JPH10274082A - Engine control system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To grasp dynamic behavior of injection fuel correctly by correcting a parameter on the basis of a difference between an exhaust air-fuel ratio and a target air-fuel ratio when the exhaust air-fuel ratio is feed back-controlled, and also obtaining its correcting result as an educator data of control logic. SOLUTION: A model base control unit 20 for deciding a fuel injection rate by a reverse model of an engine on the basis of an output of the order model of an engine to be modeled and a target air-fuel ratio Ep is arranged on a control device 10 for controlling an injection rate of fuel supplied to an engine 1. At a transient time of an engine operating condition, a correcting signal A for correcting a sliding of the model base control unit 20 on the basis of the output of a linear sensor 15 is decided by a correcting rate deciding unit 60, and is supplied to the model base control unit 20. Namely, a sliding rate is detected by the correcting signal A at the time of the transient time on the basis of the difference between the target air-fuel ratio Ep and a real measuring value E of an air-fuel ratio, and a correcting rate for correcting the sliding rate is used as an educator data for learning the correction of sliding.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an engine control system for controlling an air-fuel ratio to a predetermined value.

[0002]

2. Description of the Related Art In recent years, in order to respond to the demands for stricter exhaust gas regulations and to improve fuel efficiency and responsiveness from the market, control is performed so that the air-fuel ratio of an engine is adjusted to a target air-fuel ratio. ing. As such an air-fuel ratio control method, a so-called O2 feedback control method in which an oxygen sensor is provided in an exhaust system of an engine and the output of the oxygen sensor is fed back to correct the fuel injection amount has been conventionally used. However, this control method has a problem that the response is particularly poor in a transient state because of feedback control.

[0003]

As a control method for improving such a response problem, the applicant of the present invention has disclosed a Japanese Patent Application No.
In No. 71188, a forward model of the engine is constructed using feedforward control logic with a learning function, and an air-fuel ratio is controlled by an inverse model of the engine that determines the fuel injection amount based on the output of the forward model, and the necessary A control method for learning by correcting the deviation of the forward model by feeding back the output of the oxygen sensor accordingly has already been proposed. According to this control method, the forward model of the engine is configured using the feedforward control logic with a learning function, so that the response is better than the conventional feedback control method, and if necessary, the response from the oxygen sensor is reduced. Since the learning is performed by correcting the forward model based on the actually measured values obtained, it is possible to sufficiently cope with changes over time. However, in an actual engine, all the fuel injected from the fuel injection device does not directly enter the cylinder. After the part adheres to the wall of the intake pipe, etc., it evaporates and enters the cylinder.Therefore, as described above, a forward model of the engine is configured, and a deviation between the model and the actual engine is corrected. In order to learn this, it is necessary to accurately grasp the dynamic behavior of the injected fuel, and therefore, it is necessary to accurately correct the dynamic behavior of the injected fuel even when correcting the deviation of the forward model of the engine. It occurs. An object of the present invention is to provide an engine control method using a forward model of an engine configured with a feedforward control logic with a learning function capable of accurately grasping the dynamic behavior of the injected fuel.

[0004]

In order to achieve the above-mentioned object, an engine control system according to the present invention provides a dynamic behavior of fuel injected from a fuel injector using at least a feedforward control logic with a learning function. In an engine control system that adjusts the exhaust air-fuel ratio of the engine to the target air-fuel ratio by a forward model of the engine having a fuel system forward model that models parameters representing the following, feedback of the exhaust air-fuel ratio using an air-fuel ratio sensor The method is characterized in that the parameters are corrected based on the difference between the exhaust air-fuel ratio and the target air-fuel ratio, and the correction result is obtained as teacher data of the control logic.

[0005]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an engine control system according to the present invention will be described below with reference to an embodiment shown in the accompanying drawings. FIG. 1 is a schematic diagram showing a relationship between an engine 1 and a control device 10 capable of executing an engine control method according to the present invention. In the engine 1, an air-fuel mixture is introduced into a combustion chamber of a cylinder 5 via an air cleaner 3 and a fuel injection device 4 provided in an intake pipe 2, and exhaust gas after combustion is exhausted to the atmosphere via an exhaust pipe 6. This is a four-cycle engine, and other components such as an intake valve and an exhaust valve are omitted in the drawing. In FIG. 1, reference numeral 7 denotes a crankcase, and reference numeral 8 denotes a throttle valve. The control device 10 controls the value of the air-fuel ratio in the exhaust gas by operating the amount of fuel injection from the fuel injection device 4. As shown in FIG. 1, the control device 10 includes information α regarding the throttle opening obtained from a throttle opening detecting means 12 provided in the throttle valve 8.
And information r relating to the crank angle obtained from the crank angle detecting means 13 provided in the crankcase 7, information relating to the intake pipe wall temperature, and information relating to the atmospheric pressure.
Based on the input information, an operation amount Mf (that is, a fuel injection amount) of the fuel injection device 4 provided in the intake pipe 2 is determined and output, and the air-fuel ratio linear sensor 15 provided in the exhaust pipe 6 is determined. (Hereinafter, simply referred to as a linear sensor 15) A signal E of an actual measured value of the air-fuel ratio obtained from the input is input, and correction and learning based on this information are performed as necessary, so that optimal control can always be performed. Have been.

