EP0736680B1 - Method of self-correction of physical parameters in a dynamic system such as an internal combustion engine - Google Patents

Description

The present invention relates to a method for self-correcting physical parameters of a dynamic system, such as a combustion engine internal.

To optimally operate a combustion engine internal, it is already known to place it under the control of an electronic circuit of ordered. This electronic circuit manages all the necessary parameters the proper functioning of the engine.

Unfortunately, not all of these parameters can be measured. In effect to know certain parameters would require sensors placed in places very difficult to access, and for others the electronic circuit of command needs to know at a given moment the value that a setting in the future. In this case it is imperative to estimate or simulate the value of certain parameters instead of actually measuring them.

There are already many methods of simulating future value certain parameters, such as the inlet pressure in the manifold, or the amount of fuel to inject.

The problem inherent in these methods of simulation or prediction, is that it is imperative to ensure that they follow the actual value as closely as possible of the parameter. For this purpose, conventionally, the simulated value of a parameter at a given time is compared to the actual value of this parameter with this same instant. We deduce from this comparison a correction coefficient which achieves the registration of the simulated parameter on the measured parameter. However we don't come not act directly on the calculation of the simulated parameter. We are content with additive or subtractive corrections that do not correct the relationship to obtain the simulated parameter.

Such methods of resetting simulated values to values are not satisfactory because they do not allow to follow more closely variations in actual parameters, especially during the operating phases transient. These phases of transient regimes appear at each regime change, in particular in deceleration or acceleration. During these transient regimes the simulated parameters are always behind or ahead of the actual parameters and operation of the engine is not optimized.

So for example, the pressure simulation processes , known to date, consist in introducing additive corrections or multiplicative on the pressure measurement carried out before opening of the intake valves. These corrections are determined according to the point of engine operation and / or as a function of variations in the opening angle of the butterfly valve, to anticipate foreseeable variations in real air filling.

These correction elements on the intake pressure do not provide however not a real simulation of future pressure, but introduce only equivalent corrective terms, with no real physical meaning with future changes in pressure. Therefore the processes known to date do not use physical models sufficiently representative of the future state of the admission system, making it empirical, for example, the development of strategies for calculating the amount of fuel to inject at each engine cycle.

The aim of the present invention is to create a method allowing a automatic correction of discrepancies between simulated parameters and real parameters, acting directly on the relation allowing the calculation of parameters simulated and taking into account variations in real parameters between each calculation step.

• mesurer une erreur entre la valeur réelle mesurée du paramètre à un pas de calcul donné et la valeur simulée de ce paramètre à ce même pas de calcul,
• en déduire une correction à appliquer au calcul du gradient de ce paramètre, afin de modifier simultanément, pour le pas de calcul suivant, la fonction de simulation permettant d'obtenir la valeur simulée de ce paramètre et son gradient.

To this end, the present invention relates to a method of self-correction of at least one physical parameter of a dynamic system, such as an internal combustion engine, in which the evolution over time of said parameter is simulated by a function taking into account the gradient of this parameter, so that at each calculation step the simulation function calculates a simulated parameter, the said method according to the invention being characterized in that it consists in:

• measure an error between the actual measured value of the parameter at a given calculation step and the simulated value of this parameter at this same calculation step,
• deduce therefrom a correction to be applied to the calculation of the gradient of this parameter, in order to simultaneously modify, for the next calculation step, the simulation function making it possible to obtain the simulated value of this parameter and its gradient.

The method according to the invention thus allows at each calculation step, in general at each top dead center (TDC), to compare a simulated parameter and a real parameter. From this comparison, a correction coefficient is removed, not of the simulated parameter, but of the instantaneous derivative (gradient) of this setting.

This allows for variations in the gradient between each no calculation and not to consider that this gradient is constant over the horizon iteration. Consequently transient variations during an engine cycle are followed as closely as possible.

