METHOD FOR DETERMINING A COMBUSTION STABILITY INDICATOR IN A CYLINDER OF AN ENGINE

Description

technical field

[0001] The present invention relates to the field of combustion control in a spark-ignition internal combustion engine. In particular, the present invention relates to determining a combustion stability indicator in a spark-ignition internal combustion engine.

[0002] This type of engine comprises at least one cylinder with a combustion chamber defined by the cylinder's inner side wall, the top of the piston which slides within the cylinder, and the cylinder head. Generally, a fuel-air mixture (air and fuel, and possibly exhaust gases) is contained within this combustion chamber and undergoes a compression stage followed by a combustion stage triggered by a spark plug.

[0003] Increasing restrictions on CO2 emissions, coupled with the shrinking diesel vehicle fleet, are driving renewed interest in improving the efficiency of spark-ignition engines. One promising approach is to reduce the fuel-air ratio (fuel / air mixture) to achieve what are commonly known as lean-burn conditions. Another approach is to increase the exhaust gas recirculation (EGR) rate. These conditions increase thermodynamic efficiency and thus reduce fuel consumption. A drawback of high dilutions is increased combustion instability. Combustion stability refers to the variation, cycle by cycle, in the overall engine efficiency.The main undesirable consequences of such instabilities are a decrease in engine efficiency, potentially damaging torque fluctuations, and an increase in pollutant emissions.

[0004] Currently, combustion stability is ensured by proper calibration of the internal combustion engine, incorporating a safety margin. Therefore, a combustion stability estimator is not required. However, the advantages of lean combustion necessitate operating the engine at the limit of combustion stability, which, in this case, makes combustion stability estimation and feedback necessary. Previous technique

[0005] Combustion instability is conventionally quantified by the coefficient of variation (COV), which is the ratio of the standard deviation to the mean, both calculated over sliding windows, of a scalar quantity measured at each combustion and reflecting its quality. Quality is understood as the proportion of fuel actually burned in each cycle.

[0006] In unstable combustion conditions, a fraction of the fuel remains unburned, leading to an undesirable decrease in overall engine efficiency, an increase in pollutant emissions (HC and particulate matter in particular) and mechanical vibrations.

[0007] Several measures of combustion quality have already been proposed: exhaust gas temperature (described in particular in patent application US2007051170), exhaust gas oxygen level (described in particular in patent application KR20090065372), or measurements from an ionization sensor present in the combustion chamber (described in particular in patent application US2004084018).

[0008] However, the decreasing cost of cylinder pressure sensors makes their use in production vehicles a possibility. Mean Indicated Pressure (MIP, proportional to engine torque) can be derived from the signal of this pressure sensor. Directly reflecting engine efficiency, this measurement is an excellent candidate for determining a combustion stability indicator. An example of this application is documented in JP H08 144830 A.

[0009] In steady-state conditions, on an instrumented test bench, the VOC coefficient of variation of the PMI (Power Management Index) is traditionally calculated over long sliding windows (several hundred cycles). However, calculating VOC on production engines under real-world driving conditions (primarily transient conditions) cannot be done in the same way. Indeed, for VOC estimation to be meaningful, the statistical distribution of the combustion quality measurement must remain relatively constant. For example, during a load transient, the PMI (and therefore the torque) varies rapidly, and its distribution, particularly its average value, is altered. Even an internal combustion engine with only ideal combustion will have a significant VOC in all its torque transients (both increases and decreases), whereas a zero value would be desirable. Each variation in PMI leads to an overestimation of combustion instability.

[0010] To address this problem, patent application WO19163507 describes a method that uses the VOC coefficient of variation of the PMI, from which a linear trend estimated over a sliding window is subtracted. This method is heavily noisy due to the imprecision in estimating the two parameters modeling the linear trend. Consequently, the combustion stability indicator remains imprecise.

[0011] US patent application 2016146702 discloses a method implementing a complex indicator that sums three contributions: a term related to the VOC coefficient of variation of the PMI, a term related to the crankshaft angle at maximum cylinder pressure, and a term related to the combustion duration. However, this method does not resolve the problem of VOC overestimation during torque transients in the internal combustion engine.

