EP2549087B1 - Method for detecting and characterising abnormal combustion in internal combustion engines - Google Patents

Description

The present invention relates to the field of controlling the combustion phase of an internal combustion engine. In particular, the present invention relates to a method for detecting an abnormal combustion of the pre-ignition type at low speed and at high load, in a combustion chamber of such an engine.

It relates more particularly but not exclusively such a method applied to a spark ignition engine "downsized" operating under very heavy loads.

Spark ignition engines have the advantage of limiting local pollutant emissions (HC, CO and NO _X ) thanks to the excellent match between the operating mode (with richness 1) and their simple and low post-processing system. cost. Despite this fundamental advantage, these engines are poorly positioned in terms of greenhouse gas emissions because competing diesel engines can achieve CO ₂ emissions that are 20% lower on average.

The combination of downsizing and supercharging is one of the solutions that is spreading more and more to lower the consumption of spark ignition engines. Unfortunately, the conventional combustion mechanism in these engines can be disturbed by abnormal combustions. This type of engine comprises at least one cylinder having a combustion chamber defined by the inner side wall of the cylinder, the top of the piston which slides in this cylinder and the cylinder head. Generally, a fuel mixture is enclosed in this combustion chamber and undergoes a compression step, then a combustion step under the effect of a controlled ignition, by a candle. These steps are grouped under the term "combustion phase" in the remainder of the description.

It has been found that this fuel mixture can undergo different types of combustion, and that these types of combustion are the source of different levels of pressure, as well as mechanical and / or thermal stresses, some of which can severely damage the engine.

The first combustion, known as conventional combustion or normal combustion, is the result of the propagation of the combustion of a compressed fuel mixture during a preliminary stage of compression of the engine. This combustion propagates normally according to a flame front from the spark generated at the candle, and does not risk damaging the engine.

Another type of combustion is a knocking combustion, which results from undesirable self-ignition in the combustion chamber. Thus, after the step of compressing the fuel mixture, the spark plug is actuated to enable the ignition of this fuel mixture. Under the effect of the pressure generated by the piston, and the heat released by the beginning of the combustion of the fuel mixture, there is a sudden and localized auto-ignition of a portion of the compressed carburetted mixture, before n ' comes the flame front from the ignition of the fuel mixture by the candle. This mechanism, called rattling, leads to a local increase in pressure and temperature and can cause, in case of repetitions, destructive effects on the engine and mainly at the piston.

Finally, another type of combustion is an abnormal combustion due to a pre-ignition of the fuel mixture before the spark ignites ignition of the fuel mixture present in the combustion chamber.

This abnormal combustion particularly affects the engines that are the result of a "miniaturization" operation, better known as "downsizing". This operation aims to reduce the size and / or the engine displacement while maintaining the same power and / or the same torque as conventional engines. Generally, this type of engine is mainly gasoline type and is highly supercharged.

It has been found that this abnormal combustion is carried out at high loads, and generally at low operating speeds of the engine, when the timing of the combustion of the fuel mixture can not be optimum because of the rattling. Given the high pressures and high temperatures reached in the combustion chamber by the supercharging, an abnormal combustion start can occur, sporadically or continuously, well before the moment when ignition of the fuel mixture by the candle is achieved . This combustion is characterized by a first phase of flame propagation that is wedged too early compared to that of a conventional combustion. This propagation phase can be interrupted by a self-ignition that will affect a large part of the fuel mixture present in the combustion chamber, much larger than in the case of rattling (up to 50%, against 5 to 10% for extreme cases of intense rattling).

In the case where this abnormal combustion occurs repeatedly, engine-cycle engine-cycle, and is carried out from a hot spot of the cylinder for example, it is called "hot surface pre-ignition". If this combustion occurs violently, randomly and sporadically, it is called "snapping" or "rumble" ("pre-ignition").

This last abnormal combustion leads to very high pressure levels (120 to 250 bar), as well as an increase in heat transfer that can lead to a partial or total destruction of the mobile engine, such as the piston or the piston. rod. This type of pre-ignition currently constitutes a real limit to the miniaturization ("downsizing") of spark ignition engines. This is a very complex phenomenon that can have multiple origins. Several hypotheses have been mentioned in the literature to explain its appearance, but none of them has been clearly validated for the time being. It even seems that many of these potential causes are simultaneously manifesting and interacting with each other. This interaction, the violence of the phenomenon and its stochastic nature make its analysis extremely complicated. Moreover, the various studies on the subject all face the problem of the very identification of these abnormal combustions. It is indeed difficult to say whether an engine is more sensitive than another to pre-ignition as long as we are unable to decide on the nature of each combustion within a given sample.