Advantageously, the present invention can be applied to all simulated parameters whose gradient is taken into account in the function of simulation. This is the case, for example, with the inlet pressure in the manifold, or with engine speed. Thus the method according to the invention makes it possible to follow closely, even during a transient regime, the variations of these settings.

Advantageously again, it will be noted that the correction coefficient determined by the method according to the invention applies equally well in static (this is to say if the regime is stabilized), that in dynamics (ie in the phases transients in engine operation). Indeed the correction coefficient determined corrects not only the agreement between simulated values and measured values, but also their derivatives (gradients or variations).

Intrinsically, we therefore correct the gradient of the system, which accentuates the interest of this process, whatever the parameter concerned (pressure or regime ... etc.) and really acts on the simulation function and not only on the final result (ie the simulated parameter).

• la figure 1 est une vue schématique illustrant la mise en oeuvre du procédé selon l'invention, et
• la figure 2 est une vue schématique illustrant le fonctionnement d'un moteur classique à quatre temps, auquel le procédé selon l'invention est appliqué, à titre d'exemple.

Other objects, characteristics and advantages of the present invention will emerge from the following description, by way of non-limiting example, and with reference to the appended drawings in which:

• FIG. 1 is a schematic view illustrating the implementation of the method according to the invention, and
• Figure 2 is a schematic view illustrating the operation of a conventional four-stroke engine, to which the method according to the invention is applied, by way of example.

With reference to FIG. 1, the self-correction method according to the invention is first described in a general framework. Secondly, as example, we will show its application to the monitoring of two specific parameters such as manifold pressure and engine speed.

The self-correction method according to the invention makes it possible to follow at most close to the variations of a parameter X, of a dynamic system. In this case, dynamic system is the internal combustion engine of a motor vehicle.

The evolution of this parameter X is simulated by a function taking into account the gradient X ˙ of this parameter. Let Xs be the value of the simulated parameter at time t. We have the following relation: Xs = f ( X ).

It emerges from this relation that the simulated value Xs of the parameter is function of the derivative X ˙ of this parameter.

As shown in FIG. 1, regularly, for example at each top dead center (TDC), a comparison is made between the simulated parameter Xs and the actual measured value X of this parameter. We deduce an error called ε according to the following relation: ε = X - Xs.

From this error, we determine by a conventional process (by example by an analog or digital PID electronic circuit) a coefficient correction ± δc. This correction coefficient ± δc is applied not to the simulated value of the parameter but at the gradient X ˙ of this parameter.

In the specific cases of monitoring the intake pressure and the engine speed, the gradient X ˙ of the physical parameter is expressed more precisely as follows: X = k [a (X) -b (X)].

The correction coefficient ± δc is then applied either to the first term a (X), or to the second term b (X) according to a confidence coefficient attached to each of these two terms. We have: X = k [a (X) -b (X) ± δc].

When the gradient of the parameter has thus been corrected, the parameter Xs simulated for the next step using the corrected gradient. In consequence it is the simulation function f (X ˙) which is thus modified to follow as closely as possible the variations of the real parameter.

It should be noted that the coefficient ± δc, can be interpreted as representing a variation of the first term a (X) or the second term b (X), or apply both terms in a balanced way (half on each term) or unbalanced. It all depends on the measurement and simulation conditions and is left to the discretion of those skilled in the art. The important thing is that correction coefficient is applied in its entirety to the calculation of the gradient of the parameter X.

This self-correction method according to the invention will be better understood at using a first application example in which the parameter X is the manifold inlet pressure. To this end, to better replace the invention in context, the operation of a four-stroke engine is briefly recalled.

We have chosen to illustrate in Figure 2 the application of the process according to the invention, a four-stroke internal combustion engine. Of course those skilled in the art can extrapolate this example to combustion engines internal presenting a cycle of any type (two or more times).

