EP4281662A1 - Procede de securisation d'une fonction d'apprentissage d'un modele d'actionneurs de moteur thermique - Google Patents
Procede de securisation d'une fonction d'apprentissage d'un modele d'actionneurs de moteur thermiqueInfo
- Publication number
- EP4281662A1 EP4281662A1 EP21827204.5A EP21827204A EP4281662A1 EP 4281662 A1 EP4281662 A1 EP 4281662A1 EP 21827204 A EP21827204 A EP 21827204A EP 4281662 A1 EP4281662 A1 EP 4281662A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- value
- model
- term
- learning
- learning function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000006870 function Effects 0.000 title claims abstract description 100
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000002485 combustion reaction Methods 0.000 title abstract 2
- 238000005259 measurement Methods 0.000 claims abstract description 14
- 230000001419 dependent effect Effects 0.000 claims abstract description 5
- 238000012937 correction Methods 0.000 claims description 41
- 239000011159 matrix material Substances 0.000 claims description 24
- 238000004364 calculation method Methods 0.000 claims description 13
- 230000003044 adaptive effect Effects 0.000 claims description 12
- 230000033228 biological regulation Effects 0.000 claims description 12
- 230000015654 memory Effects 0.000 claims description 12
- 238000003745 diagnosis Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 10
- 238000005457 optimization Methods 0.000 description 6
- 230000000670 limiting effect Effects 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000002347 injection Methods 0.000 description 3
- 239000007924 injection Substances 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 230000003247 decreasing effect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 239000000446 fuel Substances 0.000 description 2
- 239000007789 gas Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000002829 reductive effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 238000009530 blood pressure measurement Methods 0.000 description 1
- 230000001143 conditioned effect Effects 0.000 description 1
- 238000007435 diagnostic evaluation Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/24—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means
- F02D41/2406—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means using essentially read only memories
- F02D41/2425—Particular ways of programming the data
- F02D41/2429—Methods of calibrating or learning
- F02D41/2441—Methods of calibrating or learning characterised by the learning conditions
- F02D41/2448—Prohibition of learning
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/24—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means
- F02D41/2406—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means using essentially read only memories
- F02D41/2425—Particular ways of programming the data
- F02D41/2429—Methods of calibrating or learning
- F02D41/2451—Methods of calibrating or learning characterised by what is learned or calibrated
- F02D41/2454—Learning of the air-fuel ratio control
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/24—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means
- F02D41/2406—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means using essentially read only memories
- F02D41/2425—Particular ways of programming the data
- F02D41/2429—Methods of calibrating or learning
- F02D41/2451—Methods of calibrating or learning characterised by what is learned or calibrated
- F02D41/2464—Characteristics of actuators
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/24—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means
- F02D41/2406—Electrical control of supply of combustible mixture or its constituents characterised by the use of digital means using essentially read only memories
- F02D41/2425—Particular ways of programming the data
- F02D41/2429—Methods of calibrating or learning
- F02D41/2477—Methods of calibrating or learning characterised by the method used for learning
- F02D41/2483—Methods of calibrating or learning characterised by the method used for learning restricting learned values
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
- F02D2041/1433—Introducing closed-loop corrections characterised by the control or regulation method using a model or simulation of the system
Definitions
- TITLE METHOD FOR SECURING A LEARNING FUNCTION OF A MODEL OF THERMAL ENGINE ACTUATORS
- the field of the invention relates to a method for securing a learning function of an actuator control model of a motor vehicle heat engine.
- the actuators of thermal vehicle engines have a behavior that may differ from the behavior models integrated in the engine control unit, due to manufacturing dispersions, wear, fouling and the quality of representation of the actuators of the motor.
- this shift can lead to drifts in the richness of the mixture. This situation causes overconsumption and/or an increase in polluting emissions. This can also have an impact on driving pleasure. It is necessary throughout the life of the vehicle to correct these richness drifts.
- this correction is carried out by engine control models, performing the richness regulation function, which permanently corrects the injector control time based on the richness measurement provided by the richness sensor present at the fuel. 'exhaust.
- the engine control unit executes in parallel an engine control learning function in charge of storing the correction necessary for engine operating fields. More precisely, on a stabilized point of the engine operating field, as soon as the learning conditions are met, the learning function stores for this point the correction value determined by the engine control model in charge of the richness regulation .
