FR2778210A1 - METHOD FOR THE CANCELLATION OF THE VARIATIONS IN THE WEALTH OF THE GASEOUS MIXTURE FROM THE CYLINDERS OF AN INTERNAL COMBUSTION ENGINE - Google Patents

Abstract

The invention relates to internal combustion engines of the injection type.The method consists in: (a) synchronizing the signal y (t) to take account of the transport time of the gas mixture in the manifold, (b) regulating the individual richness of the input mixture in each of the engine cylinders for stable operating points by introducing into the regulator (22) a disturbance pattern so as to create a periodic signal u (k) canceling the expected disturbance. (c) identifying injector gains on the signal u (k) during regulation for stable operating points, and (d) regulating the individual richness of the input mixture in each of the engine cylinders during the transitions between two operating points by multiplying injector controls by gains.

Description

METHOD FOR CANCELING VARIATIONS IN WEALTH OF THE

  GAS MIXTURE FROM CYLINDERS OF A MOTOR

INTERNAL COMBUSTION

  The invention relates to internal combustion engines and more particularly in such engines, a method for canceling the variations in richness of the gas mixtures coming from the cylinders of an internal combustion engine of the injection type. The cleanup of internal combustion engines requires increasingly fine regulation of the mixture of exhaust gases in order to meet increasingly stringent standards. For this purpose, it is no longer possible to be satisfied with regulating the gaseous composition of the exhaust gases on average over an engine cycle and it is therefore necessary to control the gaseous mixture with the

output of each cylinder.

  The composition of the exhaust gas mixture depends mainly on the ratio between the mass of fuel and the mass of air in each of the engine cylinders. This ratio is called the richness of the fuel / air mixture and this richness is measured using an oxygen sensor placed in the exhaust duct of the cylinders before the catalytic converter. More precisely, we speak of individual richnesses of the cylinders to define the fuel / air mixture in the combustion chambers and of the richness of the exhaust gases to define that measured by the oxygen sensor, the latter giving information

  combined wealth of cylinders.

  Usually, in a four-cylinder engine, all of the cylinders exhaust gases through the same exhaust duct. Depending on the order in which the cylinders are ignited, the oxygen sensor and the pollution control devices successively see the "exhaust" of the exhaust gases from each of the cylinders. Thus, a difference between the individual wealth causes a periodic variation in the richness of the exhaust gases, and therefore a fall in the effectiveness of the depollution strategies. In the case of a stoichiometric mixture of the fuel / air mixture, these periodic variations disturb the depollution organs. In the case of a lean fuel / air mixture, the increase in the richness of a single cylinder is sufficient to increase the emission of pollutants, in particular the emission of nitrogen oxides (NOx). In a four-cylinder engine, an engine cycle is broken down into four half-turns and ignition is carried out so that there is, at each half-turn, an admission of the fuel / air mixture, a compression of the mixture , a trigger and an escape of gases. Compensation for the variation in richness of the exhaust gases requires that the regulation be precise

within half a turn.

  One of the peculiarities of the physical system composed of the engine cylinders, the exhaust pipe and the oxygen sensor is the presence of a variable delay between the richness setpoint at the inlet of the cylinders and the richness of the exhaust gases. exhaust measured by the

oxygen sensor.

  This results in a phase shift between the engine cycles and the richness of the exhaust gases measured by the probe. However, this delay corresponds to the gas flow time to the probe and as this flow depends on the gas flow rate and therefore on the engine load, the delay varies from one operating point to another. Thus, if this variation of the delay is not taken into account in the regulator, the closed loop risks destabilizing. As the exhaust is a periodic phenomenon, the determination of this delay corresponds to synchronize the regulator with the four times of the engine so that the first problem & to solve is synchronization

wealth regulator.

  A second problem to be solved is the formalization of the effect of each cylinder. In the prior art, this problem is solved by estimating the individual richnesses of the cylinders from the richness signal supplied by the oxygen sensor and by associating

  each cylinder has its own regulation.