A method for detecting and characterizing in number and intensity these abnormal combustions is therefore of primary necessity, because it makes it possible precisely to establish this hierarchy and to identify the tracks which make it possible to improve the design and the adjustments of the engines. This operation is particularly interesting during engine engine development.

State of the art

The general methodology for treating these abnormal combustions is schematized on the figure 1 , with initially a phase of prevention (PP) to limit as much as possible the chances of occurrence of the phenomenon, then a phase of detection (PD) when the prevention was not enough to avoid the phenomenon, to determine if yes or not it is necessary to intervene in the same cycle where the pre-ignition was detected by means of a corrective phase (PC).

The detection phase includes a signal acquisition phase and then a signal processing phase to detect the appearance of the high-load pre-ignition, to characterize it and to quantify it.

It is known from the patent application EP 1.828.737 , a method for detecting the appearance of pre-ignition at high load, rumble type . This method is based on the measurement of a signal relating to the course of combustion, and a comparison with a threshold signal. The presence of abnormal "rumble" combustion in the combustion chamber is detected when the signal amplitude significantly exceeds that of the threshold signal. According to this method, the threshold signal corresponds to the amplitude of the signal produced during a combustion with knocking or during a normal combustion (conventional).

However, according to this method, the detection thus made does not make it possible to act during the very cycle of the detection. Corrective actions of this type of pre-ignition can therefore be performed only after the occurrence of such a phenomenon, which can seriously affect the integrity of the engine.

The method described in the patent is also known. FR 2,897,900 . According to this method, it is possible to act more quickly after the detection of the pre-ignition: one is able to act during the same cycle as the detection cycle of the phenomenon. To do this, the threshold signal is previously calculated, that is to say before the operation of the engine, and then stored in data tables of the computer, called maps.

However, the use of cartographies, does not detect at any time, that is to say, in real time, the beginning of such a phenomenon. Therefore, it is always possible that the detection is done too late. Moreover, no quantification of the evolution of the phenomenon can be achieved. Thus, the need or not to apply a correction phase is based solely on the comparison of two amplitudes at a given instant. Or such a phenomenon may well begin, then stop without causing damage to the engine, and therefore do not require a corrective phase.

It is known from the patent application FR 2.952.678 an abnormal combustion detection method for internal combustion engines with spark ignition, from several indicators of combustion. According to this method, several combustion indicators, such as CA10 and PMI, are determined and these indicators are transformed into new indicators with lower dispersions than those of unprocessed indicators for normal combustions. Then, we determine a parameter characterizing a distribution of N values of these new indicators of the combustion, acquired on N cycles preceding the current cycle. The beginning of an abnormal combustion is then detected by comparing this parameter with a threshold, and the unfolding of the abnormal combustion detected in the combustion chamber is monitored.

All these previous methodologies aim to quantify the frequency of occurrence of pre-ignition, without providing a good representation of the violence (intensity) of these detected phenomena. However, the cylinder heads can only be dimensioned in a relevant way if the potential frequency and intensity of pre-ignitions are known.

The object of the invention relates to a method for detecting in real time the occurrence of an abnormal combustion, to characterize its frequency of occurrence and its intensity, with the devices and systems commonly used in engines, so as to take measures to prevent this in the further operation of the engine, course of the same cycle as that of detection. The method is based on the definition of a multidimensional space where each dimension corresponds to an indicator of combustion, and on the definition in this space, of a closed surface delimiting the normal combustions of the abnormal combustions. The position and the distance of a point corresponding to a combustion, with respect to this surface, makes it possible to qualify the abnormal character of this combustion; as well as the severity of this abnormality.