- the learning process can be carried out by means of a neural network of the type self-adaptive.
- the control unit communicates the value being stored, which has the effect of causing the richness regulation correction value to converge.
- the work of regulation is minimized and the dynamics of correction are improved.
- the learning function plays a major role when the richness regulation model cannot correct the richness, especially when the richness sensor is not available, and significantly improves engine control during load transients by minimizing work of the richness regulation function.
- Document FR3057031A1 is known describing a richness correction method filed by the applicant.
- the process described is a least squares minimization type optimization technique based on the analysis of the errors present on the system in order to determine the terms of the correction function and the purpose of which is to correct the errors by improving the models .
- This process describes tests on the quality of the correction terms being calculated to guide the convergence of the optimization loop. These tests are included in the loop for calculating the correction terms and apply to the corrections during convergence.
- the document FR2979390A1 is also known describing an optimization technique based on a Kalman filter.
- An object of the invention is to secure the learning models in order to limit the impact of a punctually shifted value due to an atypical event.
- the invention relates to a method for securing a learning function of an actuator control model of a heat engine for a motor vehicle, the learning function being based on a model consisting of terms of controls stored in the memory of a heat engine control unit and being adapted to operate a learning cycle consisting of control term update loops.
- the method comprises the following steps:
- said security parameter is a weighting coefficient delivered by a predetermined matrix having as input parameter said deviation and the confidence index, and during the step of updating the term of the model, the part of the second current value is calculated according to the value of said weighting coefficient.
- the method further comprises, prior to the cycle of update loops, a step of initializing the confidence index of the term of the model at an initial level of confidence.
- the confidence index is greater than or equal to the initial confidence level and: If, moreover, the difference is greater than or equal to a first threshold and less than or equal to a second threshold, the share of the second value is strictly lower than the share of the first stored value,
- the part of the second value is zero so as to retain all of the part of the first stored value.
- the part of the second value is zero so as to keep in all the part of the first memorized value.
- the share of the second value is at least equal to or greater than the share of the first stored value.
- the first minimum limit is equal to the second minimum limit, or greater than the second minimum limit.
- the learning function further comprises an internal configuration parameter representative of a confidence level of the stored value of the term of the model, and in that said internal parameter is configured according to the value of the security parameter.
- the method further comprises, at each update loop, a step of updating the value of the confidence index as a function of said difference and when said difference is greater than a third threshold, the index of confidence is decremented by a predetermined value and when said deviation is less than said third threshold the confidence index is incremented by said predetermined value.
- control module is a richness regulation function provided for calculating adaptive correction terms and the learning function is adapted to memorize according to a matrix of speed and engine load variables said terms d correction adaptive.
- the invention provides a control unit for a heat engine of a motor vehicle comprising a function for learning an actuator control module of said heat engine, the learning function being based on a model consisting of control terms stored in the memory of the control unit and being adapted to operate a cycle of updates of the control terms.
- the control unit is configured to implement the method for securing the learning function according to any one of the preceding embodiments.
- the invention also relates to a motor vehicle comprising a heat engine controlled by such a control unit.
- the invention relates to a heat engine control unit implementing a learning function for an actuator control module of said heat engine, the learning function being based on a model consisting of recorded control terms in memory of the control unit and being adapted to operate a cycle of updates of the control terms, further comprising: means for calculating a difference between a first value stored by the learning function for said term of the model and a second current value of said term of the model estimated from instantaneous measurements, means for determining a confidence index attributed to the first stored value of said term of the model, means for determining the value of a parameter of security depending on said deviation and on the confidence index, a means for updating the term of the model by a third value consisting of the sum of the shares of the first value and of the second value eur, in which one of said parts or each of said parts is dependent on said security parameter.
- the method and device for securing a learning function makes it possible to establish a diagnosis concerning the reliability of the stored values of a motor actuator control model.
- the updating of a learning function is generally subject to a native convergence mechanism whose purpose is to gradually update the values stored by only a part of the new measured values.
- the diagnosis establishes that a value of a term of the model is not reliable
- the securing method makes it possible to accelerate the convergence towards the new measured values. This acceleration of convergence is also subject to additional checks to prevent an update on erroneous values from time to time.
- the invention finds a preferential application within the framework of the richness control model and therefore actively participates in a rapid and robust control of the richness error throughout the life of the vehicle.
- the invention helps to improve the performance of the learning strategies of the engine actuator control models and therefore contributes to improving its performance.