  Thus, in US Pat. No. 4,962,741, the effect of each cylinder is described by a model for the richness of exhaust over an engine cycle. Like the

  description does not contain an adaptation algorithm

  variable delay, the proposed regulator cannot take account of variations in operating points. The correction of the individual richness of the cylinders is based on the measurement of the richness of the exhaust gases. In US patent 5,524,598, the method implements an escape model associated with an estimation and a

  regulation of the individual wealth of the cylinders.

  The exhaust model chosen implies that the richness of the exhaust gases, measured by a so-called proportional probe (known by the acronym UEGO), is the weighted average of the individual richnesses of the cylinders. This model does not foresee a dynamic process in the mixture of gases and assumes that the transport times of the exhaust gases from the cylinders to the UEGO probe are

all equal.

  The object of the invention is therefore to implement a method for regulating the richness of the exhaust gases of an internal combustion engine which takes account of periodic phenomena and the effects

  of each cylinder on the exhaust gases.

  This goal is achieved by providing for synchronization of the signal supplied by the probe and regulation of the richness by rejection of disturbances according to a model.

of disturbance.

  Synchronization is obtained by introducing an adaptive delay between a proportional probe and a regulator of the individual richnesses of the cylinders according to a gradient optimization method and a variable delay in the richness signal supplied

by the probe.

  The regulation of the richness by rejection of disturbances is based on an internal model consisting in including in the regulator a dynamic model of the expected disturbance so as to create a periodic signal canceling the expected disturbance. The invention therefore relates to a method for canceling variations in richness of the gas mixture originating from the cylinders of an internal combustion engine, said engine comprising at least one injector per cylinder, a manifold for transporting the gas mixture disposed at the outlet of the cylinders and a so-called proportional probe disposed in said manifold and providing a signal y (t) representative of the richness of the gas mixture, said signal y (t) being applied to a regulator, characterized in that it comprises the following steps consisting in : (a) synchronize the signal y (t) to take account of the transport time of the gas mixture in the manifold, (b) regulate the individual richness of the input mixture in each of the engine cylinders for stable operating points in introducing a disturbance model into the regulator so as to create a periodic signal u (k) canceling the disturbance

expected.

  (c) identify injector gains on the signal u (k) during regulation for stable operating points, and (d) regulate the individual richness of the input mixture in each of the engine cylinders during the transients between two operating points by multiplying the commands of

injectors by gains.

  Step (a) mainly consists in: (ao) creating a signal y (T) by sampling the signal y (t) at a sampling frequency higher than that of the motor turn, (a1) estimating a delay (ê) between the sampled signal y (T) and a model signal s (i), (a2) introduce this estimated delay (ê) into the sampled signal y (T) to create a signal corrected z (T) according to determined criteria, and (a3) create a signal y (k) by resampling the signal z (rT) at the frequency of the U-turn

engine.

  Step (b) consists in: (b1) choosing a perturbation model P (q) according to determined criteria, (b2) combining P (q) with a model S (q) of gas mixture in such a way that a disturbance signal w (k), corresponding to the output signal of P (q), is added to the input signal S (q), (b3) choose a regulator R (q) for the combined model P (q ) .S (q) according to determined criteria, (b4) compare the sampled signal y (k) with a setpoint signal (yc) supplied by the injection computer to create a difference signal Sy (k) in the form (y (k) - yc), (b5) apply the difference signal Sy (k) to the regulator R (q) which provides a signal Eu (k), (b6) add the signal Su (k) to the reference signal yc in order to create the signal u (k), (b7) demultiplexing the command u (k) according to determined criteria to generate four independent commands ul (k), u2 (k) u3 (k) and u4 (k) of the injectors, and

  (b8) apply the four commands to the injectors.

