly=Polytropic constant
Once POffset is known, the polytropic index may be estimated logarithmically by Linear-Regression taking all samples, preferably more than three, over the compression stroke. However, the direct measurement of POffset using an intake manifold pressure sensor is not trivial due to pressure fluctuations near IVC (intake valve closure) and sensor noise, and would often lead to errors in the polytropic index. An alternative approach is therefore described below.
The invention herein describes a technique enabling the explicit derivation of NPoly and POffset.
Turning firstly to NPoly, this can be obtained from a linear expression related to pressure samples taken shortly after IVC up to around 20° before TDC for each cylinder. An accurate TDC point of each cylinder taking into account system delays such as but not limited to thermodynamic loss, processor delays, phase lag of sensors and analogue/digital filters is preferably calibrated on a test-bed at manufacture and is stored as a thermodynamic loss angle and mapped against engine condition. This allows for non-adiabatic thermal loss to the environment and other system delays wherein the peak pressure is non-aligned with the TDC point of the piston within the cylinder which would otherwise create inaccuracies between the timing of the control system and engine cycle/piston position.
Applying energy balance to the trapped mass (in a cylinder) in the continuous time domain gives:
{dot over (U)}+{dot over (W)}={dot over (Q)} (3)
where {dot over (U)} is the rate of change of internal energy, {dot over (W)} is the rate of work done on the environment (heat transfer to the surrounding engine parts) and {dot over (Q)} is the rate of net heat gained.
The rate of change of internal energy for a gas of fixed mass m at temperature T is given by:
where cv is the specific heat capacity of the gas at constant volume. Applying the perfect gas law PV=mRT gives:
where P and V are the pressure and volume of the enclosed gas and R is the gas constant. Since cv/R=1(γ−1), where γ is the ratio of specific heats, and assuming this remains constant, (5) can be rewritten as:
The rate of work done by the gas on the environment is given by:
Substituting (6) and (7) back into (3) gives:
Integrating with respect to time:
where suffix ‘0’ denotes initial conditions.
By assuming that the rate of heat exchange is governed by a polytropic gas relationship of polytropic index NPoly, (9) can be approximated to:
where the left most term includes the heat transfer represented by the closed integral on the right-hand side of (9).
Allowing for inherent errors in sensing the pressure, the sensed pressure PSens equals actual pressure P modified by an offset POffset:
P
Sens
=P−P
Offset (11)
hence P=PSens+POffset
and assuming the offset remains constant during the compression stroke of the engine, (10) is modified to:
Rearranging this gives:
where K1=1/(NPoly−1) and K2=−NPolyPOffset/(NPoly−1).
over continuous time.
Converting (14) to the discrete crank-synchronous domain and applying trapezoidal integration, for each sample i we can approximate:
X
i
K
1
+Y
i
K
2
=W
i (15)
where:
X
i
=P
Sens
V
i
−P
Sens
V
0
Y
i
=V
0
−V
i
W
i
=W
i−1−(PSens
Vi is known at any point as it is directly derivable from the crank (or piston) position and the known volume V0 of the cylinder, and it can be shown that K1 and K2 in (15) can be solved by linear regression (that is to say finding a best solution for the multiple values of Xi, Yi, and Wi) to give numerical values using:
where Xi, Yi and Wi are calculated at each sample i=1, 2, . . . , N. K1 and K2 may be re-arranged from (14) to give:
As a result, from measured Tint and derived NPoly, the corresponding value of Zx for that cylinder can be obtained from the look-up table of
It should be noted that the linear regression proposed its only one method of obtaining a “best fit”. There are many alternative approaches include nonlinear regression, the Maximum Likelihood method and Bayesian Statistics. Iterative approaches would involve constructing a penalty function, E; at each iteration j, such that for example:
where:
e
i
=X
i
K
1,j
+Y
i
K
2,j
−W
i
Here K1,j and K2,j are values calculated at each iteration so as to minimise E such that eventually
Sufficient convergence will take place after a finite number of iterations and can be achieved using well known minimisation algorithms such as steepest descent and the simplex method. The computational overhead of performing multiple iterations within each engine cycle in any case can be mitigated by spreading the number of iterations over multiple cycles such that after, say, 3 iterations in one cycle the calculated values of K1 and K2 can be carried over to the next. Convergence will therefore take place after a number of engine cycles. The maximum number of iterations per cycle is selected to ensure overall convergence takes place, especially during transients.