- FIG.1 represents a device for securing a learning function of a thermal engine actuator control model according to the invention
- FIG.2 represents a preferred embodiment of the algorithm of the securing method according to the invention
- FIG.3 represents an example of a matrix for calculating the security parameter operating the weighting of an adaptive richness correction value for the preferred embodiment of the security method according to the invention
- FIG.4 represents a second variant of the device for securing a learning function according to the invention.
- the invention finds an application for thermal engine actuator control models and in particular for securing a learning function for such control models.
- the invention provides an additional function for securing a native learning function making it possible to establish a security diagnosis relating to the stored adaptives and the current adaptives before the recording of the new value in the learning matrix.
- the additional security function cooperates with the learning function so as to modulate the model noise of the terms of the model according to the result of the diagnosis carried out by the security function.
- Model noise refers to any configuration parameter of the model representative of a level of confidence in the model. According to the result, the securitization function accelerates the process of native convergence towards the value of the new adaptives by decreasing in this variant the confidence level of the model.
- the term actuator control model refers to any motor actuator control function.
- the learning function implements the function of memorizing the terms calculated by the control model, the function of updating the terms of the model and the function of restoring said terms.
- the term stored value of the learning function refers to the value of the control model which is stored in the memory of the engine control unit.
- the term current value of the learning function designates the value estimated from instantaneous measurements of physical quantities from the engine sensor, for example the richness sensor at the exhaust.
- the embodiments of the invention will be described more precisely by way of non-limiting example for the case of a learning function of a wealth regulation model.
- the control model is a richness regulation function in charge of calculating the correction terms of the engine actuators making it possible to regulate the richness.
- a control term designates the richness correction adaptives and the learning function is adapted to memorize said richness correction adaptives, following for example a matrix of engine speed and load variables .
- the current value is estimated from engine manifold pressure measurement and an oxygen content of an engine exhaust gas.
- the adaptive corrections adjust the actuators of the intake branch and/or the injection branch of the heat engine making it possible to control the richness, either the gas admission valve opening control and/or the duration control fuel injection.
- FIG. 1 represents the first preferred embodiment of the functional modules of a thermal engine control unit allowing the realization of the securing method according to the invention.
- the control unit is provided with an integrated circuit computer and electronic memories, the computer and the memories being configured to execute the securing method. But this is not mandatory. Indeed, the computer could be external to the engine control unit, while being coupled to the latter. In this last case, it can itself be arranged in the form of a dedicated computer comprising a possible dedicated program, for example. Consequently, the control unit, according to the invention, can be produced in the form of software (or computer (or even “software”)) modules, or else of electronic circuits (or “hardware”), or even of a combination of electronic circuits and software modules.
- control unit comprises a learning function 11 of the richness regulation actuator control model and a module 1 for securing the value of the term to be learned by the learning function 11 .
- the learning function 11 is configured to memorize an adaptive correction under pre-established learning conditions, in particular on an operating point stabilized under load and engine speed or an operating zone stabilized under engine speed and load.
- the learning process continuously executes during the operation of the engine loops of updates of the terms of the learning matrix when the learning conditions are established.
- the learning function proceeds iteratively, or recursively, to calculations of values of the terms of the model resulting from instantaneous measurements of physical quantities recorded by one or more sensors of the engine for an evaluation in comparison with a value stored from the previous update loop.
- the new value memorized by an update loop includes all or part of the current value, or fully retains the value memorized during the previous loop.
- the values updated by the learning function can be obtained by using a least squares minimization type optimization technique, as illustrated by document FR3057031A1, or a Kalman filter as illustrated by application FR2979390A1.
- a least squares minimization type optimization technique as illustrated by document FR3057031A1
- a Kalman filter as illustrated by application FR2979390A1.
- the security module 1 comprises a calculation means 2 capable of receiving as input a first value stored 10 by the learning function 11 for a term of the model and a second current value 12 of said term estimated from instantaneous measurements. Calculation means 2 is configured to calculate the difference between these two values.
- the security module 1 can be used for the evaluation of a point, several points, a zone or each zone of motor operation defined by the learning matrix. The calculation of the difference is capable of being operated on each update loop of the learning function. The value of deviation 6 is delivered at the output of module 2.
- the security module 1 further comprises means 3 for configuring and determining a confidence index 7 assigned to the first stored value 10 of said term of the model in the learning matrix.