[0012] US patent application US2019010890 describes a method that uses the VOC coefficient of variation of the PMI and a target function of the engine speed, torque, and temperature to control the engine condition and minimize the difference between the target and the PMI VOC. However, this method does not address the problems of VOC overestimation during internal combustion engine torque transients and uncertainty in VOC estimation if it relies on short sliding windows. Summary of the invention

[0013] The present invention aims to precisely determine a combustion stability indicator, even during transient regimes of the internal combustion engine. To this end, the invention relates to a method for determining a combustion stability indicator that calculates the coefficient of variation (COV) of a ratio of the mean indicated pressure (MIP) divided by the product of the amount of fuel injected and the engine efficiency. This ratio is as invariant as possible to any variation other than a change in combustion quality, thereby improving accuracy, particularly during torque transients of the internal combustion engine. This allows for optimized control of the internal combustion engine at the limits of stability, thereby improving the efficiency of the internal combustion engine.

[0014] The invention relates to a method for determining a combustion stability indicator in at least one cylinder of a spark-ignition internal combustion engine, said at least one cylinder of said internal combustion engine being equipped with a pressure sensor, in which the following steps are implemented: a. For each cycle, a quantity of fuel injected into said at least one cylinder is determined; b. For each cycle, said pressure within said at least one cylinder is measured by means of said pressure sensor, and an average indicated pressure is deduced; c. For each cycle, an efficiency of said internal combustion engine is determined; and d. A ratio is determined, formed by said average indicated pressure divided by a product of said efficiency and said quantity of fuel injected into said at least one cylinder; e. Said stability indicator is characterized based on said ratio determined for each cycle.

[0015] According to one embodiment, the quantity of fuel injected into said at least one cylinder is determined by means of a setpoint for the quantity of fuel injected.

[0016] According to the invention, said efficiency is determined by means of an efficiency map, preferably said efficiency map depends on the speed and torque of the internal combustion engine.

[0017] According to the invention, the stability indicator is characterized by determining the ratio of the standard deviation of said determined ratio to the mean of said determined ratio. In one embodiment, the stability indicator is characterized by determining an estimator of a distribution of said combustion stability indicator, said estimator determining a confidence interval for the combustion stability indicator based on a number N of cycles preceding the current cycle.

[0018] Advantageously, the estimator in question is Bayesian.

[0019] Preferably, the number N of cycles is between 5 and 20.

[0020] Advantageously, said confidence interval contains the value of the ratio of the standard deviation to the mean of the ratio of said determined ratio to the mean of said determined ratio in exactly 98% of cases.

[0021] According to one aspect, said internal combustion engine is an internal combustion engine operating with lean mixtures, and / or said internal combustion engine is equipped with a pre-combustion chamber and / or said internal combustion engine is equipped with an exhaust gas recirculation system.

[0022] Furthermore, the invention relates to a method for controlling a spark-ignition internal combustion engine, at least one cylinder of said internal combustion engine being equipped with a pressure sensor, in which the following steps are implemented: a. A combustion stability indicator is determined by means of the method for determining a combustion stability indicator according to one of the preceding characteristics; and b. The internal combustion engine is controlled according to the combustion stability indicator determined.

[0023] According to one embodiment, said internal combustion engine is controlled by controlling an actuator of said internal combustion engine, based on the comparison of said stability indicator with a predefined threshold.

[0024] According to one implementation, the determination of said estimator is carried out according to one of the preceding characteristics, and said internal combustion engine is controlled by means of the following steps, as a function of a predefined threshold θ of said indicator: i) if said threshold of said indicator θ is less than a lower bound cov min of said confidence interval of said ratio of the standard deviation of said determined ratio to the mean of said determined ratio, then the actuator is decreased; ii) if said threshold of said indicator θ is greater than an upper bound cov max of said confidence interval of said ratio of the standard deviation of said determined ratio to the mean of said determined ratio, then the actuator is increased; and iii) if said threshold of said indicator θ is within said confidence interval (cov min , cov max ), then the actuator is not modified.

[0025] The invention also relates to a control system for a spark-ignition internal combustion engine comprising at least one computer implementing the control method according to one of the preceding characteristics.

[0026] Furthermore, the invention relates to a computer program product downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor, comprising program code instructions for implementing a process according to one of the characteristics, when said program is executed on a computer or calculator.