The process according to the invention

• on choisit des indicateurs de la combustion que l'on peut déduire dudit signal, et l'on définit un espace multidimensionnel dont chaque dimension correspond à un desdits indicateurs, et dans lequel toute combustion peut être représentée par un point ;
• on définit dans ledit espace une surface fermée de façon à envelopper des points correspondant à des combustions normales et à ne pas envelopper des points correspondant à des combustions anormales ; dues à un pré-allumage puis pour chaque combustion d'un cycle moteur :
  • on représente ladite combustion du cycle par un point dans ledit espace multidimensionnel en déterminant pour cette combustion lesdits indicateurs ;
  • on détermine une position dudit point par rapport à ladite surface et l'on en déduit un caractère anormal de ladite combustion ;
  • on détermine une distance entre ledit point et ladite surface, et l'on en déduit une sévérité du caractère anormal ; et
  • on contrôle le déroulement de la combustion anormale détectée en fonction de la sévérité du caractère anormal. On définit la surface en réalisant les étapes suivantes :
    • on choisit une équation définissant ladite surface, ladite équation comportant au moins un paramètre ;
    • on réalise un ensemble de combustions dans lequel des combustions normales et des combustions anormales sont connues, et l'on représente ledit ensemble de combustions dans ledit espace multidimensionnel formant un nuage de points ;
    • on détermine au moyen d'une analyse en composantes principales des directions principales dudit nuage de points, et on détermine une dispersion des points selon chaque direction principale ;
    • on modifie ledit paramètre de façon à ce que l'extension de la surface dans chaque direction principale soit égale à la dispersion dans cette direction. Selon un mode de réalisation, on peut définir un coefficient multiplicateur que l'on applique à chaque dispersion avant la modification du paramètre. Ce coefficient multiplicateur peut être choisi entre 2,4 et 2,6, de préférence égale à 2,5.

In general, the invention relates to a combustion control method of a spark-ignition internal combustion engine, in which at least one signal representative of a state of combustion is recorded by means of at least one sensor placed on the engine. The method comprises the following steps:

• combustion indicators are selected which can be deduced from said signal, and a multidimensional space is defined in which each dimension corresponds to one of said indicators, and in which any combustion can be represented by a point;
• defining in said space a closed surface so as to envelop points corresponding to normal combustions and not to wrap points corresponding to abnormal combustions; due to a pre-ignition then for each combustion of an engine cycle:
  • said combustion of the cycle is represented by a point in said multidimensional space by determining for said combustion said indicators;
  • determining a position of said point relative to said surface and deducing an abnormal character of said combustion;
  • a distance between said point and said surface is determined, and a severity of the abnormality is deduced therefrom; and
  • the course of the abnormal combustion detected is checked as a function of the severity of the abnormal character. The surface is defined by carrying out the following steps:
    • an equation defining said surface is chosen, said equation comprising at least one parameter;
    • a set of combustions is made in which normal combustions and abnormal combustions are known, and said set of combustions is represented in said multidimensional space forming a cloud of points;
    • a principal component analysis of the principal directions of said cloud of points is determined, and a dispersion of the points according to each principal direction is determined;
    • said parameter is modified so that the extension of the surface in each main direction is equal to the dispersion in this direction. According to one embodiment, it is possible to define a multiplying coefficient that is applied to each dispersion before the modification of the parameter. This multiplier coefficient can be chosen between 2.4 and 2.6, preferably equal to 2.5.

According to the invention, said surface can be updated from a point resulting from a new combustion. Said surface may be a quadric type surface.

Finally, according to one embodiment, the indicators are standardized.

Presentation of figures

• la figure 1 montre la méthodologie générale de traitement des combustions anormales de type pré allumage ;
• la figure 2 montre un moteur utilisant la méthode de détection selon l'invention ;
• la figure 3 donne un exemple de représentation tridimensionnelle de données calculées sur un point de fonctionnement avec pré allumage ;
• la figure 4 présente une superposition de données normalisées obtenues sur différents points de fonctionnement ;
• la figure 5 illustre un exemple d'identification des directions principales - Zoom sur la figure de droite (les axes principaux peuvent ne pas paraître orthogonaux à cause des différentes échelles) ;
• la figure 6 illustre la détermination de l'épaisseur optimale de la surface de normalité ;
• la figure 7 représente une estimation de l'enveloppe des combustions normales par une surface quadrique en utilisant un coefficient multiplicateur de 2.5 ;
• la figure 8 illustre la distance des pré allumages à la normalité (représentée par la taille des cercles).

The other features and advantages of the invention will appear on reading the description given below with reference to the appended figures in which:

• the figure 1 shows the general methodology for the treatment of abnormal combustions of the pre-ignition type;
• the figure 2 shows a motor using the detection method according to the invention;
• the figure 3 gives an example of a three-dimensional representation of calculated data on an operating point with pre-ignition;
• the figure 4 presents a superposition of standardized data obtained on different operating points;
• the figure 5 illustrates an example of identifying the main directions - Zoom on the figure on the right (the main axes may not appear orthogonal because of the different scales);
• the figure 6 illustrates the determination of the optimal thickness of the normality surface;
• the figure 7 represents an estimate of the envelope of normal combustions by a quadric surface using a multiplying coefficient of 2.5;
• the figure 8 illustrates the distance of pre-ignitions to normality (represented by the size of the circles).