- the calculation means 3 is able to configure an initial predetermined confidence level and to modify the value of the confidence index according to the evaluation of the difference 6, delivered by the means 2, in comparison with a predetermined threshold .
- the confidence index 7 is a parameter used to develop the security diagnosis relating to the stored values of the model.
- the confidence index 7 acts as an evaluation scale, for example as a percentage between 0% and 100%. But it is not mandatory. The purpose of this confidence index 7 is to quantify the degree of confidence in the reliability of the stored value and to have an indication of what happened in the previous update loops.
- the value of the initial predetermined level configurable at the start of the learning cycle for the first memorized value of a term of the learning matrix is in this example the average intermediate value 50%.
- the configuration and determination means 3 is capable of incrementing and decrementing the value of the index by predetermined stages of constant value, comprised between 1% and 10% for example. In the context of the invention, the step determines the number of update loops before triggering a questioning of the stored value.
- the security module 1 further comprises a means 4 for determining the value of a security parameter 8 taking as input parameter the value of the difference 6 and the value of the confidence index 7 making it possible to establish the security diagnosis in the stored values 10 and the current values 12 of the control model during an update loop.
- the determining means 4 delivers in this embodiment a weighting coefficient value 8 making it possible to adjust the part of the current value 12 of the term of the model calculated from the instantaneous measurements for the update of the learning model 13 More specifically, in function of the confidence index 7 and the deviation 6, the weighting coefficient makes it possible to memorize during the update only part of the current value 12, to ignore the current value 12 in the event of very large difference if the confidence index is high, a large part if the confidence index of the stored value 10 is low, that is to say has reached a predetermined minimum limit, and the complete value if the difference is zero.
- a method of calculating the security parameter 8 will be described by way of example in the remainder of the description.
- the calculation of the security parameter 8, in the form of a weighting parameter can take the form of a matrix, delivering predetermined values, with two input variables which are the deviation 6 and the confidence index 7.
- the values of the weighting parameter 8 are for example between 0 and 1.
- the security module 1 further comprises a determination means 5 whose function is to update a term of the model by a third value to be stored consisting of the sum of parts of the first value 10 and of the second value 12, during which the part of the second value 12 is calculated from the value of the weighting parameter 8 attributed to the current value of the term of the model.
- the determining means 5 delivers the new value 14 corresponding to the current value of the richness correction to which the weighting parameter 8 is assigned for its update by the learning function 11 .
- the new value memorized in the learning matrix 13 by the learning function consists of the sum of the parts of the first value 10 and of the second value 12 which are calculated according to the weighting coefficient 8. The new memorized value is obtained by calculating the barycenter.
- the part of the second value 12 corresponds to the product of this second value 12 with the weighting coefficient 8 and the part of the first value 10 corresponds to the product of this first value 10 with a weighting term equal to one minus the value of the weighting parameter 8.
- the updated values to be stored are recorded in the memory of the engine control unit, in a storage means with volatile memory and non-volatile memory and programmable so as to be reused later, in particular when the engine is stopped. motor and restart.
- the security parameter 8 is a weighting parameter determining the share of the current value 12 of the term of the model to be stored in the learning matrix.
- a learning cycle in accordance with the invention provides for a point, each operating point, a zone or each motor operating zone, an initialization phase 200 of the learning model.
- the learning matrix does not contain any control term values for the operating point or operating area subject to control.
- the method performs a check on the learning conditions, specific to the function, during which it is checked in particular whether the engine is operating at stabilized speed and load.
- This stabilization condition is not limiting. Other conditions may determine the triggering of a calculation of a model richness correction term.
- the method stores, at a step 21 , the entire current value of the control term resulting from the instantaneous measurements, the weighting coefficient 8 is therefore configured at 1 , and, at a step 22 configures the value of the confidence index attributed to said term to the initialization value, for example 50%.
- the initial confidence level is representative of an average level, that is to say below which the security diagnostic doubts the value and above which the security diagnostic considers that the value is reliable.
- the confidence index will be used in combination with the value of the difference during the learning cycle to determine from which update loop we can accelerate the convergence of the stored value towards the current value.
- the securing method enters a phase 201 of updating the learning matrix for the control term or each control term for which a first value is recorded.
- This update phase is recursive and the triggering of which is conditioned by the detection of optimal learning conditions, by operation at stabilized speed and load.