Such an engine 10 has four cylinders 11 (only one is shown in Figure 2) which, during an engine cycle, fill with a mixture air / fuel. Each cylinder 11 is supplied with a mixture when a valve intake 12 formed in this cylinder opens. Upstream of this valve intake 12 there is a manifold 13, optionally provided with an air filter 14. The manifold 13 is provided with a throttle valve 15, the role of which is to leave more or less penetrate air inside the collector. This throttle 15 is coupled to an accelerator pedal (not shown in Figure 2) operated by a driver.

By pressing more or less on the accelerator pedal, the driver varies the opening angle a of the butterfly which results in vary the amount of air admitted into the cylinder.

A fuel injector 16, sends in the collector, a quantity of fuel predetermined by a computer 18.

When the inlet valve 12 of the cylinder 11 opens, the mixture air / fuel accumulated in the manifold 13, enters the cylinder in question.

FIG. 2 also shows a system for regulating idle r, an All ignition device of the compressed air / fuel mixture in the cylinder and a catalyst 17 recycling the exhaust gases discharged by the cylinder 11. These devices of known type are not detailed.

The electronic computer 18, associated with the engine 10 receives the value of the pressure P prevailing in the intake manifold 13. This pressure is measured by an appropriate sensor 19, known per se. The computer 18 is also kept informed by suitable sensors of the speed of rotation of the engine N (engine speed), water temperature , etc.

One of the functions of the electronic calculator is to calculate the quantity of fuel to be injected into a cylinder, to make a mixture air / fuel in specified proportions. For this purpose the calculator needs know the pressure prevailing in the intake manifold when the valve admission will close. We consider that the pressure prevailing in the intake manifold at the time a valve closes equals the pressure in the cylinder in question. Knowing the prevailing pressure in a cylinder, and knowing the volume of this cylinder and the temperature of gas, we deduce the amount of air present inside the cylinder. The amount of gas which had to be injected to have an air / fuel mixture in given proportions is therefore easy to deduce. It is therefore important to be able to predict the pressure that will prevail in the manifold of a cylinder when the This cylinder's intake valve opens and closes.

The method according to the present invention makes it possible to correct automatically all the differences that may exist between the intake pressure simulated or predicted and the actual intake pressure as measured.

The simulated inlet pressure Ps is a function of the inlet pressure gradient as measured. So we have: Ps = f ( P ).

The gradient P ˙ of the intake pressure is expressed as follows: P = 1 VS × [Q (P, Pa, α) -Q (P, N)].

In this relation the coefficient 1 / C is known, Pa is the pressure atmospheric, the first term Q (P, Pa, α) is representative of the air flow entering the intake manifold, and the second term Q (P, N) is representative of the cylinder filling.

The first and second terms of the pressure derivative are determined by mapping for each type of engine.

According to the invention, the difference ε = P - Ps is measured at a given time. we deduce from this difference a correction coefficient ± δc, which we apply to calculation of the derivative of the intake pressure P ˙.

When the engine is running at medium or heavy loads the correction coefficient ± δc is applied to the first term Q (P, Pa, α), because it is this term which has the lowest confidence coefficient. Indeed this term varies the most without any direct means of verifying the importance of these variations.

In the example given, the flow rate Q (P, Pa, α) indeed reflects the variations in throttle flow, idle control valve and altimetric correction. However, all of these variables are difficult to measure, and highly unstable. The term Q (P, N) is better known and its variations are easier to determine.

If the engine is running at low loads, i.e. at slowed down, unlike the above it is the term Q (P, N) which presents the lowest confidence coefficient. Indeed the incoming air flow Q (P, Pa, α) is weak and regulated since the engine idles, however due to motor suction phenomena (motor pulsations at low loads) it is the flow Q (P, N) which is difficult to control.

In this case the correction coefficient of ± δc applies to the second term of the relation allowing the computation of the derivative P ˙.

Thus the comparison between the value of the pressure predicted at a step of calculation and the actual value of this pressure at this same step, is not directly used to reset the predicted value to the measured value using a additive correction coefficient, but is used to redefine the relationship allowing the calculation of the predicted pressure.

