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
  • F02D41/02—Circuit arrangements for generating control signals
  • F02D41/18—Circuit arrangements for generating control signals by measuring intake air flow
  • F02D41/182—Circuit arrangements for generating control signals by measuring intake air flow for the control of a fuel injection device
• • G—PHYSICS
  • G05—CONTROLLING; REGULATING
  • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
  • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
  • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
  • G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
  • G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only
• • 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
  • F02D41/0002—Controlling intake air
  • F02D2041/001—Controlling intake air for engines with variable valve actuation
• • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
  • F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
  • F02D—CONTROLLING COMBUSTION ENGINES
  • F02D2200/00—Input parameters for engine control
  • F02D2200/02—Input parameters for engine control the parameters being related to the engine
  • F02D2200/04—Engine intake system parameters
  • F02D2200/0411—Volumetric efficiency
• • 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
  • F02D41/1405—Neural network 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
• • 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

Definitions

• the invention relates to the control of vehicle engines, in particular gasoline engines.
• a gasoline engine generally uses an injection system calculating a fuel requirement based on measured information regarding manifold pressure, engine speed, and air temperature in the manifold. We then try to model the relative filling efficiency to characterize the amount of air actually entering the engine. This quantity of air is then translated into the quantity of fuel to be injected according to a goal of richness. This calculation can be done as follows:
• Vcyl Unit cubic displacement of the motor (in m3)
• the volumetric filling efficiency ⁇ re mp is characteristic of the engine configuration (the associated volumes, the length and shape of the intake and exhaust manifolds, the materials used and the surface condition of the tubes constituting them). It is also characterized by the laws of lift of the intake and exhaust valves and their phasing in the engine cycle. It depends on the pressure in the manifold, the engine speed and finally the timing of the camshafts (called AAC or VVT) intake and exhaust, especially if the engine is equipped with a variable timing system of these trees.
• AAC or VVT camshafts
• the engine comprises a mapping by position, on or off, the camshaft shifter, function of the pressure in the manifold and the speed.
• the document "Modeling the Volumetric Efficiency of IC Engines: Parametric, Non-Parametric and Neural Techniques" by G. DE NICOLAO proposes a method of controlling an engine in which a volumetric air filling efficiency is determined.
• map corrections are known in motors comprising a continuous intake camshaft shifter.
• a mapping is implemented as a function of the pressure and the speed for a reference position of the shifter and parabolic corrections are made, associated with coefficients mapped as a function of the pressure and the speed.
• An object of the invention is therefore to improve the control of the vehicle engines, and in particular to improve the estimation of the relative filling efficiency, for example in the case of an engine equipped with a double shifter. intake camshaft and exhaust.
• a control method of a vehicle engine in which a volumetric air filling efficiency of the engine is determined, characterized in that a base value of the output is determined, and a correction value by means of an array of artificial neurons, and adding the base value and the correction value.
• the efficiency is determined according to a speed of the engine; the efficiency is determined as a function of a pressure in a manifold of the engine;
• the yield is determined as a function of a difference between a set point of a camshaft of intake cams and a measurement of a position of the shaft;
• the efficiency is determined as a function of a difference between a stall setting of an exhaust camshaft and a measurement of a position of the shaft;
• the efficiency is determined as a function of a difference between an intake valve lift law and a position of the valve
• the efficiency is determined as a function of a difference between an exhaust valve lift law and a position of the valve
• the efficiency is determined as a function of a ratio between a setpoint torque of the motor and a maximum torque;
• the base value is determined as a function of a pressure in the intake manifold and / or a speed of the engine;
• the network comprises a single hidden layer
• the network has the function of activating the tansig function
• a vehicle engine comprising a control member adapted to determine a volumetric air filling efficiency of the engine, the control member comprising a neural network.
• the engine will comprise an intake camshaft variable timing device and / or exhaust camshaft or a variable lift device of intake valves and / or exhaust valves.
• FIG. 1 is a flowchart illustrating the general implementation of the method in the present example
• FIG. 2 comprises two flowcharts illustrating the obtaining of the correction values associated with variable timing of the intake and exhaust camshafts;
• FIG. 3 is a flowchart illustrating the use of the neural network in the method of FIG. 1;
• FIG. 4 is a graph of the activation function implemented in the network of FIG. 3;
• FIG. 6 is a flow chart illustrating the process of choosing the neural network and its calibration.
• Figure 7 is a view similar to Figure 3 illustrating an alternative embodiment.
• the invention relates to a motor vehicle internal combustion engine.
• This engine comprises a computer providing control of the engine and comprising in this case a network of artificial neurons.
• a network of this type is known per se and will not be described in detail here.
• the purpose of the network is to determine a volumetric air filling efficiency of the engine, and more precisely to model this parameter.
• a reference map developed on the nominal reference engine settings, is used to define a reference fill value or base fill value.
• a filling correction value modeled by the neural network is used to correct this filling, in this case by adding to the base value. This gives a characterization of the effective filling of the engine outside the nominal operating points of the engine.
• Box 2 shows the implementation of a map showing a reference value of the filling efficiency of the engine in Z according to a measurement of the engine speed following the X axis on the abscissa and as a function of a pressure measured in an intake manifold of the motor and Y-bearing along the y-axis.
• the engine being in this case provided with intake camshaft and variable-pitch exhaust camshaft, the map provides the filling values on the calibration points chosen by the developer.