- the one-step learning function 24 calculates a new current value of the model richness correction term from the instantaneous measurements.
- the securing method then comprises a step 25 of calculating the difference between the stored value of the correction term by the learning function 11, resulting from the previous loop, and the current value of said term estimated from instantaneous measurements by the learning function 11 .
- the method includes a step 26, during which the security function 1 determines the value of the confidence index assigned to the stored value of said term of the model in the learning matrix.
- the security function reads the value 50% in this example.
- the confidence index will depend on the learning revolution.
- the method includes a step 27 of determining the value of a weighting coefficient to be assigned to the current value for updating the learning matrix.
- the weighting coefficient is the result of a security diagnostic evaluation depending on said deviation calculated during this update loop and the confidence index associated with the stored value of the model term.
- the securing function defines the value of the weighting coefficient which will be multiplied with the current value of the richness correction term of the model.
- the determination of the weighting coefficient is carried out in this case of example by a diagnostic matrix having as input parameter said difference and the confidence index.
- the first column represents values of the confidence index, graduated between 0% and 100%, of the memorized value
- the first line represents the value of the difference between the memorized value and the current value of the term of the model, between 0 and 0.1.
- the security diagnosis operates according to the following strategy to calculate the weighting coefficient.
- the weighting coefficient is equal to 1. Not showing any difference between the learning model and the real measurements, the learning strategy saves the full current value. If the confidence index is greater than or equal to the initial confidence level, in this example 50% and:
- the weighting coefficient is graduated decreasing, between the values 0.3 and 0.05 for example, in proportion to the difference so that the share of the current value of the correction term is strictly lower than the share of the value stored in the learning matrix .
- the new current richness correction value is partially, or only slightly, taken into account for the update, due in particular to the fact that the stored value is, at this stage of the learning cycle, judged to still be reliable.
- the objective of the security diagnosis is that the update takes place according to a first rate of convergence graduated according to the value of the difference. The more the gap increases and as long as the confidence index is reliable, the more the speed of convergence is reduced.
- the weighting coefficient is equal to 0 so as to fully retain the part of the first stored value.
- the new current value of the wealth correction term is therefore not taken into account.
- the convergence of the correction term towards the current value is stopped as long as the confidence index is greater than the level initial 50%.
- the weighting coefficient is equal to 0. This implies that the part of the current value of the wealth correction term is zero and is not taken into account.
- the securing diagnostic acts in such a way as to retain all of the part of the first value stored for this zone of uncertainty based on the value of the index.
- the diagnostic strategy consists of temporarily inhibiting the modification of the stored value during a certain number of update loops. At this stage of the learning cycle, there is a phase during which the security diagnostic doubts the stored value and the current value.
- the security function is configured to update the stored value and for this purpose triggers an accelerated process of convergence towards the current measured values. More precisely, with the aim of accelerating convergence, if the confidence index is equal to the minimum limit Lim, here 0%, the weighting coefficient is configured at the value 0.5. It is envisaged that other higher values between 0.5 and 1 can be configured so that the part of the current value of the correction term is at least equal to or greater than the part of the stored value in order to accelerate convergence to the current value.
- the value of the weighting coefficient is specifically provided to accelerate the convergence cycle with respect to the native learning mechanism implemented by the learning function 11 . For this purpose, in this situation the weighting coefficient has a value controlling a speed of convergence towards the current value which is greater than that of the native convergence mechanism of the learning function 11 .
- the security function 1 calculates the new value to be stored in the learning matrix, during which the current value of the richness correction term is weighted by the weighting coefficient resulting from the diagnosis of the previous step.
- the securing method comprises a step 29 of updating the value of the confidence index according to the difference calculated in step 25 between the measured value and the current value of an update loop. day.
- a predetermined threshold for example S2 equal to 0.06
- the confidence index is decremented by the predetermined value of the plateau.
- the security function 1 detects in this case a significant inconsistency between the stored value and the current value of the richness correction term. The function therefore decreases the confidence level in the stored value.
- the threshold value is different, for example greater than S2, or between S1 and S2, or even equal to S1.
- the securing method passes to the step 23 of monitoring an update loop.
- Learning consists of executing a cycle of update loops, during which each loop is triggered when the learning conditions are met to calculate, in this example, a richness correction value.
- FIG. 4 now describes the second embodiment of the securing function 11 .