[0027] Other features and advantages of the process and system according to the invention will become apparent from the following description of non-limiting examples of implementations, with reference to the figures attached and described below. List of figures

[0028] There figure 1illustrates the steps in the process of determining a stability indicator according to a first embodiment of the invention. figure 2 illustrates the steps in the process of determining a stability indicator according to a second embodiment of the invention. figure 3 illustrates the steps in the process of controlling an internal combustion engine according to a first implementation of the invention. figure 4 illustrates the steps in the process of controlling an internal combustion engine according to a second embodiment of the invention. figure 5 represents an example of internal combustion engine efficiency mapping. figure 6 represents the SME and the size PMI ηQ inj depending on the time available, for example, the following: figure 7 illustrates the comparison step for the estimator implemented in the second implementation of the control method according to the invention. Description of the implementation methods

[0029] The present invention relates to determining a combustion stability indicator for a combustion process occurring in a cylinder of a spark-ignition internal combustion engine. The combustion stability indicator reflects the stability of the combustion (cycle-to-cycle variation in the overall engine efficiency). The internal combustion engine includes at least one pressure sensor for at least one cylinder.

[0030] The invention can be implemented for any type of spark-ignition internal combustion engine. However, the invention is particularly well-suited for spark-ignition internal combustion engines with a reduced fuel-air ratio, in order to achieve what are known as lean-mixture conditions. Typically, lean mixture conditions are defined as a mixture where the proportion of air is greater than 14.5 times that of fuel. Furthermore, the invention is particularly well-suited for internal combustion engines equipped with an exhaust gas recirculation (EGR) system. In addition, the invention is particularly well-suited for pre-chamber (or pre-combustion chamber) internal combustion engines.

[0031] According to the invention, the method for determining, in real time, a combustion stability indicator involves the following steps: 1 / Determining the fuel quantity 2 / Measuring the pressure and determining the PMI (Positive Pressure Intake) 3 / Determining the engine efficiency 4 / Determining the ratio PMI ηQ inj 5 / Determination of the stability indicator

[0032] Steps 1 through 5 can be implemented using computer systems, specifically a computer / controller for an internal combustion engine control system. Steps 1 through 3 are performed simultaneously for each cycle; there is no specific order for these steps. These steps will be detailed later in the description.

[0033] There figure 1This illustrates, schematically and without limitation, the steps of the process for determining the stability indicator according to a first embodiment of the invention. The process includes a pressure measurement step (MES), which allows the indicated mean pressure (PMI) to be deduced. In parallel, the efficiency of the internal combustion engine (η) and the quantity of fuel injected (Qinj) are determined. These three parameters are then used to determine a ratio (RAT), which in turn determines the ratio PMI ηQ inj . We can then deduce EST as the coefficient of variation cov to form the combustion stability indicator.

[0034] According to a second embodiment of the invention, the method for determining a combustion stability indicator can determine an estimator of the distribution of the combustion stability indicator. For this embodiment, the method for determining a combustion stability indicator implements the following steps: 1 / Determining the fuel quantity 2 / Measuring the pressure and determining the PMI (Positive Pressure Intake) 3 / Determining the engine efficiency 4 / Determining the ratio PMI ηQ inj 6 / Estimation of the indicator distribution

[0035] Steps 1 through 4 and 6 can be implemented using computer systems, specifically a computer / controller for an internal combustion engine control system. Steps 1 through 3 are performed simultaneously for each cycle; there is no specific order for these steps. These steps will be detailed later in the description.

[0036] There figure 2 This illustrates, schematically and without limitation, the steps of the method for determining the stability indicator according to a second embodiment of the invention. The method includes a pressure measurement step (MES), which allows the indicated mean pressure (PMI) to be deduced. In parallel, the efficiency of the internal combustion engine (η) and the quantity of fuel injected (Qinj) are determined. These three determined parameters are then used to calculate a ratio (RAT) which allows the ratio to be determined. PMI ηQ inj We then determine an estimator EST of the distribution of the stability indicator. This estimator determines a confidence interval idc of the indicator, namely the cov of the ratio PMI ηQ inj .

[0037] According to one embodiment of the invention, the invention also relates to a method for real-time control of a spark-ignition internal combustion engine. For this control method, the internal combustion engine is controlled based on a determined combustion stability indicator.

[0038] Thus, according to a first implementation of the control method, the method for determining the combustion stability indicator can be implemented according to the first embodiment (for example, as illustrated in figure 1 For this implementation, the ordering process may include the following steps: 1 / Determining the fuel quantity 2 / Measuring the pressure and determining the PMI (Positive Pressure Intake) 3 / Determining the engine efficiency 4 / Determining the ratio PMI ηQ inj 5 / Determination of the stability indicator 7 / Control of the internal combustion engine

[0039] Steps 1 through 5 and 7 can be implemented using computer systems, specifically a computer / controller for an internal combustion engine control system. Steps 1 through 3 are performed simultaneously for each cycle; there is no specific order for these steps. These steps will be detailed later in the description.

