FR2898936A1 - Exhaust gas fuel and air mixture estimating method for e.g. turbocharged oil engine, involves estimating rate of fuel injected in combustion engine using estimator based on information relative to temperature of gas in upstream of turbine - Google Patents

Abstract

The method involves estimating a rate of fuel injected in an internal combustion engine using an estimator (15) with neural networks (17). The rate is estimated based on information relative to a temperature of an exhaust gas in an upstream of a turbine arranged in an exhaust line of the engine. The richness of the fuel and air mixture of the exhaust gas is estimated based on the estimated fuel rate. Independent claims are also included for the following: (1) a device for estimating fuel and air mixture (2) an internal combustion engine comprising a fuel and air mixture estimating device.

Description

The field of the invention is that of the depollution of a motor with

  internal combustion, and more particularly exhaust devices mounted downstream of the exhaust manifold of the engine.

  The invention more specifically relates to a method and a wealth estimation system, and an internal combustion engine equipped with such a system. To avoid or limit the emission of gaseous pollutants from motor vehicles, gas after-treatment systems are generally disposed in the exhaust line of the engines. These systems are designed to reduce carbon monoxide emissions, unburnt hydrocarbons as well as particulate and nitrogen oxides. These systems operate discontinuously or alternately in that they alternate pollutant storage phases and regeneration phases of pollutant gas traps (that is to say conversion of stored pollutants into non-polluting substances). In order to optimize the treatment of all the pollutants, it is necessary to better control these storage and regeneration phases. In particular, it is necessary to estimate over time the trapped masses (ie the particles in the case of the particulate filter, the nitrogen oxides in the case of the nitrogen oxide trap). Similarly, it is necessary to know the evolution over time of the masses converted during the regeneration phases. However, during these regeneration phases, the polluting species (mass stored) react with certain exhaust gases whose nature strongly depends on the ratio between the amount of fuel on the quantity of air injected into the cylinders of the engine, it is that is, the richness of the air / fuel mixture. We thus seek to know, if not to control, the richness of this mixture.

  Moreover, the increase in the complexity of the engines and their modes of operation requires electronic management means increasingly sophisticated and thereby the means of measurement or estimation more and more numerous. But it turns out that many physical quantities are not directly measurable, or that the necessary sensors are too expensive or even unsuitable. It therefore appears necessary, for the control of the regeneration of the traps and for the control of the engine itself, to estimate the quantity of polluting species present in the exhaust gas. To do this, the estimation of the richness of the air / fuel mixture is particularly useful. To know the wealth, we measure today, knowing the amount of fuel injected (quantity provided by an electronic calculator), the proportions of the species present in the exhaust line. However, because of the dispersion of the injectors, there is a difference between the amount of fuel injected that the electronic computer indicates (setpoint) and the amount of fuel actually injected into the engine cylinders. This dispersion of the injectors is related to the conditions that prevail in the combustion chamber (temperature, high pressure, ...), conditions that greatly alter the behavior of the injectors over time. To date, and to know the evolution of the species present in the exhaust line, using a probe 02 which provides a value of wealth. The knowledge of this wealth, coupled with the amount of fuel control that the electronic computer sends to the injector, allows it to perform a correction of the setpoint over time.

  Several disadvantages are however related to the use of such a probe placed in the exhaust line.

  The accuracy of a sensor is indeed inversely proportional to the field of use. That is, the more it is desired to measure over a wide range of use, the lower the measurement accuracy. This accuracy can also derive with the aging and fouling of the sensor. In addition, the cost of using a probe can be important. It is actually necessary to associate the intrinsic cost to the probe that of the connector, the input port in the computer and the software driver. It is also necessary to have suitable means to diagnose the operating state of the probe. Ultimately, the electronic computer associated with the probe 02 makes it possible to overcome the error on the quantity of fuel actually injected only very moderately. It is also possible to use theoretical models to determine the fuel flow injected into the engine, the engine output richness and then the quantity of species present in the exhaust line, but this use of models also has drawbacks. Indeed, if these models are generally faithful in steady state gas, they are rather poor transient. In addition, many parameters (physical parameters, state variables) necessary for these models are difficult to identify or to measure on the motor. Most: time, they are too greedy in computing time and / or memory. An object of the invention is to improve existing methods and systems, in particular for estimating the fuel flow injected into the engine. Another objective of the invention is to provide an estimate of the richness of an air / fuel mixture with a model that is more dynamic and less expensive in terms of memory and calculation time. For this purpose, it is provided in the context of the present invention a method for estimating the richness of the air / fuel mixture of the exhaust gas of an engine, characterized in that it estimates a fuel flow injected into the motor by means of a neural network estimator. The method according to the invention may furthermore have at least one of the following characteristics: the estimation of the fuel flow is carried out on the basis of information relating to the temperature of the gases upstream of a turbine disposed in the exhaust line of the engine; an estimate of the richness is calculated based on the estimated fuel start estimate provided by the neural network estimator; at least one of the estimates upstream of at least one post-processing system arranged in the exhaust line of the engine is carried out; The fuel flow rate injected into the engine is post-processed; a new estimation of the fuel flow is made according to an outcome of the post-processing step; pretreatment of one or more variables at the input of the estimator is performed by performing calculations based on predetermined physical relationships; - Reprocessing of some of the quantities from the neural network is carried out before these are returned to the input of the network. According to another aspect, the invention relates to a device for estimating the richness of an air / fuel mixture of the exhaust gases of an engine, characterized in that it comprises a neural network estimator for providing an estimate of the fuel flow injected into the engine. According to still other aspects, the invention relates to an internal combustion engine equipped with an estimation device according to the invention.