- the security function 1 uses the same functional blocks 2, 3 and 4 as those implemented for the first embodiment.
- the references are kept identical.
- This embodiment nevertheless differs in that the security function 1 does not execute the function 5 of calculating the value of the richness correction term to be stored which is described in the first embodiment.
- the security parameter 8 calculated by the module 4 is used to configure an internal parameter of the learning function 11, in which the internal parameter is representative of a level of confidence in the stored value of the term of the model.
- a calculation module 51 of the learning function 11 uses this internal parameter to define a model noise associated with the stored terms of the model and to drive natively the process of convergence towards the current richness correction value.
- this internal parameter depends on the modification history of the model during a learning cycle.
- the learning function 11 is adapted to modify said internal parameter of the learning function 11 according to the value of the security parameter 8.
- This internal parameter is for example used by functions of Kalman filter type learning.
- the module 4 makes it possible to establish the diagnosis of securing the stored values 10 and the current values 12 according to the difference 6 between the values 10 and 12 and the value of the confidence index 7 calculated by module 3.
- the means of determination 4 increases or decreases the confidence level of the internal parameter of the learning function, which has the effect, during the update process, of slowing down or accelerating the convergence towards the current value calculated by the learning function 11.
- the modification of the internal parameter during an update loop "n” determines the part of the stored value of the model for the calculation of the new value to be saved 141 of the update loop next “n+1”.
- the securing method does not directly modify the value of the stored correction term, but retains the native management mode of the convergence of the control terms of the learning function 11 and the value of this internal parameter representative of the model noise to control the speed of convergence towards the current values.
- the invention makes it possible to accelerate the convergence towards the new current value.
- the securing method also envisages a function for selecting the mode of convergence of a stored value towards a current value of the model.
- the selection is configurable and makes it possible to select for the learning cycle either the implementation of the first embodiment or the implementation of the second embodiment.
- the security parameter 8 directly affects the calculation of the share of the current value constituting the new stored value, while in the second embodiment, the security parameter 8 directly affects the native internal parameter used by the learning function to determine the part of the stored value constituting the new value to be recorded.
- the securing method differs from the sequence of FIG. 2 in that step 27 calculates a securing parameter intended to modify the value of the internal parameter of the learning function representative of the level of confidence of the model and in that step 28 consists in modifying this value of the internal parameter to slow down or accelerate the process of convergence towards the current value, by increasing the share or reducing the share of the stored value in the calculation of the new value to store.
- step 27 calculates a securing parameter intended to modify the value of the internal parameter of the learning function representative of the level of confidence of the model
- step 28 consists in modifying this value of the internal parameter to slow down or accelerate the process of convergence towards the current value, by increasing the share or reducing the share of the stored value in the calculation of the new value to store.
- the part of the stored value depends on the value of the security parameter calculated during the previous loop.
- the invention applies to any function for learning thermal engine actuator models, in particular for motor vehicle engines.
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Combined Controls Of Internal Combustion Engines (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
FR2100573A FR3118995B1 (fr) | 2021-01-21 | 2021-01-21 | Procede de securisation d’une fonction d’apprentissage d’un modele d’actionneurs de moteur thermique |
PCT/FR2021/052135 WO2022157429A1 (fr) | 2021-01-21 | 2021-11-30 | Procede de securisation d'une fonction d'apprentissage d'un modele d'actionneurs de moteur thermique |