  Step (c) consists in: (cl) demultiplexing the command u (k) according to determined criteria to generate four independent commands ul (k), u2 (k) u3 (k) and u4 (k) of the injectors, ( c2) create four signals gl (k), g2 (k), g3 (k) and g4 (k) by dividing each of the four signals ul (k), u2 (k), u3 (k) and u4 (k) by the mean on ul (k) to u4 (k), (c3) filter the signals gl (k) to g4 (k) with a low-pass filter, and (c4) replace the gains G1, G2, G3 and G4 respectively by the values gl (k), g2 (k), g3 (k) and g4 (k) when the regulation according to

step (b) is no longer possible.

  Step (d) consists of: (d1) calculating the global command Ug (k) of the injectors by regulating the average richness, (d2) creating four different injector commands by multiplying the global command Ug (k) respectively by gains G1, G2, G3 and G4, and

  (d3) apply the four commands to the injectors.

  The invention will be better understood using the

  following description of a particular example of

  realization, said description being made in relation

  with the attached drawings in which: - Figure 1 is a block diagram of a system implementing the method according to the invention, Figure 2 is a diagram illustrating one aspect of the method according to the invention, - Figures 3a, 3b, 3c and 3d are diagrams illustrating a disturbance model used in the method according to the invention, - Figure 4 is a functional diagram illustrating the formalization of the control problem for the synthesis of the disturbance rejection regulator according to l 'invention. - Figure 5 is a block diagram implementing the signal synchronization device

y (t) according to the invention.

  - Figure 6 is a block diagram showing

  operates the regulation device according to the invention.

  - Figure 7 is a block diagram illustrating one aspect of the method according to the invention, - Figure 8 is a block diagram illustrating another aspect of the method according to the invention, - Figure 9 is a diagram illustrating the evolution of the signal u (k) over time, - Figure 10 is a diagram illustrating the evolution of the signals gl (k) to g4 (k) over time, and - Figure 11 is a block diagram illustrating a

  aspect of the process according to the invention.

  In FIG. 1, an internal combustion engine 10 of the injection type produced by an injection device 26 comprises, for example, four cylinders including the exhaust gases, materialized by the four arrows 101, 102, 103 and 104, are collected in a collector 12 to be directed to a catalytic exhaust 24 preceded by a proportional probe 16. The element 14 materializes, functionally, a delay time e introduced by the length of the collector 12 and corresponding at the time of transport of the exhaust gases. The proportional type probe 16 provides a signal y (t), representing the richness of the exhaust gases, which is applied to a synchronization circuit 20 introducing a variable delay. The output signal y (k) of the synchronization circuit 20 is applied to a regulating device 22 which generates the control signals of the injectors of the injection device 26 from the signal y (k) and a set value yc supplied by an injection computer 28 known elsewhere. In the diagram of FIG. 5, which represents the synchronization circuit 20, the signal y (t) is sampled (circuit 30) at a frequency fe higher than that of the motor half-turn, for example ten times, to create a signal y (T). This latter signal is applied, on the one hand, to a delay estimator device 32 and, on the other hand, to a variable delay device 34 under the control of the estimator device

  delay which provides an estimated delay ê.

  The signal y (T), delayed by a value (Tc - ê), provides

  the signal corrected according to the formula: z (T) = y (T-Tc + ê).

  In this formula, Tc is the duration of a cycle of the

  motor and ê is the delay estimated by the device 32.

  The corrected signal is then resampled at the frequency of a half-turn of the motor (reference 36) to

create the signal y (k).

  The estimator device 32 calculates the estimated delay ê

  between the measured signal y (r) and a model signal s (T).

  The model signal s (r) is a signal which is determined during the calibration of the synchronization device for a determined distribution of the wealth and for a known delay. It is in the form of a sample at the frequency fe, the values of which are recorded in a memory 38. To calculate this estimated delay, the signal y (T) must correspond to the same distribution of the wealth as for s (T) and, for this purpose, the regulating device 22 is provided to introduce this

  distribution at instants determined by a signal Sy.