Stage 2 of the process comprises obtaining an estimate of Zx. Depending on the specification of the pressure sensors in use, one of two methods may be employed to execute stage 2. By way of explanation, the following example relates to ZO2 using the fact that additional information is available in the form of the oxygen mass in the in-take manifold (26). Method A estimates the distribution of cylinder O2 concentration assuming that the mass is the same in each cylinder and method B provides an improved estimate of the distribution of O2 concentration and, in addition, estimates the respective masses. The difference in intake temperature of the inducted mixture between cylinders is assumed to be small relative to absolute temperatures.
A first estimate of the cylinder O2 concentration is obtained as described above from (1):
Z
O2indi
*=f
O2(NPolyi,TInt) (19)
where:
ZO2indi*=First estimate of O2 concentration in cylinder i (0-1)
NPolyi=Polytropic index of cylinder i (−)
fO2=O2 concentration function (may be implemented as a 2-D lookup table as described above)
TInt=Intake manifold temperature (K)
The intake manifold temperature is assumed to be the same for all cylinders.
This first estimate obtained is an empirical value from the test-bed model calibrated look-up table of
ZO2Indi=αZO2Indi* (20)
where ZO2Indi is the corrected oxygen concentration for cylinder i.
The mass balance relationship is as follows:
where MO2Int is the oxygen intake manifold mass per cycle and MO2Indi is the inducted oxygen mass for cylinder i of a 4-cylinder engine. This can be re-expressed as functions of oxygen concentrations and total inducted masses as
Applying (20) and re-arranging gives:
This gives the following for cylinder i:
By assuming the difference in total trapped charge masses between cylinders is small, the M's cancel due to
leaving:
where:
ZO2Ind=Corrected concentration of inducted cylinder O2 (0-1), based on the average O2 concentration obtained from mean-value observer models.
ZO2int=Bulk O2 concentration in the intake manifold (0-1)
This results in correctional factor
effectively related to how close the sum of the concentrations ZO2Indj* is to the expected value ZO2Int. If the summed first estimates of ZO2Indj* are less than ZO2Int, the correctional factor increases the original estimate of ZO2Indi*, and if they are more, the factor decreases the original estimate.
The bulk O2 concentration ZO2Int can be approximated by the following known steady-state expression that applies to lean mixtures:
Z
O2Int
=Z
O2Atm(1−ZEGR/λ) (26)
where:
ZO2Atm=Ambient O2 concentration (0.23 as standard based on mass) (0-1)
ZEGR=EGR rate (0-1)
λ=Excess air ratio (−)=AFR/stoichiometricAFR
Known observer models such as mean-value models in some of today's ECUs can be applied to obtain ZEGR. The excess air ratio, λ, can be obtained from an EGO sensor.
In equations (15-18), polytropic index NPoly was found from sensed pressure reading PSens
According to the invention, a further complementary, more precise correction may be applied to the O2 concentration obtained from method A that additionally takes into consideration the differences in total charge masses between cylinders i.e. without the assumption on which (25) is based.
Using (24):
where:
MInt=Total intake mass per engine cycle (sum of all cylinders) (kg)
MIndj=Total inducted mass in cylinder j (kg)
The O2 mass is given by:
where the bulk estimate, MInt, is obtained from known observer models in today's ECUs.
The individual cylinder masses are obtained directly from the cylinder pressure sensors as follows:
The inducted mass of cylinder i can be expressed as:
where:
ηVoli=Volumetric efficiency of cylinder i (0-1)
PInt=Intake manifold pressure (Pa)
R=Gas constant (J/kg/K)
TInt=Intake temperature (K)
VCylDisp=Cylinder displacement volume (m3)
Note that PInt and TInt are assumed to be the same for all cylinders and the variation of R with gas properties is assumed to be negligible.
It can be shown for example in Taylor, C., The Internal Combustion Engine in Theory and Practice, Volume 1, MIT Press, 1985 that by assuming the valve overlap period is negligible the volumetric efficiency can be estimated directly from cylinder pressure thus:
where:
VCylDisp=Cylinder displacement volume (m3)
VCyl=Cylinder volume (m3)
VIVC=Cylinder volume at IVC (m3)
CCompRat=Compression ratio (−)
PCyli=Cylinder i pressure (Pa)
PIVOi=Cylinder i pressure at intake valve open (IVO) (Pa)
PIVCi=Cylinder i pressure at IVC (Pa)
γ=Ratio of specific heats (−)
ΔTi=Temperature increase from in-take manifold to cylinder
Substituting (29) for ηVoli in (28) results in the cancellation of PInt and VCylDisp thus:
The cylinder pressures are corrected by;
P
Cylj
=P
Sensj
+P
Offset, j=IVO, . . . , IVC
where Poffset is obtained from Stage 1.