[0040] There figure 3 This illustrates, schematically and without limitation, the steps of the control process according to a first implementation of the invention. The process includes a pressure measurement step (MES), which allows the indicated mean pressure (PMI) to be deduced. In parallel, the efficiency of the internal combustion engine (η) and the quantity of fuel injected (Qinj) are determined. These three parameters are then used to determine a ratio (RAT), which in turn determines the ratio PMI ηQ inj .From this, we deduce by estimation EST the coefficient of variation cov to form the combustion stability indicator. Then, we control CON the internal combustion engine after a comparison COMP between the combustion stability indicator cov and a predefined threshold of the indicator θ.

[0041] According to a second implementation of the control method, the method for determining the combustion stability indicator can be implemented according to the second embodiment (for example, as illustrated in figure 2 For this implementation, the ordering process may include the following steps: 1 / Determining the fuel quantity 2 / Measuring the pressure and determining the PMI (Positive Pressure Intake) 3 / Determining the engine efficiency 4 / Determining the ratio PMI ηQ inj 6 / Estimation of the indicator distribution 7 / Internal combustion engine control

[0042] Steps 1 through 4 and 6 through 7 can be implemented using computer systems, specifically a computer / controller for an internal combustion engine control system. Steps 1 through 3 are performed simultaneously for each cycle; there is no specific order for these steps. These steps will be detailed later in the description.

[0043] There figure 4 This illustrates, schematically and without limitation, the steps of the control process according to a second embodiment of the invention. The process includes a pressure measurement step (MES), which allows the indicated mean pressure (PMI) to be deduced. In parallel, the efficiency of the internal combustion engine (η) and the quantity of fuel injected (Qinj) are determined. These three parameters are then used to determine a ratio (RAT), which in turn determines the ratio PMI ηQ inj We then determine an estimator EST of the distribution of the stability indicator. This estimator determines a confidence interval idc of the indicator, namely the coefficient of variation cov of the ratio PMI ηQ inj . Then, we control CON the internal combustion engine according to the combustion stability indicator ind, after a comparison COMP of the confidence interval idc and a predefined threshold of the indicator θ. 1 / Determining the quantity of fuel

[0044] During this step, the quantity of fuel injected into at least one cylinder of the internal combustion engine is determined for each cycle. A cycle is defined as the set of four operating phases of an internal combustion engine: intake, compression, expansion, and exhaust, corresponding to two crankshaft revolutions.

[0045] According to one embodiment of the invention, the quantity of fuel injected into the cylinder can be a setpoint for the quantity of fuel injected into the cylinder. This setpoint for the quantity of fuel injected can be determined by the internal combustion engine's computer / controller, based in particular on a setpoint for the torque and engine speed. For example, the setpoint for the quantity of fuel injected may only be known in open loop. In this case, the internal combustion engine controller can impose an injection duration. A model can then be used that transforms this injection duration into a mass of fuel.

[0046] Alternatively, the amount of fuel injected into the cylinder can be measured using at least one sensor. 2 / Pressure measurement and determination of the PMI

[0047] During this step, the pressure within the cylinder is determined for each cycle using the pressure sensor. Then, the mean indicated pressure (MIP) is calculated from the sensor signal by an integral calculation of the measured pressure over each cycle. The mean indicated pressure (MIP) is the average pressure in the combustion chamber of an internal combustion engine during one cycle. It corresponds to the ratio of the work done by the engine fluid (fuel-air mixture) to the engine displacement. 3 / Determining engine efficiency

[0048] During this step, the expected efficiency of the internal combustion engine is determined for each cycle.

[0049] In one embodiment, the expected efficiency can be determined by means of an efficiency map that depends on the operating point of the internal combustion engine; the efficiency can depend, for example, on the engine speed and torque. However, to improve the accuracy of the efficiency determination, it is possible to take into account other parameters, such as the ignition timing.