  The estimated delay is calculated, at each appearance of the signal Sy, using a synchronization algorithm seeking to minimize the area fn (FIG. 2) between the measured signal y (r) and the model signal s (r) by shifting the signal s (T) by a value (Tc-ê) which is then expressed by the formula:

s (t-Tc + ê).

  This minimization of the area is obtained by minimizing

  on a cycle the difference between y (T) and s (r-Tc + ê).

  The determination of the minimum value of the area n is obtained by a gradient method. Such a method is for example described in the book "SYSTEM IDENTIFICATION THEORY FOR THE USER" by Lennart Ljung

  published by PRENTICE-HALL Inc in 1987.

  This synchronization algorithm can be used in two different ways: - Recursion over several cycles: This corresponds to a Sy signal which lasts several

engine cycles.

  The gradient is calculated after each engine cycle and is used to calculate the next cycle. To converge, this way of doing things requires that the distribution profile and the average wealth be constant over several cycles: this is then the average

  of e over several cycles which is calculated.

  - Recursion on a cycle: This corresponds to a signal Sy which lasts only one cycle of the motor. The estimated delay ê is evaluated by several iterations over a cycle. Sufficient number

  iterations is performed to ensure convergence.

  In this second way of doing things, it suffices that the distribution profile and the average wealth are

constants over a cycle.

  The synchronization of the signal y (t) having been obtained as described above, the signal y (k) is applied to the regulator 22. This regulator is of the disturbance rejection type, that is to say it cancels the disturbances or variations in the richness of the exhaust gases due to a distribution of wealth

individual cylinders.

  To cancel the disturbances, the invention uses a disturbance model which must meet the following three criteria: - the disturbance of the richness of the exhaust gases due to a distribution of the individual richnesses is periodic over a cycle, - the dominant frequencies of the disturbance is that of a cycle and a half-cycle, and - the disturbance signal can have four values

different over a cycle.

  Figures 3a, 3b, 3c and 3d show a perturbation model P (q) (Figure 4) with three modes (q being the shift operator), two real modes and one complex mode. The three modes are located on the unit circle in order to be periodic. The real mode at 1 (figure 3b) makes it possible to take into account the constant errors, the second real mode -1 (figure 3d) and the complex mode (+ i, -i) (figure 3c) taking into account the cycle frequencies and half cycle. The three modes are initialized with four initial conditions, which

  allows to have four different values on a cycle.

  The disturbance model is used for the synthesis of the disturbance rejection regulator. The starting point for the synthesis is FIG. 4 which shows the combination of the perturbation model P (q) (reference) with a model S (q) (reference 42) of gas mixture in the computer and the application of a regulator R (q) (reference 44) to the combined model P (q) .S (q). A noise signal e (k) is applied to the input of the disturbance model P (q) and describes the modification of the periodic disturbance w (k) over time. The periodic disturbance signal w (k) is applied to a comparator 46 which also receives the output signal Eu (k) from the regulator 44. The difference signal supplied by the comparator 46 is applied to the mixture S (q) model. gas which provides a deviation signal Sy (k). This deviation signal constitutes the

  regulator input signal R (q).

  To cancel the periodic disturbance w (k), the regulator must stabilize the closed loop comprising the elements 42, 44 and 46. The internal model method for the synthesis of the regulator consists in choosing the regulator among the dynamic model having the following form:

R (q) = Rl (q) / (R2 (q) .P2 (q)).

  In this formula Rl (q), R2 (q) and P2 (q) are polynomials whose shift operator is q. The polynomial P2 (q) is the characteristic equation of the perturbation model. The perturbation model is defined by the quotient between a polynomial Pl (q) and the polynomial P2 (q) according to the formula:

P (q) = Pl (q) / P2 (q).

  The modes of a dynamic model correspond to the roots of its characteristic equation which determines its stability. A dynamic model is stable if its modes, represented in the complex plane, are at

  inside the complex unit circle.