• this reference filling value is then corrected to block 4 by adding a corrective term determined in turn in parallel to block 6 by the neural network.
• the neural network calculates this correction according to the following parameters:
• the sum of the base values and the corrective term provides the final value of the fill performance to be modeled.
• FIG. 2 illustrates the detail of obtaining positional offsets of the intake and exhaust camshafts.
• the first diagram indicates that for the intake shaft the difference is made between the calibration set position and the actual position of this shaft as measured.
• the subtraction of these two values in block 8 makes it possible to determine the difference of wedging on this tree.
• the offset value of the shaft is indicated in degrees of crank angle.
• the calibration setpoint value it is determined beforehand in block 10 by mapping from: - the engine speed measured; - the ratio of the target torque (derived from the driver's will via the position of the accelerator, the engine speed and other parameters) and the maximum torque available on the engine (essentially a function of engine speed and engine temperature; the air).
• the offset value for the exhaust shaft is determined in exactly the same way.
• Input to this network includes the engine speed parameters, the pressure in the collector and the differences in timing on the shafts. intake and exhaust relative to the reference timing.
• the neural network implements an algorithm for optimizing the terms of weight (wi) and bias (b) for each neuron as a function of the activation function chosen by the user.
• each neuron is then used in the output layer 20 where a combination of the outputs of each neuron is performed according to the same calculation (but with different weights and a different bias) than for the hidden layer 14.
• the output value of the neural network is then denormalized at block 22 in order to best describe the desired variable which is here the filling efficiency.
• the hidden layer 14 is in this case unique. It has been shown that any continuous piecewise function can be approximated by such an architecture.
• the choice of the number of neurons of the hidden layer is in turn to be determined according to two essential constraints: On the one hand, the precision of the filling modeled by the neural network, on the other hand the number of operations and acceptable calibrations for real-time processing by the engine control computer. It is important to carefully choose the activation function of each neuron to ensure network performance.
• the activation function used in this case is the tansig function. This mathematical function is defined by the following formula:
• FIG. 4 illustrates the shape of the curve of the logig and tansig functions.
• this function In order to be able to be used by the computer but also during the optimization procedure of the neural network, this function must be discretized in a table. When calculating point optimization and network bias or for calculating fill efficiency, the table is then used discretely by linear interpolation.
• the choice of the optimization criterion of the function makes it possible to minimize the error made by replacing a continuous function with a piecewise linear function. It is found that the solution of discretizing the function into a large number of equidistant points is not the best because it is expensive in number of calibrations. It is more advantageous to use a discretization implementing non-equidistant points to reduce the size of the map while maintaining a good accuracy on the output data.
• the optimization criterion retained between the linear function and the discretized function is the optimization of the positioning of the points of support (or breakpoints) by minimizing the second derivative difference between the linear function and the discretized function. It is clear that a linear interpolation between the points is even less correct that the variation of the slope of the function between these points is important.
• FIG. 5 illustrates the result of an optimization of the positioning of the points (the number of which is in this case set at 22) by minimizing the second derivative difference between the continuous function and the interpolated function.
• FIG. 5 illustrates an "s" curve which is that of the activation function used throughout the neural network and in this case in each neuron of the network.
• the linear discretization aims at representing the continuous tansig function according to a one-dimensional table which is easily usable in the engine control software. This discretization has been illustrated on the same graph. It is optimized here to minimize calculation errors when using the weight optimization algorithm (wi) and the bias (b) of the neural network.
• This curve shows on the one hand the interpolation by equidistant breakpoints and on the other hand the interpolation by optimized breakpoints.
• the second figure represents the second derivative, called d 2 , of different functions:
• FIG. 6 illustrates the process of choosing the neural network and its calibration. Indeed, the choice of the number of neurons is important for the computing load in the microprocessor and for the accuracy of the modeling obtained.
• block 30 the generation of the engine database takes place. It implements the scanning of the input parameters of the neural network in the field of the complete engine. This block leads to block 32 which extracts part of the database to create a validation base.
• a pretreatment is carried out on the basis of the data (verification, cleaning, ...) as well as an achievement of learning the neural network according to a convergence criterion (quadratic error + standard deviation + .. .).
• an iteration loop 36 returns if necessary between the blocks 32 and 33 for a modification of the database, the type of learning, the number of neurons, etc.
• next block 36 implements the choice of the network (in particular the number of neurons and the refinement of the activation function).
• This block then leads to block 38 which relates to the implantation of the neural network in the computer and the characterization of the performance in operation.
• block 38 At the output of this block begins a second iteration loop 40 on the revision of the number of neurons as a function of the computing load or further complements on the database. If no iteration is necessary, the block 38 leads to the end block 42.
• the method according to the invention makes it possible to take into account in the frame of the engine each of the continuous camshafts for intake and exhaust cams. in the parametric corrections of the filling.
• the estimation of the relative filling efficiency of the engine equipped with this double camshaft shifter is implemented on the basis of an estimate of the mass of air admitted using the collector pressure sensor, the intake air temperature and engine speed. It makes it possible to ensure optimal control of the injection without any correction of the injection time in a closed loop by using information relating to an exhaust richness probe.
• An implementation variant of the method has been illustrated in FIG.