Detailed description of the process

On the figure 2 a supercharged spark ignition internal combustion engine, in particular of the gasoline type, comprises at least one cylinder 12 with a combustion chamber 14 inside which combustion of a mixture of supercharged air and fuel.

The cylinder comprises at least one pressurized fuel supply means 16, for example in the form of a fuel injector 18 controlled by a valve 20, which opens into the combustion chamber, at least one intake means of 22 with a valve 24 associated with an intake manifold 26 terminating in a plenum 26b (not shown in the figure), at least one exhaust gas exhaust means 28 with a valve 30 and an exhaust manifold 32 and at least one ignition means 34, such as a spark plug, which makes it possible to generate one or more sparks making it possible to ignite the fuel mixture present in the combustion chamber.

The pipes 32 of the exhaust means 28 of this engine are connected to an exhaust manifold 36 which is itself connected to an exhaust line 38. A supercharging device 40, for example a turbocharger, is placed on this line. exhaust and comprises a drive stage 42 with a turbine swept by the exhaust gas flowing in the exhaust line and a compression stage 44 which makes it possible to admit an intake air under pressure into the combustion chambers 14 through the intake manifolds 26.

The engine comprises means 46a for measuring the cylinder pressure, arranged within the cylinder 12 of the engine. These measuring means are generally constituted by a pressure sensor which makes it possible to generate a signal representative of the evolution of the pressure in a cylinder.

The engine may also include means 46b for measuring the intake pressure, arranged in the plenum 26b. These measuring means are generally constituted by an absolute pressure sensor, of piezoelectric type, which makes it possible to generate a signal representative of the evolution of the admission pressure in the intake plenum.

The engine also comprises a calculation and control unit 48, called engine calculator, which is connected by conductors (for some bidirectional) to different components and sensors of the engine so as to be able to receive the various signals emitted by these sensors, such as the temperature of the water or the temperature of the oil, to treat them by calculation and then to control the organs of this engine to ensure its smooth operation.

Thus, in the case of the example shown in figure 2 , the spark plugs 34 are connected by conductors 50 to the engine control unit 48 so as to control the moment of ignition of the fuel mixture, the cylinder pressure sensor 46a is connected by a line 52 to the same engine computer to send the signals to it. Representative of the evolution of the pressure in the cylinder, and the control valves 20 of the injectors 18, are connected by conductors 54 to the computer 48 to control the injection of fuel into the combustion chambers. The means 46b are also connected by a line 53 to the motor calculator 48.

Within such an engine, the method according to the invention makes it possible to detect the occurrence of a high-load pre-ignition phenomenon (of the rumble type), to characterize its frequency of occurrence and its intensity, in particular pressing on a simultaneous characterization of values of several indicators of combustion (CA10, PMI, ...).

• la première de ces étapes concerne les actions de prévention qui ont pour but de limiter au maximum les chances d'apparition du pré allumage ;
• dans le cas où cette phase de prévention ne suffit pas, une seconde étape de détection physique du pré allumage doit être mise en oeuvre (par un choix de capteurs par exemple) ;
• vient ensuite une phase de traitement des données qui doit permettre de caractériser le pré allumage ;
• et finalement, une dernière phase d'action corrective est réalisée pour déterminer si oui ou non il y a lieu d'intervenir dans le cycle même où le pré allumage a été détecté ou au cours des cycles qui suivent.

The general methodology for treating these abnormal combustions involves several steps:

• the first of these steps concerns preventive actions that aim to minimize the chances of the occurrence of pre-ignition;
• in the case where this prevention phase is not sufficient, a second step of physical detection of the pre-ignition must be implemented (by a choice of sensors for example);
• then there is a data processing phase which should make it possible to characterize the pre-ignition;
• and finally, a last phase of corrective action is performed to determine whether or not it is necessary to intervene in the same cycle where the pre-ignition was detected or in the cycles that follow.