  Other characteristics, objects and advantages of the present invention will appear on reading the detailed description which follows, and with reference to the appended drawings, given by way of non-limiting examples and in which: FIG. schematic view of an internal combustion engine according to one aspect of the invention; FIG. 2 is a schematic view of a system for estimating the richness of the exhaust gases according to one aspect of the invention; FIG. 3 is a diagram showing the steps of the method for estimating the richness of the exhaust gases according to one aspect of the invention. In Figure 1, only the elements necessary for the understanding of the invention have been shown. An internal combustion engine 1. is intended to equip a vehicle such as an automobile. The engine 1 is for example a diesel engine turbocharged four-cylinder in-line and direct fuel injection. The engine 1 comprises an intake circuit 2 providing its air supply, an engine control computer 3, a pressurized fuel circuit 4, and an exhaust line 5 of the gases. The injection of the fuel into the cylinders is ensured by injectors, not shown, opening into the combustion chambers and driven by the computer 3 from the circuit 4. At the output of the engine 1, the exhaust gases discharged into the Exhaust line 5 passes through one or more after-treatment devices 6. Non-limiting examples of post-treatment devices include a particulate filter, a nitrogen oxide filter and a nitrogen oxide catalytic trap. Nox or a 4-way system.

  A turbocharger 7 comprises a compressor disposed on the intake circuit 2 and a turbine disposed on the exhaust line 5. Between the compressor and the engine 1, the intake circuit 2 comprises a heat exchanger 8 for cooling the engine. compressed air at the outlet of the compressor and thus to increase its density, an air intake flap 9 controlled by the computer 3, and a pressure sensor 10 connected to the computer 3. At the output of the engine 1, upstream of the turbine, the exhaust line 5 further comprises means 30 adapted to provide information relating to the temperature of the exhaust gas upstream of the turbine. Said means 30 are for example constituted by a temperature measuring probe or by an estimator provided to provide an estimate of said exhaust gas temperature. The engine 1 also comprises an exhaust gas recycling circuit 11 equipped with a valve 12 whose opening is controlled by the computer 3; it is thus possible to reintroduce exhaust gases into the intake circuit 2. An air flowmeter 13 is mounted in the intake circuit 2 upstream of the compressor to supply the computer 3 with information relating to the air flow rate. intake feeding the engine. Sensors 14, for example pressure or temperature, may also be provided. The computer 3 comprises, in a conventional manner, a microprocessor or central unit, memory zones, analog / digital converters and various input and output interfaces. The microprocessor of the computer comprises electronic circuits and appropriate software for processing the signals from the various sensors, deduce the engine conditions and generate the appropriate control signals to the various actuators controlled such as injectors.