Alternatively, this can be pegged to intake manifold pressure, PInt, ie:
P
Cyl
=P
Sens
+P
Offset
−P
IVCLR
+P
Int
where PIVCLR is the first estimate of IVC pressure taken from the linear regression fit in Stage 1.
Applying (30) to (27), the cylinder O2 mass is now given by:
If the intake temperature sensor is located midway between the intake ports of all cylinders, any differences in ΔTi, (i=1, . . . , 4) can be assumed small compared to Tint. This important assumption results in the following solution for inducted O2 mass, MO2Indi, for cylinder i:
where the bulk estimate, MInt, is obtained from known observer models found in some of today's ECUs. All other variables are either known or measurable as described herein.
Unlike Method A, this method requires gain calibration of the cylinder pressure sensor as the absolute pressure is required, derived from the sensed value and the offset as found in stage 1.
It should be appreciated that while the above methods employ multiplicative corrections to a first estimate, ZO2Indi*, additive corrections to equations (24) and (25) such as but not limited to
would be equally valid.
Furthermore, it should be understood that the concentration of other species present may be estimated using the same principle as the O2 estimation described in stage 2 above.
When performing calculations with instantaneous cylinder pressures and volumes, such as in (16), (29) and (31), it is preferred that the crank angles for pressure and volume match as closely as possible such that the pressure is known fairly accurately at each position of the crankshaft. As discussed above, the accuracy depends on knowing precisely where the TDC occurs in the pressure trace. In practice there is a small but noticeable offset between the TDC as “seen” by the ECU and its true location due to crank sensor offset.
Furthermore, this can be slightly different for each cylinder due to crank pin offset of each piston and even crankshaft flexibility. For the control system described herein, a further apparent offset can occur due to the measuring chain delay such as sensor response time, phase lag in the filtering of raw pressure signals, signal acquisition delay and a further effect to be accounted for is the thermodynamic loss angle. In an ideal case where there is no heat transfer between the enclosed gas mixture and the cylinder walls (ie. adiabatic compression), the maximum pressure would occur at TDC. In practice, because of heat transfer, this maximum will always occur before TDC by an amount called the thermodynamic loss angle. This angle varies with engine speed and wall temperature, the latter of which can result in a noticeable difference between cylinders. A further correction is therefore necessary to the TDC 115 position to accommodate for this effect. The total correction is therefore given by:
ΔθOffset=θPmax+ΔθTLA−ΔθMC
where:
If, at pressure sample i, the corresponding crank-angle is taken to be θi, then for all angles, i=1 to N, the following correction will need to be applied:
θi,k=θi,k−1−βΔθOffset,k
where ΔθOffset,k is the TDC offset calculated in the kth engine cycle and β is a tuning constant less than 1 to ensure these corrections occur gradually.
It will be seen that the invention as described provides a range of solutions to common engine problems. The measurement of parameters such as but not limited to volume and pressure of the gas species present within each individual cylinder provides data that, together with the methodologies described in stages 1 and 2 of the invention, allow increased control of engine parameters on a species by species and cylinder by cylinder basis, including where required, an accurate value of POffset derived by linear regression. Variables such as proportion of EGR within any one cylinder at any one time provide the advantage of reduced emissions particularly in the case of a diesel engine. Improved fuelling allows optimum AFR or O2/fuel ratio leading to increased fuel economy and in the case of diesel engines, reduced particulate content of the exhaust gas leading to the avoidance of expensive add-on cleaning systems in order to meet emissions regulations. Control can be in any appropriate manner for example EGR control by variable valve actuation (VVA).
A further advantage of the individual cylinder approach is the avoidance of one “culprit” cylinder affecting the control of variables such as fuelling, ignition, EGR, and air content of every other cylinder in the same way.
It should be appreciated that the two stage method herein described of identifying engine gas composition may equally be applied to other engine configurations and types such as but not limited to differing engine types, such as rotary, differing stroke cycles and differing number of cylinders employed, and differing fuel types, such as diesel or gasoline, wherein the ignition may additionally be controlled as a result of the data obtained.
It should be further appreciated that, as well as direct sensing of in-cylinder pressure, the pressure sensor can be mounted external to the cylinder, in the form of a spark-plug washer, gasket displacement sensor or integrated into a glow-plug.
Number | Date | Country | Kind |
---|---|---|---|
0601727.1 | Jan 2006 | GB | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/GB2007/000274 | 1/26/2007 | WO | 00 | 1/21/2009 |