  The regulator R (q) stabilizes the closed loop and cancels the periodic disturbances if all the controllable modes of the dynamic model corresponding to the closed loop are inside the unit circle. The controllable modes of the closed loop are those which can be modified by the choice of R (q). The uncontrollable modes of the closed loop are those which are invariant with respect to the choice of R (q). In this case, they correspond to the modes of the

  disturbance included in the regulator.

  The diagram in FIG. 6 shows a part of the regulator 22 (FIG. 1) comprising a comparator 50 for comparing the output signal y (k) of the synchronization circuit 20 with the reference signal yc. It provides the deviation signal Sy (k) which is applied to the regulator 44 whose output signal Su (k) is added in an adder circuit 52 to the reference signal yc to give the signal u (k) for controlling the injectors . The four injector commands ul (k), u2 (k), u3 (k) and u4 (k) are obtained by a demultiplexing device 60 of the signal u (k) as shown in FIG. 7. the demultiplexing device 60 ensures the update of the four injector controls at the

motor cycle frequency.

  The commands ul (k), u2 (k), u3 (k) and u4 (k) are applied respectively to the injectors 261, 262, 263

  and 264 each associated with a cylinder of engine 10.

  Synthesis calculations of such regulators can

  be conducted according to the LQG method (acronym

  Saxon for Linear Quadratic Gaussian) and Control Robust. The first method called LQG is for example described in the book "COMPUTER CONTROLLED SYSTEMS" by Karl J. Astrbm and Bjôrn Wittenmark published by PRENTICE-HALL International Inc. in 1984. The second method is for example described in the book "ROBUST PROCESS CONTROL "by Manfred Morari and Evanghelos Zafiriou edited by

PRENTICE-HALL Inc in 1989.

  The regulation of the individual wealth of the input mixture in each of the engine cylinders, hereinafter called "regulation of the individual wealth", is possible as long as the engine remains at the point of

  operation for which the delay ê has been identified.

  Each transition between two operating points requires a new identification of the delay ê, in order to

  ensure good synchronization of the regulator.

  Consequently, the regulation of wealth

  is not possible during a transition.

  It must be replaced by a regulation of the average wealth which is more robust with respect to transitions. Nevertheless, the identification of injector gains during the regulation of individual wealth makes it possible to cancel the periodic disturbances due to

  injector dispersions during transitions.

  Dispersion of injectors is manifested by the fact that an injector control applied to two different injectors provides two different individual cylinder richnesses. Injector gain is the quotient between the individual cylinder richness and

the injector control.

  The multiplication of the injector control with the inverse of its gain cancels the distributions of the individual richnesses of the cylinders and consequently cancels the periodic disturbances on the richness of the exhaust gases due to the dispersions of gains. The injector gains are identified during the regulation of individual wealth on the signal u (k). The identification procedure is illustrated in figure 8. An example of the input signal u (k) of the gain identification device is shown in figure 9. The demultiplexing of u (k) gives four different signals ul ( k), u2 (k), u3 (k) and u4 (k) corresponding to the injector commands. Each of the signals ul (k) to u4 (k) is then multiplied (multiplications 621, 622, 623, 624) by 1 / um, um being the average over ul (k) at u4 (k) and then filtered by a filter. low pass (641, 642, 643, 644). This results in four signals gl (k), g2 (k), g3 (k) and g4 (k) respectively converging towards four constants G1, G2, G3 and G4, as shown in Figure 10. The four constants G1 & G4 correspond respectively to the inverses of the injector gains of cylinders 1 to 4. During the regulation of the individual wealth, the ul (k), u2 (k), u3 (k) and u4 (k) commands ensure that the periodic disturbances on the richness of engine exhaust is canceled. In this case, the signal u (k) supplied by the regulator is applied as

shown in figure 7.