Publications (1)
Publication Number | Publication Date |
---|---|
EP4281662A1 true EP4281662A1 (fr) | 2023-11-29 |
Family
ID=75278181
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP21827204.5A Pending EP4281662A1 (fr) | 2021-01-21 | 2021-11-30 | Procede de securisation d'une fonction d'apprentissage d'un modele d'actionneurs de moteur thermique |
Country Status (4)
Country | Link |
---|---|
EP (1) | EP4281662A1 (fr) |
CN (1) | CN116745513A (fr) |
FR (1) | FR3118995B1 (fr) |
WO (1) | WO2022157429A1 (fr) |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB1251223A (fr) | 1968-08-09 | 1971-10-27 | ||
FR2979390B1 (fr) | 2011-08-23 | 2013-08-23 | Valeo Sys Controle Moteur Sas | Procede et systeme de commande du fonctionnement d'un moteur de vehicule |
FR3033840B1 (fr) * | 2015-03-16 | 2017-03-31 | Peugeot Citroen Automobiles Sa | Procede de gestion de l’alimentation d’un moteur a combustion interne |
FR3055667B1 (fr) * | 2016-09-06 | 2020-08-21 | Peugeot Citroen Automobiles Sa | Procede de gestion de l’alimentation d’un moteur thermique, et calculateur mettant en œuvre ledit procede |
FR3057031A1 (fr) | 2016-10-03 | 2018-04-06 | Peugeot Citroen Automobiles Sa | Procede de recalage des modeles de comportement d’actionneurs de lignes d’admission et d’injection de moteur a combustion interne |
-
2021
- 2021-01-21 FR FR2100573A patent/FR3118995B1/fr active Active
- 2021-11-30 WO PCT/FR2021/052135 patent/WO2022157429A1/fr active Application Filing
- 2021-11-30 EP EP21827204.5A patent/EP4281662A1/fr active Pending
- 2021-11-30 CN CN202180091362.4A patent/CN116745513A/zh active Pending
Also Published As
Publication number | Publication date |
---|---|
FR3118995A1 (fr) | 2022-07-22 |
CN116745513A (zh) | 2023-09-12 |
FR3118995B1 (fr) | 2023-04-28 |
WO2022157429A1 (fr) | 2022-07-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP4424417B2 (ja) | 燃料中のアルコール成分量の推定装置 | |
FR2901847A1 (fr) | Dispositif et procede de commande d'injection de carburant | |
FR2721351A1 (fr) | Système de commande de dosage de carburant d'un moteur combustion interne. | |
FR2945077A1 (fr) | Dispositif de commande d'injection de carburant | |
EP1058781B1 (fr) | Procede et dispositif d'autoadaptation rapide de richesse pour moteur a combustion interne | |
EP1936156B1 (fr) | Procédé de régulation d'un moteur à combustion interne | |
FR2979390A1 (fr) | Procede et systeme de commande du fonctionnement d'un moteur de vehicule | |
EP4281662A1 (fr) | Procede de securisation d'une fonction d'apprentissage d'un modele d'actionneurs de moteur thermique | |
EP3212920A1 (fr) | Procede de diagnostic du fonctionnement de la purge d'un canister | |
WO2016207556A1 (fr) | Procédé et calculateur pour le contrôle de la pression à l'intérieur d'un réservoir de carburant d'un véhicule automobile | |
WO2022053256A1 (fr) | Diagnostic defaillance jeu aux soupapes ou papillon de tondeuse | |
FR3088965A1 (fr) | Procede de correction de commande d’un moteur thermique | |
WO2022018332A1 (fr) | Procede de correction d'une derive de mesure de richesse | |
FR2892771A1 (fr) | Procede et dispositif de gestion d'un moteur a combustion interne | |
FR3033840A1 (fr) | Procede de gestion de l’alimentation d’un moteur a combustion interne | |
FR3073570A1 (fr) | Procede de correction de richesse moteur | |
WO2022214747A1 (fr) | Procede de validation d'un diagnostic de deviation de la richesse sous critere de conditions similaires | |
EP3006716B1 (fr) | Procédé de commande d'un moteur à combustion interne muni d'un dispositif de traitement de gaz | |
FR3123387A1 (fr) | Procede de surveillance d’adaptatifs dans un controle moteur | |
FR3139162A1 (fr) | Procede de la regulation de la richesse aval en phase de purge d’un catalyseur | |
FR3061746A1 (fr) | Procede de correction d'une duree d'injection de carburant dans un cylindre de moteur thermique de vehicule automobile | |
WO2020245522A1 (fr) | Procédé de correction de richesse de carburant lors d'un démarrage a froid d'un moteur thermique | |
EP4348024A1 (fr) | Procede de limitation d'une correction de parametre effectuee par plusieurs adaptatifs dans un controle moteur | |
FR3122218A1 (fr) | Procede de surveillance des adaptatifs de correction delivres par une fonction d’apprentissage de regulation de la richesse d’un moteur thermique | |
KR20060003541A (ko) | 엔진의 퍼지 농도 보정 제어방법 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
17P | Request for examination filed |
Effective date: 20230605 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
RAP3 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: STELLANTIS AUTO SAS |
|
DAV | Request for validation of the european patent (deleted) | ||
DAX | Request for extension of the european patent (deleted) |