[0050] There figure 5 This is a non-limiting example of an efficiency mapping (η in %) as a function of the internal combustion engine speed (Ne in rpm) and the maximum permissible pressure (MPP) in bar (10⁻¹ < MPa). The shade of gray within the map represents the efficiency. It is important to remember that the MPP is related to the torque of the internal combustion engine; therefore, this map can be based on both the engine speed and the torque of the internal combustion engine. 4 / Determining the ratio PMI ηQ inj

[0051] In this step, for each cycle, a ratio is determined by dividing the average pressure (determined in step 2) by the product of the efficiency (determined in step 3) and the quantity of fuel injected (determined in step 1). This ratio can be written PMI ηQ inj with PMI the indicated average pressure, η the efficiency, and Q inj the quantity of fuel injected. 5 / Determination of the stability indicator

[0052] In this step, the stability indicator is characterized by directly determining the combustion indicator, as the coefficient of variation of the ratio determined in the previous step.

[0053] In other words, the ratio of the standard deviation divided by the mean of the ratio of the indicated mean pressure divided by the product of the efficiency and the quantity of fuel injected is determined. The mean and standard deviation are determined over sliding windows of N cycles (preceding the current cycle). Advantageously, the number N can be greater than or equal to 5, preferably between 5 and 20, and preferably between 7 and 15, and can be approximately 10. Indeed, at least 5 values ​​allow for a useful time window. Moreover, the limit of 20 measurements allows for the formation of a responsive window and limits the RAM used for the process according to this embodiment of the invention. However, for an application not requiring high responsiveness, the number N can be taken as large as possible. The indicator IND can therefore be written as: IND = COV PMI ηQ inj where COV is the ratio of the standard deviation divided by the mean, PMI is the mean indicated pressure, η is the engine efficiency, and Qinj is the quantity of fuel injected into the cylinder. The coefficient of variation COV of a variable X can be written as: COV X = σ X μ X with σ(X) the standard deviation of the variable X, and µ ( X ) the mean of the variable X.

[0054] Thus, the combustion stability indicator is the coefficient of variation of a quantity that reflects the quality of combustion, independent of engine torque variations. Indeed, the ratio of the average pressure indicated by the amount of fuel injected is proportional to the instantaneous efficiency of the engine, on the scale of each cycle.

[0055] Cycle-to-cycle variations of the ratio PMI Q inj reflect the variations in overall engine efficiency. Although the amount of fuel injected may be biased (injector inaccuracy), as long as this bias, expressed as a multiplicative factor, is constant, the VOC of the quantity PMI Q inj is not flawed. The consistency of this bias, across the calculation windows (N cycles), is reasonable. In order to correct the overestimation of VOCs that would result from a transient linking two operating points with significantly different expected efficiencies (the overall engine efficiency will then necessarily vary, regardless of any combustion instability), the mapped engine efficiency is taken into account by calculating, at each cycle, the ratio PMI η Q inj . The mapped efficiency may indeed give a biased value for the ideal engine efficiency. But as long as this bias, expressed as a multiplicative factor, remains constant across the calculation windows (N cycles), the ratio PMI η carto Q inj remains invariant to changes in the engine's ideal efficiency. This ratio is less sensitive to load transients, and only sensitive to combustion instabilities.

[0056] There figure 6 The upper section illustrates the evolution of the PMI as a function of time T in s for an example of a portion of a driving cycle, simulated without combustion instabilities. Therefore, for this example, the combustion stability indicator should be zero. figure 7 illustrates, for the same example, in the lower part the evolution of the ratio PMI ηQ inj as a function of time T in s. We note that the ratio PMI ηQ inj is more constant than the PMI, particularly for transient regimes. Therefore, the coefficient of variation of the ratio PMI ηQ inj is closer to 0 than the VOC of the PMI. As a result, the combustion stability indicator defined by the invention remains accurate during the transients of the internal combustion engine. 6 / Estimation of the indicator distribution

[0057] In this optional step, the stability indicator can be characterized using an estimator of the distribution of the combustion stability indicator. This estimator determines a confidence interval for the coefficient of variation of the ratio determined in step 5.

[0058] Estimating the confidence interval allows, in particular, for robust control. The estimator determines a confidence interval for the combustion stability indicator based on a number N of cycles preceding the current cycle.

[0059] For this step, we consider the ratios PMI ηQ inj The values ​​determined in step 4 follow a stochastic distribution. Preferably, the stochastic distribution can be approximated by a normal distribution. Indeed, this distribution is particularly suitable because it corresponds to experimental observations. Such a distribution is defined by the following two parameters: the mean and the standard deviation. VOCs are a good indicator of the stability of combustion.

[0060] Such an estimation can be performed quickly with a limited number of measurements, enabling real-time control with minimal memory requirements. Furthermore, this estimator ensures the stability of the process according to the invention, regardless of the number of measurements.