  The computer: 3 thus controls the fuel pressure in the circuit 4 and the opening of the injectors, and this, from the information supplied by the various sensors and in particular the air mass admitted, the engine speed, and stored calibrations to achieve the desired levels of consumption and performance. The computer 3 further comprises a neural network estimator 15 provided for making an estimate of the fuel flow injected into the cylinders, that is to say into the engine. For the sake of simplicity and the sharing of a certain number of resources, in particular calculation and memory resources, it is particularly advantageous to have the estimator 15 within the computer 3. As illustrated in FIG. The estimator 15 comprises a preprocessing module 16, a neural network 17 and a post-processing module 18. The estimator 15 further comprises a loop provided for returning to the input of the neural network one or more of the output variables. said network. The neural network 17 has an output channel relating to the estimation of the desired fuel flow rate, as well as possibly one or more other output channels relating to non-measurable state quantities to be looped back directly. or indirectly, to the inlet of the network 17. The estimator 15 furthermore comprises entries relating to the exhaust gas temperature information upstream of the turbine disposed on the exhaust line, information provided by the means 30 The estimator may also include one or more other entries relating to physical quantities representative of the state of the engine, such as, for example: the injected air flow, the injection timing, the pressure; the engine torque, - the engine speed, the gear ratio, 5 - the gas flow rate in the exchanger. The neural network 17 consists of a certain number of neurons defined by their parameters (weight, bias) and by their activation functions. The output s of a neuron is related to the inputs (el, e2, en) of the neuron by s = F (el * wl + e2 * w2 + ... + in * wn + b), where F is the io function of activation of the neuron, wl, w2, ... wn the weights and b the bias. The number of neurons in the network, the values of the weights and the biases of the different neurons are calibration parameters which can in particular be determined during a learning phase which will be described further later. The preprocessing module 16 makes it possible to process one or more input variables of the estimator 15 by performing calculations based on predetermined physical relationships. The pretreatment module 16 makes it possible in particular to reduce the number of inputs of the neural network 17 by calculating one or more variables al, a2 each representative of one or more physical quantity (s), measurable (s). or not, absent at the input of the computer 3, from several input variables. The module 16 also makes it possible to filter certain inputs that may take outliers. In the case in point, the quantity of fuel injected into the engine is not an input of the estimator 15, unlike the temperature upstream of the turbine which is arranged on the exhaust line 5 The post-processing module 18 makes it possible to process the estimation of the fuel flow injected into all the cylinders and therefore into the engine (mass or volume flow) available at the output of the neural network 17 to transmit information on this fuel flow estimated at the output of the estimator, available to the computer 3. The calculator 3 then determines the richness R of the air / fuel mixture according to the formula R = 14.5 'or another wealth estimator where Qair Qa; air flow injected into the cylinders (this flow rate is supplied by the flow meter 13) and Q; r ,; the fuel flow injected into the engine by the neural network. Insofar as the quantity of fuel injected is not an input of the estimator 15, the estimator 15 is insensitive to the dispersion of the injectors over time. It is therefore within the scope of the invention, knowing the richness, correct the drift injectors. Furthermore, it is noted that the richness is estimated upstream of the turbine 7, and therefore upstream of the exhaust line 5. Thus, better control and optimization of the regeneration phases of the post-treatment elements are carried out. 6. The processing carried out by the post-processing module 18 may be, for example, a filtering making the estimator robust by limiting the return to the input of an outlier. As mentioned above, one or more of the output variables of said network follow a loop and are returned to the input of the neural network. Some of the output variables of the network may be processed (e.g., filtered, delayed, or associated with other entries in the preprocessing module) before being returned to the input of the network. With reference to FIG. 2, the case is thus illustrated in which state variables X31, 132, 133 (which can not be measured and determined only during learning) available at the output of network 17 are looped back to the input of said 17. There is also illustrated the case where the output S of the estimator (ie the estimate of the richness) available at the output of the post-processing module 18 (and whose value is here measurable during learning) is looped back after being delayed (see retarder represented by the symbol z-1) to the input of the network 1 and to the preprocessing module 16. By implementing such a return (also called feedback) of one or more of the variables at the output of the neural network, said network uses as input of its estimation function one or more of the values predicted at the preceding calculation steps by this same estimation function, and that these values were measured or not measured rs of the learning phase. Such a return of variables has the advantage of associating with the management of highly dynamic or even non-linear phenomena the taking into account of the physical history of the system. The neural network 17 can thus be described as recurrent or dynamic, and differs in this respect from the current non-recurrent simplistic models. The estimator 15 can be seen as implementing a transfer function taking input temperature information upstream of the turbine 7, as well as possibly one or more other information relating in particular to physical quantities, to provide at output the estimation of the fuel flow injected into all the cylinders of the engine (and therefore into the engine), the calculator calculating the wealth on the basis of this information. The description below relates to the design and learning of the estimator 15. For example, an estimator developed and implemented in the electronic engine management software, it is a question of choosing the neural network, in particular by determining the number of neurons and calibrating the parameters (weight, bias). Learning the neural network can be done on a computer from data collected on the vehicle.