• on enregistre au moins un signal (pression dans le cylindre) représentatif de l'état de la combustion au moyen d'au moins un capteur placé dans le moteur ;
• on choisit des indicateurs de la combustion que l'on peut déduire de ce signal, et l'on définit un espace multidimensionnel dont chaque dimension correspond à un des indicateurs, et dans lequel toute combustion peut être représentée par un point ;
  • on définit dans cet espace une surface fermée de façon à envelopper des points correspondant à des combustions normales et à ne pas envelopper des points correspondant à des combustions anormales ;
  puis pour chaque combustion d'un cycle moteur :
  • on représente la combustion du cycle en cours par un point dans l'espace multidimensionnel en déterminant pour cette combustion les indicateurs ;
  • on détermine la position du point par rapport à la surface et l'on en déduit le caractère anormal de la combustion en cours ;
  • on détermine la distance entre le point et la surface, et l'on en déduit la sévérité du caractère anormal ; et
  • on contrôle le déroulement de la combustion anormale détectée en fonction de la sévérité du caractère anormal.

The invention falls within the scope of the third step. According to an exemplary embodiment, the method comprises the following steps:

• at least one signal (pressure in the cylinder) representative of the state of combustion is recorded by means of at least one sensor placed in the engine;
• we choose indicators of the combustion that can be deduced from this signal, and we define a multidimensional space where each dimension corresponds to one of the indicators, and in which any combustion can be represented by a point;
  • a closed surface is defined in this space so as to envelop points corresponding to normal combustions and not to envelop points corresponding to abnormal combustions;
  then for each combustion of an engine cycle:
  • the combustion of the current cycle is represented by a point in the multidimensional space by determining for this combustion the indicators;
  • the position of the point with respect to the surface is determined and the abnormal character of the current combustion is deduced therefrom;
  • the distance between the point and the surface is determined, and the severity of the abnormality is deduced; and
  • the course of the abnormal combustion detected is checked as a function of the severity of the abnormal character.

At least one signal representative of the state of combustion is recorded by means of a sensor placed in the engine. According to one embodiment, the cylinder pressure is chosen. The measurement of the cylinder pressure is carried out from the means 46a for measuring the cylinder pressure. Cylinder instrumentation for pressure measurement is becoming more common on vehicles.

The invention makes it possible to use measurements other than cylinder pressure, such as instantaneous torque, instantaneous speed, vibration level (accelerometer sensors), ionization signal, etc.

Then, a preliminary phase (steps 1 and 2 below) is carried out to detect in real time an abnormal combustion.

1. Choice of combustion indicators and definition of a multidimensional space

During this step, combustion indicators are selected that can be deduced from the measured signal, and a multidimensional space is defined in which each dimension corresponds to one of the indicators, and in which any combustion can be represented by a point.

According to one embodiment, the CA10 is chosen. The CA10 represents the crankshaft angle at which only 10% of the feed introduced was consumed. Because of this, it is particularly well suited to highlighting an anomaly occurring at the beginning of combustion, such as pre-ignition.

However, a simple identification of pre-ignitions is not sufficient since the aim is also to characterize the dangerousness of these abnormal combustions. Therefore, it is necessary to also choose variables that explicitly represent the violence of pre-ignitions.

The figure 3 gives an example of three-dimensional representation of calculated data on an operating point with pre-ignition. In addition to the CA10, the CA10 pressure (PCA10) and pressure derivative (DPCA10) values were retained. Intuitively, it is understood that the values taken by the pressure and the pressure derivative at CA10 are decisive for the values which they will then take during the cycle, in particular for their maximum values (in other words, a combustion which strong share is very likely to continue and finish very strong ...).

• à partir de la pression cylindre : PMI, pression cylindre maximale, angle vilebrequin à la pression maximale, CAxx, maximum de dégagement d'énergie,... ;
• à partir du couple instantané : maximum de couple, dérivée maximale de couple, ... ;
• à partir du régime instantané : maximum de vitesse, accélération maximale, ... ;
• le volume de la chambre de combustion, ou le gradient de volume à certains moments (au CA10 par exemple).

The invention can also use other combustion indicators:

• from cylinder pressure: PMI, maximum cylinder pressure, crankshaft angle to maximum pressure, CAxx, maximum energy release, ...;
• from the instantaneous torque: maximum torque, maximum torque derivative, ...;
• from instantaneous speed: maximum speed, maximum acceleration, ...;
• the volume of the combustion chamber, or the volume gradient at certain times (CA10 for example).

Several tests of three-dimensional representations of data such as that of the figure 3 have been made on different operating points, but also on different engines and with different fuels. Systematically, these tests revealed correlations between the different variables and the standardization of the data made it possible to show that these tendencies were repeatable. The figure 4 presents a superposition of standardized data (CA10n, PCA10n, DPCA10n) obtained on different operating points. Normal combustions occupy a fairly compact area of space and form a condensed cloud of data while pre-ignitions tend to come out of this cloud (as well as late but to a lesser extent).