Likewise, the method according to the present invention can be applied to self-correction of the engine rotation speed N, in relation to the value simulated Ns of this scheme.

The speed of an internal combustion engine can be simulated according to the following law: Ns = g ( NOT ).

The gradient N ˙ of this regime is given by the following relation: NOT = k [Γ 1 (N) -Γ 2 (NOT)].

In this relation Γ ₁ is the driving torque and Γ ₂ is the resisting torque. In the same way as previously described, an error ε = N - Ns is measured at a given time. From this error, a correction coefficient ± δc is conventionally drawn which is applied to the torque (motor or resistor) having the lowest confidence coefficient. In the case of monitoring the engine speed, the torque whose variations are the most difficult to simulate is the resisting torque. According to the invention, the correction coefficient ± δc therefore applies as follows: NOT = k [Γ 1 (N) -Γ 2 (N) ± δc].

This corrected gradient makes it possible to recalculate a simulated rotation regime taking account of the variations of the gradient, and therefore as closely as possible the variations of the real parameter N. In addition the term [Γ ₂ (N) ± δc] is an image of the resistant couple, that it is difficult to apprehend by known measurement means.

Of course, the self-correction method according to the invention is not limited to the embodiments described above. So this self-correcting process can be implemented to correct discrepancies between a parameter simulated according to a certain model (for example a so-called failed model) with this same parameter but simulated according to a second model (for example a model not readjusted says "free").

Claims (8)

1. Method of self-correction of at least one physical parameter X of a dynamic system of a technical process, such as an internal combustion engine, in which the evolution over time of the said parameter is simulated by a function that takes account of the gradient X ˙ of this parameter, so that on each calculation step, the simulation function calculates a simulated parameter Xs, the method according to the invention consisting in:

   measuring an error (ε) between the actual measured value (X) of the parameter on a given calculation step and the simulated value (Xs) of this parameter on this same calculation step,

   and being characterized in that it consists in:

   from this, deducing a correction (±δc) to be applied to the calculation of the gradient (X ˙) of this parameter, so as to modify simultaneously, for the next calculation step, the simulation function making it possible to obtain the simulated value (Xs) of this parameter and its gradient (X ˙).
2. Method according to Claim 1, characterized in that the gradient (X ˙) of the parameter can be expressed by the following relationship: X = k[a(X)-b(X)], and in that the correction coefficient ±δc is applied to the first term a(X) or to the second term (b(X) depending on a confidence coefficient associated with each of these two terms.
3. Method according to Claim 2, characterized in that the parameter monitored is the inlet pressure (P) of an internal combustion engine, and in that the gradient P ˙ of this pressure is given by the following relationship: P=1V ×[Q(P,Pa,α)-Q(P,N)], in which 1/C is a known coefficient, the first term Q(P,Pa,α) represents an airflow rate entering the inlet manifold, the second term Q(P,N) represents an amount of mixture actually absorbed by a cylinder, (P) is the inlet pressure, (Pa) is atmospheric pressure, (α) is the throttle-valve angle, and (N) is the rotational speed of the engine.
4. Self-correction method according to Claim 3, characterized in that when the engine is at heavy or medium load, the correction coefficient (±δc) is applied to the first term.
5. Method according to Claim 3, characterized in that when the engine is at light load, the correction coefficient (±δc) is applied to the second term.
6. Method according to Claim 2, characterized in that the parameter monitored is the rotational speed (N) of an internal combustion engine, the gradient (N ˙) of this parameter being expressed by the following relationship: N=k(Γ1(N)-Γ2(N)), in which Γ₁ is an engine torque and Γ₂ is a resistive torque, the coefficient (±δc) being applied to the torque which has the lowest confidence coefficient.
7. Method according to Claim 6, characterized in that the corrected gradient of the rotational speed of the engine can be expressed by the following relationship: N=k(Γ1(N)-Γ2(N)±δc).
8. Method according to one of the preceding claims, characterized in that the parameter which evolves over time is itself a simulated parameter.

Images