Landscapes

• Engineering & Computer Science (AREA)
• Artificial Intelligence (AREA)
• Evolutionary Computation (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Software Systems (AREA)
• Mechanical Engineering (AREA)
• Combustion & Propulsion (AREA)
• Health & Medical Sciences (AREA)
• Chemical & Material Sciences (AREA)
• Medical Informatics (AREA)
• General Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Automation & Control Theory (AREA)
• Combined Controls Of Internal Combustion Engines (AREA)
• Output Control And Ontrol Of Special Type Engine (AREA)
• Valve Device For Special Equipments (AREA)
• Feedback Control In General (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=35376955

Family Applications (1)

Country Status (6)

Families Citing this family (10)

Family Cites Families (12)

• 2005
  • 2005-04-28 FR FR0504285A patent/FR2885175B1/fr not_active Expired - Fee Related
• 2006
  • 2006-04-25 WO PCT/FR2006/050387 patent/WO2006114550A1/fr active Application Filing
  • 2006-04-25 US US11/912,937 patent/US7774127B2/en not_active Expired - Fee Related
  • 2006-04-25 EP EP06743844A patent/EP1877658A1/de not_active Withdrawn
  • 2006-04-25 CN CN2006800147090A patent/CN101198783B/zh not_active Expired - Fee Related
  • 2006-04-25 JP JP2008508271A patent/JP2008539360A/ja not_active Ceased

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events