2. Definition of a closed surface delimiting normal combustions

An object of the invention is to delimit normal combustions to then more easily extract information on abnormal combustions in terms of distance to normality.

To do this, a closed surface is defined in the multidimensional space so as to envelop points corresponding to normal combustions and not to envelop points corresponding to abnormal combustions.

To do this, we can use a first set of points corresponding to normal combustions of the engine, and a second set of points corresponding to abnormal combustion of the engine. These sets are represented in the multidimensional space, and one adjusts a surface enveloping the points corresponding to the first set, avoiding the points of the second set.

1. i. en identifiant tout d'abord les directions principales présentes au sein du jeu de données (les directions sont représentées par des flèches blanches sur la figure 5) ;
2. ii. puis en déterminant une modélisation pertinente des combustions normales (figure 7).

An exemplary implementation is described below, in which the delimitation is carried out in two stages:

1. i. by first identifying the main directions present within the dataset (the directions are represented by white arrows on the figure 5 );
2. ii. then determining a relevant modeling of normal combustions ( figure 7 ).

Each combustion is represented in the multidimensional representation space in the form of a point whose coordinates are the values of the indicators calculated in the previous step. After several cycles, the combustions form a cloud of points in this space of representation.

First, we determine the principal directions of this cloud of points, that is to say the directions in which the cloud extends, or in other words the directions in which the dispersion is maximum. In one example, the identification of the main directions is robustly performed via a Principal Component Analysis (PCA) algorithm. But other algorithms can be used for this purpose.

The figure 5 illustrates an example of identification of the main directions: the main axes are represented by white arrows; they may not appear orthogonal because of different scales.

In a second time, we build an envelope (surface) around the points corresponding to normal combustions. It is essential to correctly determine this optimal surface, which should not be too large (the risk would then be to include pre-ignitions), nor too small (the risk would be to over estimate the number of pre-ignitions in considering some normal combustions as abnormal). To build this envelope, one chooses a form of envelope, then one adjusts it to the cloud of points along the principal directions of the cloud.

According to one example, the first three main directions are calculated, and a quadric type envelope is chosen (other types of surface could also be used). A quadric, or quadratic surface, is a surface of the Euclidean space of dimension 3, place of the points satisfying a cartesian equation of degree 2. We can quote for example: the ellipsoid, the hyperboloid, the elliptic paraboloid, the paraboloid hyperbolic, the cylinder (elliptical, hyperbolic or parabolic).

The parameters of the quadric surface are then adjusted so that it is centered on the center of the cloud.

• x, y, et z représentent les trois directions principales formant un repère orthonormé, dont le centre est le centre du nuage de points.
• a, b et c sont les paramètres de la surface quadrique à ajuster.

In the case of an ellipsoid, the equation is: $$\frac{x^2}{{\mathrm{at}}^2}+\frac{{\mathrm{there}}^2}{{\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="imgf0003.png,imgf0005.png"\; state="\{\{state\}\}">b</figure-callout>}}^2}+\frac{{\mathrm{<figure-callout\; id="2"\; label="z"\; filenames="imgf0003.png,imgf0005.png"\; state="\{\{state\}\}">z</figure-callout>}}^2}{{\mathrm{vs}}^2}-1=0$$

• x, y, and z represent the three principal directions forming an orthonormal coordinate system, the center of which is the center of the point cloud.
• a, b and c are the parameters of the quadric surface to be adjusted.

Then, we estimate the dispersion (for example the standard deviation) of the data on each of the main directions. This estimate can be advantageously performed following the PCA, that is to say at the same time as the determination of the main directions. The dispersion on each of the principal directions x, y and z defines the extension of the quadric surface: the parameters a, b and c are chosen so that the extension of the quadric surface in the direction x (respectively y and z) is equal to the dispersion in the x direction (respectively y and z).

According to one embodiment, a multiplying coefficient is calculated for the calculated dispersions. The progressive increase of this multiplying coefficient makes it possible to increase the size of the surface enveloping the normal combustions. Thus, according to our example of a quadric surface of ellipsoid type, the parameters a, b and c are chosen so that the extension of the quadric surface in the direction x (respectively y and z) is equal to the dispersion in the direction x (respectively y and z), multiplied by a multiplying coefficient.