[0061] According to one embodiment of the invention, the estimator can be a Bayesian estimator. This estimator allows the confidence interval for the VOC estimate to be determined reactively with a limited number of measurements.

[0062] According to one embodiment of the invention, the confidence interval contains the true VOC value in exactly 98% of cases. In other words, the true quantile value is less than the confidence interval in 1% of cases and greater than the confidence interval in 1% of cases.

[0063] An example of Bayesian estimation of a confidence interval is given, in a non-limiting manner, in the appendix to this application (at the end of the description).

[0064] According to one embodiment of the invention, the number N can be greater than or equal to 5, preferably between 5 and 20, and preferably between 7 and 15. Indeed, at least 5 values ​​allow for the construction of a representative estimator. Moreover, the limit of 20 measurements allows for the formation of a reactive estimator and limits the RAM used for the method according to the invention. However, for an application not requiring a reactive estimator, the number N can be taken as large as possible. Furthermore, preferably, the number N of cycles can be the same as the number N used in step 2. 7 / Internal combustion engine control

[0065] In this optional step, the internal combustion engine is controlled according to the combustion stability indicator determined in step 5, and / or where appropriate by means of the estimator determined in step 6. The control of the internal combustion engine aims to maintain combustion stability within the cylinder of the internal combustion engine, to improve its efficiency, and possibly to limit pollutant emissions.

[0066] According to one embodiment of the invention, the internal combustion engine can be controlled by controlling the amount of fuel injected into the pre-chamber. Alternatively and / or additionally, the air-fuel mixture in the cylinder can be controlled, for example, by controlling a throttle valve, an intake valve, or a turbocharger. Alternatively and / or additionally, the exhaust gas recirculation (EGR) level can be controlled, for example, by controlling an EGR valve, etc. Alternatively and / or additionally, the ignition timing can be controlled.

[0067] In the following description, an actuator is any component of the internal combustion engine capable of being controlled to modify combustion conditions. This can include the components listed above, such as the injector, throttle valve, intake valve, turbocharger, EGR valve, etc. By convention, a decrease in the actuator's power decreases the stability indicator, and an increase in the actuator's power increases the stability indicator.

[0068] According to one embodiment of the invention, at least one actuator can be controlled based on a comparison of the stability indicator with a predefined threshold. For example, if the stability indicator is below the predefined threshold, then the actuator can be increased, and if the stability indicator is above or equal to the predefined threshold, then the actuator can be decreased.

[0069] In the embodiment implementing the estimator determined in step 6, the internal combustion engine can be controlled by comparing the confidence interval with a predefined threshold of the combustion stability indicator (this threshold is also called the target). In other words, it is necessary to verify whether the combustion stability threshold is within the confidence interval. The threshold of the combustion stability indicator can be obtained through engine bench tests during the engine calibration phase.

[0070] Depending on one aspect, the predefined threshold of the indicator can be between 1 and 7%, and can be 3% for example.

[0071] The ignition timing can be controlled as follows: i) if the indicator threshold is less than the lower bound of the confidence interval enclosing the estimated expectation of cov, then the actuator is decreased; ii) if the indicator threshold is greater than the upper bound of the confidence interval enclosing the estimated expectation of cov, then the actuator is increased; and iii) otherwise (if said indicator threshold is within the confidence interval (cov min , cov max )) the actuator is not modified.

[0072] Thus, the control method according to this embodiment of the invention allows for only relevant corrective actions to be taken: the ignition or fuel mixture control is modified only if it is certain (for example, with 99% confidence, if the centered confidence interval represents 98% of the timing events) that the target VOC is outside the confidence interval. The ignition timing, fuel mixture, or EGR control is therefore more stable, as changes are only imposed approximately once every fifty cycles under steady-state conditions. Reducing control dispersion and engine state variation over short timescales, in turn, makes it possible to approach the combustion stability limit, achieve better efficiency, and limit the number of unstable cycles.

[0073] This comparison and control step can be implemented in a buffer memory containing the indicators from the previous N cycles. Alternatively, it can be implemented using finite impulse response filters. This approach allows only a few variables to be stored from cycle to cycle, unlike the buffer memory which requires more memory.

[0074] In each strategy, an actuator affecting both stability and return is controlled.