  During the regulation of the average richness, a signal, Ug (k) provided by a regulation of the average richness, is applied to the demultiplexer 60 as shown in FIG. 11. The signal Ug (k) is demultiplexed to generate four commands of injectors Ugl (k), Ug2 (k), Ug3 (k) and Ug4 (k). Before they are applied to the injectors, they are respectively multiplied by the constants G1, G2, G3 and G4 in order to compensate for the

injector gains.

  As soon as the engine is again in a stable operating point, the wealth regulation

individual takes over.

Claims (5)

  1. Method for canceling variations in the richness of the gas mixture coming from the cylinders of an internal combustion engine (10), said engine comprising at least one injector (26) per cylinder, a manifold (12) for transporting the gas mixture disposed at the outlet of the cylinders and a so-called proportional probe (16) disposed in said manifold (12) and providing a signal y (t) representative of the richness of the gas mixture, said signal y (t) being applied to a regulator (22 ), characterized in that it comprises the following stages consisting in: (a) synchronizing the signal y (t) to take account of the time of transport of the gas mixture in the collector, (b) regulating the individual richness of the mixture of entry into each of the engine cylinders for stable operating points by introducing into the regulator (22) a disturbance model so as to create a periodic signal u (k) canceling the expected disturbance. (c) identify injector gains on the signal u (k) during regulation for stable operating points, and (d) regulate the individual richness of the input mixture in each of the engine cylinders during the transients between two operating points by multiplying the commands of injectors by gains.   2. Method according to claim 1, characterized in that step (a) comprises the following intermediate steps consisting in: (a0) creating a signal y (r) by sampling the signal y (t) at a sampling frequency (fe) higher than that of the motor half-turn, (a1) estimate a delay (à) between the sampled signal y (T) and a model signal s (T), (a2) enter this estimated delay (ê) in the sampled signal y (r) to create a corrected signal z (T) according to determined criteria, and (a3) create a signal y (k) by resampling the signal z (T) at the frequency of the U-turn engine.   3. Method according to claim 1 or 2, characterized in that step (b) comprises the following intermediate steps consisting in: (b1) choosing a perturbation model P (q) according to determined criteria, (b2) combining P (q) with a gas mixture model S (q) in such a way that a disturbance signal w (k), corresponding to the output signal of P (q), is added to the input signal S (q ), (b3) choose a regulator R (q) for the combined model P (q) .S (q) according to determined criteria, (b4) compare the sampled signal y (k) with a setpoint signal (yc) supplied by the injection computer to create a difference signal Sy (k) in the form (y (k) - yc), (b5) apply the difference signal Sy (k) to the regulator R (q) which provides a signal 6u (k), (b6) add the signal Su (k) to the setpoint signal yc in order to create the signal u (k), (b7) demultiplex the command u (k) according to determined criteria to generate four independent commands (ul (k ), u2 (k) u3 (k) and u4 (k)) of the injectors, and   (b8) apply the four commands to the injectors.   4. Method according to any one of the claims   1 to 3, characterized in that step (c) comprises the following intermediate steps consisting in: (cl) demultiplexing the command u (k) according to determined criteria to generate four independent commands (ul (k), u2 ( k) u3 (k) and u4 (k)) of injectors, (c2) create four signals gl (k), g2 (k) g3 (k) and g4 (k) by dividing each of the four signals ul (k) , u2 (k) u3 (k) and u4 (k) by the average on ul (k) to u4 (k), (c3) filter the signals gl (k) to g4 (k) by a low-pass filter, and (c4) replace the gains G1, G2, G3 and G4 respectively by the values gl (k), g2 (k), g3 (k) and g4 (k) when the regulation according to step (b) is no longer possible.   5. Method according to any one of the claims   previous 1 to 4, characterized in that step (d) comprises the following intermediate steps consisting in: (d1) calculating the overall command Ug (k) of the injectors by regulating the conventional average richness, (d2) creating four orders of different injectors by multiplying the order of the mean richness um respectively by the gains G1, G2, G3 and G4, and   (d3) apply the four commands to the injectors.