Using a synthetic dataset, in which the normal combustions and pre-ignited combustions are known, can define these multiplier coefficients. The observation made is illustrated on the figure 6 . The goal is to determine the inflection (PI) of the curve (C) representing the number of points (n) contained within the normality surface, as a function of the multiplier coefficient (CM). Indeed, this inflection corresponds to the moment when, in spite of the increase of the size of the surface of normality, less and less points enter this surface. We then reach the separation between normal and abnormal combustions which are much more dispersed, and which therefore require higher multiplying coefficients to be included in the normality surface. In this figure, a multiplying coefficient of 2.5 thus makes it possible to encompass all normal combustions. The same approach applied to nearly 600 manually generated datasets led to the same result with a value around 2.5 (between 2.4 and 2.6).

The figure 7 represents an estimate of the envelope of normal combustions by a quadric surface using a multiplying coefficient of 2.5.

This surface defined before the detection phase at each cycle of an abnormal combustion, can be refined at each cycle, by integrating the point cloud from the combustions of the cycles preceding the current cycle.

Once these preliminary steps (definition of a multidimensional space and a reference surface) have been carried out, it is possible, from the signal, to detect an abnormal combustion at each engine cycle.

3. Identification and qualification of abnormal combustions

• on représente la combustion du cycle en cours par un point dans l'espace multidimensionnel en déterminant pour cette combustion les indicateurs ;
• on détermine la position du point par rapport à la surface et l'on en déduit le caractère anormal de la combustion en cours ;
• on détermine la distance entre le point et la surface, et l'on en déduit la sévérité du caractère anormal.

During each cycle, the combustion indicators are calculated from the signal and for each combustion. Then, the following steps are carried out:

• the combustion of the current cycle is represented by a point in the multidimensional space by determining for this combustion the indicators;
• the position of the point with respect to the surface is determined and the abnormal character of the current combustion is deduced therefrom;
• the distance between the point and the surface is determined, and the severity of the abnormality is deduced.

A method for calculating the distance from a point to an ellipsoid is described for example in the following document: David Eberly, 2011, "Ellipsoid's Ellipsoid, Ellipsoid, a Hyperellipsoid", Geometric Tools, LLC . In the case of a quadric surface of degree dg, a calculation method consists in calculating the distance d1 from the point to the ellipsoid of the same parameters a, b, c, which provides a good practical approximation of the exact distance. Another possible type of calculation consists of determining the radial line joining the center of the quadric surface at the point considered, then calculating the smallest "radial" distance d2 between the point considered and the two intersections (the radial line generally intersects the two-point surface: one close, the other farther (on the other side of the center), take the distance to the nearest point.) between the quadric surface and the radial line, and to consider the smaller of the two distances d1 and d2. The distance can thus be slightly overestimated, which preserves the preventive aspect of the proposed detection.

This distance is an indicator of the combustion at each cycle. If the distance indicates that the point characterizing the combustion is outside the cloud, this indicates a pre-ignition, and the greater this distance, the greater the intensity of the phenomenon is important.

The simultaneous taking into account of several variables thus makes it possible to construct through this distance a "combined" criterion (and even if these different variables had to be partially correlated).

The figure 8 illustrates this distance by the size of the circles used to represent the different cycles. This figure represents the CA10 as a function of the NbC cycle. We thus find an expected result, namely that more pre-ignition is triggered early in the cycle (low CA10) and more likely to be violent (large circle size).

• le nombre de variables utilisées ;
• la méthode d'identification des directions principales ;
• la méthode et le type de modélisation des combustions normales ;
• la méthode de calcul de la distance d'un point à la surface modélisée.

However, the method according to the invention makes it possible to better classify these different pre-ignitions because the distance to normality is not only a function of the CA10 but of several combined variables. In other words, the processing process makes it possible to associate cycles of similar CA10 with circles of different sizes, ie different intensities. The two examples from the bottom of the figure 8 illustrate this phenomenon with cycles 650 and 671 selected at iso CA10 (about 374 ° V). Here it is important to emphasize that the distances are different although the CA10 are equivalent.
Remarks: the process has several degrees of freedom:

• the number of variables used;
• the method of identifying the main directions;
• the method and type of modeling of normal combustions;
• the method of calculating the distance from a point to the modeled surface.