  The neural network may be subject to learning by a learning algorithm method, in particular using a database 19. The database 19 may be fed from results of real track tests of a vehicle with tests at different gear ratios, with different accelerations and decelerations, all chosen to be representative of the normal operating conditions of the vehicle. In addition, it will be possible to split the database fed from the track tests into a learning base and a test base, so as to reduce the size of the database 19 which forms the basis of the database. 'learning. The database 19 may be part of the computer 3 or, alternatively, be a separate base of the vehicle and implemented during initialization operations of the vehicle or during maintenance operations. Figure 3 illustrates the design and learning steps of the neural network estimator. In step 20, tests are performed to generate the data to inform the database 19 (see arrow to the base 19 in Figure 2), the data being representative of interesting operating areas. It is indeed a question of representative travel of the variation space (or at least a part) of the inputs and outputs of the estimator. It may be useful to perform data cleansing by signal processing, for example to remove outliers, to remove redundant dots, to re-synchronize or to filter the data. In step 21, the data is split into two parts to form a test base and a learning base. In step 22, the pretreatments to be executed by the preprocessing module 16 are determined and the neural network 17 performs the training from the training database. In step 23, the performance is tested with one or more validation criteria, on the test basis and on the learning basis. In step 24, the selection of the neural network 17 is made and in step 25, the characterizations of the performances and tests on the vehicle are carried out. The arrow from the base 19 in FIG. 2 illustrates how the data of the base 19 is used for the learning and validation of the estimator 15. Between the steps 23 and 24, a loopback can be provided. allowing to go upstream of step 22 to perform a certain number of iterations with possible modifications on the number of neurons, the inputs, the outputs, the division into number of networks, the architecture of the network or networks neurons 17, type of learning, optimization criteria, etc.

  This iterative process (loopback 26) allows to elaborate the desired estimator (precision, robustness), at lower cost (number of entries, number of calculations, number and difficulty of the tests, etc.). Between steps 24 and 25, loopback 27 may be provided when it turns out that the data generated in step 20 is insufficient. It then returns to step 20 to generate new data to replace the previously acquired data or complete them to establish a sufficient number of measurement points. For example, there may be 5000 measurement points in the learning base.

  During the test step 23, selection criteria can be provided for selecting the set of parameters allowing the most accurate estimation (for example by removing points whose error is too important, by looking for an average of error close to 0.05, or looking for a low rolling average error).

  Finally, the implementation of the invention for estimating the richness of the exhaust gas consumes little of the resources of the computer (computing load, memory required), uses a limited number of parameters, and can thus be easily implemented. Moreover, since the database useful for learning is sufficiently representative of all the subsequent conditions of use, the richness of the gases upstream or downstream of an after-treatment system can be estimated with accuracy (of the order of a few degrees), and this whether steady state or transient. Finally, the combination of the invention with a physical sensor makes it possible either to diagnose the sensor or to use adjustment or adaptation strategies for managing the post-processing systems. 30

Claims (9)

  A method for estimating the richness of the air / fuel mixture of the exhaust gas of an engine (1), characterized in that a fuel flow rate injected into the engine is estimated by means of an estimator (15). ) with neural network (17).   2. Method according to claim 1, characterized in that the estimate of the fuel flow from an information relating to the temperature of the gas upstream of a turbine (7) disposed in the exhaust line. (5) motor.   3. Method according to one of the preceding claims, characterized in that calculates an estimate of the wealth on the basis of the estimation of the fuel flow provided by the neural network estimator.   4. Method according to one of the preceding claims, characterized in that carries out at least one of the estimates upstream of at least one post-processing system (6) disposed in the line (5) exhaust of the motor.   5. Method according to one of the preceding claims, characterized in that it carries out a post-treatment of the estimation of the fuel flow injected into the engine.   6. Method according to the preceding claim, characterized in that realizes a new estimate of the fuel flow as a function of an outcome of the post-processing step.   7. Method according to one of the preceding claims, characterized in that performs a pretreatment of one or more variables input to the estimator by performing calculations based on predetermined physical relationships.   8. Method according to one of the preceding claims, characterized in that it carries out a reprocessing of some of the quantities from the neural network before they are returned to the input of the network.   9. Device for estimating the richness of an air / fuel mixture 10 of the exhaust gas of an engine (1), characterized in that it comprises an estimator (15) with a neural network (17) for provide an estimate of the fuel flow injected into the engine. 1O.Internal combustion engine (1) comprising a device for estimating the richness of the air / fuel mixture of the engine exhaust gas according to claim 9.