[0075] Rather than reacting incorrectly to each cycle, the stochastic control offered by this (non-limiting) implementation is a complete control, but one that is deactivated in the intermediate situation, when no conclusion can be drawn (when the indicator threshold is within the confidence interval). This strategy has several advantages: The actuator's setpoint varies less frequently than in simple integral control (for example, 50 times less frequently if we consider a confidence interval that contains the true quantile value in 98% of cases), if the system is close to the target. θ of the VOC. The entire system will be more easily controllable because unnecessary fluctuations are eliminated. If the system is far from its target θ The controller consistently sees the target as outside the confidence interval. In this situation, the proposed controller is therefore as fast as an integral controller.

[0076] Furthermore, the invention relates to a control system for a spark-ignition internal combustion engine comprising means for implementing the method according to any of the combinations of variants of the control method described above.

[0077] In particular, the control system may include: a pressure sensor, means for processing the pressure sensor signal, means for calculating the combustion quality indicator, a memory for recording N indicators (indicators of the N previous combustion cycles), means for calculating the Bayesian estimator and for performing the comparison, and means for controlling at least one actuator.

[0078] Signal processing means, memory, computing means, and control means can be integrated within an on-board computer of a vehicle.

[0079] The invention also relates to an internal combustion engine equipped with such a control system.

[0080] Furthermore, the invention relates to a computer program product downloadable from a communication network and / or stored on a computer-readable medium (embedded computer) and / or executable by a processor. This program includes program code instructions for implementing the method as described above, when the program is executed on a computer or a computer / controller. Appendix

[0081] Bayesian estimation of the quantile confidence interval.

[0082] For the sake of simplicity, this appendix concerns Bayesian estimation for a normal distribution. I. MOTIVATIONS

[0083] We assume that we have n measurements of the physical phenomenon under study, whose random characteristics cannot be neglected. We also assume that these measurements are independent and normally distributed. We then have n realizations, which we denote X₁, ..., Xₙ. Let cov be the true coefficient of variation of this distribution. Bayes' theorem allows us to express the distribution of the value of cov, with respect to the observed measurements X₁, ..., Xₙ. Naturally, cov can be estimated without uncertainty as n → ∞. However, as n < ∞ (for example, n = 10), cov can no longer be estimated with certainty.

[0084] On the other hand, the Bayesian formalism allows us to define an interval which effectively encompasses cov for 98% of the n possible tuples of n realizations X 1 ,... ,X n from this same distribution. II. ESTIMATION OF THE CONFIDENCE INTERVAL OF THE MEAN

[0085] First, we present the method for determining the average. µ of the distribution (and not its cov), while assuming the variance σ 2< of the known distribution. For this purpose, we introduce Bayes' theorem: π μ i σ i 2 X 1 , … , X n ︷ distribution a posteriori = π X 1 , … , X n μ i σ i 2 ︷ vraisemblance π μ i σ i 2 ︷ distribution a priori π X 1 , … , X n ︸ facteur de normalisation with π the probability and | the conditioning operator. The normalization factor does not depend on µ and behaves like a constant. The prior law reflects the prior knowledge we have about the distribution of µ (regardless of subsequent observations). In order to avoid introducing biased information (which in this context should be considered prejudice), we use the non-informative a priori. π ( µ ) ∝ 1. The realizations (X 1 ,...,X n ) being assumed conditionally (knowing µ ) independent, we can factor the likelihood: π X 1 , … , X n μ ∝ ∏ k = 1 n π X k μ

[0086] We then know the formula for all the factors of the posterior distribution. The calculation leads us to π μ X 1 , … , X n ∝ exp μ − ∑ k = 1 n X k n 2 2 σ / n 2 which is none other than a normal distribution centered on the mean of the X k and with standard deviation σ n It is then easy to obtain the confidence interval of µ , namely: mean ± constant x standard deviation. III. ESTIMATION OF THE CONFIDENCE INTERVAL FOR VOCs

[0087] Experimental data preclude us, on the one hand, from the hypothesis σ = constant and, on the other hand, from the hypothesis σ as a function of µ We are then forced to perform the parallel estimation of the parameter pair ( µ ; σ ) . The methodology has 3 main differences compared to simply estimating the average: The a priori distribution of the couple ( µ ; σ) must be judiciously selected to remain uninformative and avoid introducing unjustified bias during the estimation. The prior π ( µ ) ∝ 1 has a very natural form. On the other hand, in the multidimensional case where one does not only estimate µ but ( µ ; σ ) , Defining the notion of "non-informative" (absence of bias) presents a higher-order difficulty. The concepts of mutual information and information entropy maximization were employed. The posterior distribution has four parameters, requiring a suitable renormalization step to significantly reduce the necessary computing power (avoiding any integral calculus). The posterior distribution of ( µ ; σ) must be processed in order to obtain the posterior distribution of cov. The posterior distribution of cov does not admit a simple analytical form (without integrals) and can be achieved by separating the calculation into two steps: a costly offline calculation of the invariants of the distribution of cov and a lightweight online calculation at each cycle.