Another advantage of the methodology: as can be seen on the upper part of the figure 8 the late combustions also emerge because of their abnormal distances from normality. This methodology can therefore also be used to characterize late combustions and misfires. Late combustions are combustions that have been properly initiated at the candle but are developing slowly and thus leading to a loss of yield. We call "combustion misfires", combustions that have not been initiated at all to the candle (following a lack of wealth for example). In terms of cylinder pressure, it is observed that a simple compression / relaxation (or in the best case just a very low combustion).

These two types of combustion do not represent a danger unlike pre-ignition but there is still an interest in detecting them because they are synonymous with poor performance or significant emissions due to combustion defects.

4- Control of abnormal combustion

Finally, the course of the abnormal combustion detected is checked as a function of the severity of the abnormal character.

By means of the position relative to the surface, the engine computer can detect the beginning of an abnormal combustion of the "pre-ignition" type in the combustion chamber. And thanks to the distance from the surface, the engine computer can detect the severity of this abnormal combustion.

In the event of abnormal combustion, and if the severity is proven, this calculator then initiates the actions necessary to control this combustion in order to avoid the continuation of such combustion.

By controlling the abnormal combustion, it is understood not only the possibility of controlling the course of this combustion to avoid sudden increases in destructive pressures but also to completely stop such combustion, such as by choking.

Preferably, this control of the combustion is achieved by a fuel injection at a crankshaft angle determined by the injectors 18. More specifically, the computer controls the valves 20 so that the cylinder injector concerned allows for introducing into the combustion chamber a quantity of fuel in liquid form. The amount of fuel reinjected depends on the constitution of the engine and can range from 10% to 200% of the amount of fuel initially introduced into the combustion chamber. As a result, the reinjected fuel serves to thwart the flame that begins to deploy during abnormal combustion. This reinjection allows either to blow this flame, or to stifle this flame by increasing the richness of the fuel mixture. In addition, the fuel injected in liquid form uses the heat present around this flame to vaporize and the temperature conditions around the flame will drop by retarding the combustion of the fuel mixture and especially its auto-ignition.

After this fuel injection, the pressure in the cylinder increases but less suddenly. This pressure then decreases to a level compatible with the pressure level of a conventional combustion.
By this mechanism, any development of abnormal combustion with a high rate of combustion and high pressures is prohibited. Of course, the implementation of the means for controlling the abnormal combustion is done at each cycle during which such a combustion is detected by the computer.

The actions of the process as described above can be combined with other slower actions, such as butterfly closure, to prevent the pressure conditions of the combustion chamber from being conducive to abnormal combustion in cycles which follow. The choice of action depends on the severity of the abnormal nature of the combustion.

Claims (6)

1. A method for controlling the combustion of a spark-ignition internal-combustion engine, wherein at least one signal representative of a state of the combustion is recorded by means of at least one detector arranged on the engine, characterized in that it comprises:

   - selecting indicators of the combustion that can be deduced from said signal and defining a multidimensional space each dimension of which corresponds to one of said indicators, and wherein any combustion can be represented by a point,

   - defining in said space a closed surface so as to surround points corresponding to normal combustions and not to surround points corresponding to abnormal combustions due to pre-ignition, defining said closed surface by carrying out the following steps:

   • selecting an equation defining said surface, said equation comprising at least one parameter,

   • carrying out a set of combustions wherein normal combustions and abnormal combustions are known, and representing said set of combustions in said multidimensional space forming a cluster of points,

   • determining, by means of a principal component analysis, principal directions of said cluster of points, and determining a dispersion of the points in each principal direction,

   • modifying said parameter so that the extension of the surface in each principal direction is equal to the dispersion in this direction;

   then, for each combustion of an engine cycle:

   - representing said combustion of the cycle by a point in said multidimensional space by determining for this combustion said indicators,

   - determining a position of said point with respect to said surface and deducing therefrom an abnormal nature of said combustion,

   - determining a distance between said point and said surface, and deducing therefrom a severity of the abnormal nature, and

   - controlling the progress of the abnormal combustion detected as a function of the severity of the abnormal nature.
2. A method as claimed in claim 1, wherein a multiplying coefficient that is applied to each dispersion prior to modifying the parameter is defined.
3. A method as claimed in claim 2, wherein said multiplying coefficient is selected between 2.4 and 2.6, preferably equal to 2.5.
4. A method as claimed in any one of the previous claims, wherein said surface is updated from a point obtained from a new combustion.
5. A method as claimed in any one of the previous claims, wherein a surface of quadric type is selected.
6. A method as claimed in any one of the previous claims, wherein said indicators are normalized.

Images