Claims

1. Method for determining a combustion stability indicator in at least one cylinder of a controlled-ignition internal combustion engine, said at least one cylinder of said internal combustion engine being equipped with a pressure sensor, wherein the following steps are implemented: a. for each cycle, an amount of fuel injected (Qinj) into said at least one cylinder is determined, b. for each cycle, said pressure within said at least one cylinder is measured (MES) by means of said pressure sensor, and a mean indicated pressure (MIP) is deduced therefrom, c. for each cycle, an efficiency (η) of said internal combustion engine is determined by means of an efficiency map, and d. a ratio (RAT) formed by said mean indicated pressure divided by a product of said efficiency and said amount of fuel injected into said at least one cylinder is determined, e. said stability indicator is characterized (EST) on the basis of said determined ratio, for each cycle, by determining the ratio of the standard deviation of said determined ratio to the mean of said determined ratio.

2. Method for determining a combustion stability indicator according to Claim 1, wherein said amount of fuel injected (Qinj) into said at least one cylinder is determined by means of a setpoint for the amount of fuel injected.

3. Method for determining a combustion stability indicator according to either of the preceding claims, wherein said efficiency map is dependent on the speed and the torque of the internal combustion engine.

4. Method for determining a combustion stability indicator according to any of the preceding claims, wherein said stability indicator is characterized (EST) by determining an estimator of a distribution of said combustion stability indicator, said estimator determining a confidence interval (idc) of the combustion stability indicator on the basis of a number N of cycles preceding the current cycle.

5. Method for determining a combustion stability indicator according to Claim 4, wherein said estimator is Bayesian.

6. Method for determining a combustion stability indicator according to either of Claims 4 and 5, wherein said number N of cycles is between 5 and 20.

7. Method for determining a combustion stability indicator according to any of Claims 4 to 6, wherein said confidence interval (idc) contains the value of the ratio of the standard deviation to the mean of the ratio of the standard deviation of said determined ratio to the mean of said actual determined ratio in exactly 98% of cases.

8. Method for determining a combustion stability indicator according to any of the preceding claims, wherein said internal combustion engine is a lean-burn internal combustion engine, and / or said internal combustion engine is equipped with a combustion pre-chamber and / or said internal combustion engine is equipped with an exhaust gas recirculation system.

9. Method for controlling a controlled-ignition internal combustion engine, at least one cylinder of said internal combustion engine being equipped with a pressure sensor, characterized in that the following steps are implemented: a. a combustion stability indicator is determined by means of the method for determining a combustion stability indicator according to any of the preceding claims; and b. said internal combustion engine is controlled (CON) as a function of said determined combustion stability indicator.

10. Control method according to Claim 9, wherein said internal combustion engine is controlled (CON) by controlling an actuator of said internal combustion engine as a function of a comparison (COMP) of said stability indicator with a predefined threshold (θ).

11. Control method according to Claim 10, wherein the determination of said estimator according to any of Claims 5 to 8 is carried out, and wherein said internal combustion engine is controlled (CON) by means of said following steps, as a function of a predefined threshold θ of said indicator: i) if said threshold of said indicator θ is lower than a lower limit covmin of said confidence interval of said ratio of the standard deviation of said determined ratio to the mean of said determined ratio, then the actuator is decreased in order to decrease the stability indicator; ii) if said threshold of said indicator θ is greater than an upper limit covmax of said confidence interval of said ratio of the standard deviation of said determined ratio to the mean of said determined ratio, then the actuator is increased in order to increase the stability indicator; and iii) if said threshold of said indicator θ is within said confidence interval (covmin, covmax), then the actuator is not modified.

12. System for controlling a controlled-ignition internal combustion engine, said controlled-ignition internal combustion engine comprising a cylinder of said internal combustion engine that is equipped with a pressure sensor, said system comprising at least one calculator implementing the control method according to any of Claims 9 to 11.

13. Computer program product downloadable from a communication network and / or recorded on a medium that is readable by computer and / or executable by a processor, comprising program code instructions for implementing a method according to any of Claims 1 to 11, when said program is executed on a computer or a calculator.