FIG. 2 is a schematic block diagram showing the configuration of the control device 10. The control device 10 determines a fuel injection amount by an inverse model of the engine based on the output of the forward model of the engine modeled using the feedforward control logic with the learning function and the target air-fuel ratio Ep. A target air-fuel ratio calculation unit 30 for calculating the target air-fuel ratio Ep, an engine speed calculation unit 40, a conversion unit 50,
And a correction signal A (A1 to A3) for correcting the displacement of the model base control unit 20 based on the output of the linear sensor 15 when the operating state of the engine is in a transient state, and outputs it to the model base control unit 20. Deviation correction amount determination unit 6
0 is provided. The model base control unit 20 calculates the engine speed n, the throttle opening α, the atmospheric pressure, the intake pipe wall temperature calculated by the engine speed calculation unit 40, and the target air-fuel ratio calculated by the target air-fuel ratio calculation unit 30. Ep, and inputs the basic operation amount Mf of the fuel injection device 4 based on the information.
n is determined, and the conversion unit 50 converts the basic operation amount Mfn into a fuel injection cycle of the engine 1 and outputs the operation amount Mf from the control device 10. Target air-fuel ratio calculation unit 30
Inputs the engine speed n, the throttle opening α, and the crank angle r, determines a target air-fuel ratio Ep that matches the engine operating state at that time based on these information, and determines the model-based control unit 20. Output to The shift correction amount determination unit 60 determines whether the operating state of the engine is in a transient state or a steady state based on the throttle opening signal α of the throttle opening detection signal 12, and determines whether the engine is in a steady state. The amount of deviation between the engine 20 and the actual engine 1 is detected based on the difference between the target air-fuel ratio Ep and the measured air-fuel ratio E at a predetermined time while the operating state is in the steady state,
In addition, the deviation between the model base control unit 20 and the actual engine 1 in the transient state is detected based on the difference between the target air-fuel ratio Ep and the measured air-fuel ratio E while the operating state is in the transient state. And correction amounts A1 to A3 for correcting these deviations.
Is determined and output to the model base control unit 20. The correction amounts A1 to A output from the shift correction amount determination unit 60
Reference numeral 3 is used to correct a model shift of the model-based control unit 20, and the correction result is used as teacher data for the model-based control unit 20 to learn the shift correction.

(Regarding Model-Based Control Unit) The configuration of the model-based control unit 20 will be described below with reference to FIGS. FIG. 3 is a schematic block diagram showing the configuration of the model base control unit 20 in FIG. The model-based control unit 20 includes an air system forward model 21 that models the dynamic behavior of air in the intake pipe 2 using feedforward control logic with a learning function, and a fuel injection device using feedforward control logic with a learning function. Fuel system forward model 2 that models the dynamic behavior of the fuel injected from 4
2, and the estimated air-fuel ratio Ev based on the estimated air amount Av and the estimated fuel amount Fv output from the forward models 21 and 22.
Is calculated. Also,
The model-based control unit 20 includes a feedback loop that feeds back the estimated air-fuel ratio Ev output from the estimated air-fuel ratio calculation unit 23 to the basic operation amount calculation unit 24.
The basic manipulated variable calculator 24 includes an estimated air-fuel ratio Ev output from the estimated air-fuel ratio calculator 23 and a target air-fuel ratio calculator 30.
The target air-fuel ratio Ep output from the
The basic operation amount Mfn (basic fuel injection amount) for the fuel injection device 4 is calculated. The basic operation amount Mfn is output from the model base control unit 20 and also input to the fuel system forward model 22, and the fuel system forward model 22 calculates an estimated fuel amount Fv based on the basic operation amount Mfn. As described above, in the model-based control unit 20, the air system forward model 21, the fuel system forward model 22, and the estimated air-fuel ratio calculation unit 23
, And is output from the forward model of the engine 1 using a feedback loop including the fuel system forward model 22, the estimated air-fuel ratio calculation unit 23, and the basic operation amount calculation unit 24. An inverse model of the engine that outputs the basic operation amount Mfn by feeding back the estimated air-fuel ratio Ev is configured.

FIG. 4 is a schematic block diagram showing the structure of the fuel system forward model 22. The fuel system forward model 22 models the dynamic behavior of the fuel injected from the fuel injection device 4 as described above. The fuel system forward model 22 includes a non-adhered fuel operation unit 22a, an adhering fuel operation unit 22b, a first-order lag unit 22
c, 22d, and a fuel adhesion rate / evaporation time constant estimating unit 22e
And estimates the amount of fuel actually entering the cylinder 8 from the basic operation amount Mfn (basic fuel injection amount) input from the basic operation amount calculation unit 24. The fuel adhesion rate / evaporation time constant estimating unit 22e uses a fuzzy neural network that has learned in advance the relationship between the engine speed n and the throttle opening α, the fuel adhesion rate x, and the evaporation time constant τf of the attached fuel. FIG. 5 shows a model of a fuel adhesion rate x and an evaporation time constant τf as parameters of the dynamic behavior of the injected fuel.
), The engine speed n and the throttle opening α are input, and based on these information, the ratio x (that is, the fuel adhesion rate) at which the fuel injected from the fuel injection device 4 adheres to the wall surface of the intake pipe 2 or the like. x) and the time constant τf at which the attached fuel evaporates
(That is, the evaporation time constant τf). Note that the evaporation time constant τv of the fuel injected from the fuel injection device 4 and directly entering the cylinder is a very small value and does not affect the estimation of the injected fuel entering the cylinder. The calculation is performed as a predetermined constant. Non-adhered fuel calculation unit 22a
Is based on the fuel adhesion rate x obtained from the fuel adhesion rate / evaporation time constant estimating unit 22e, the fuel injection device at the basic operation amount Mfn (that is, the basic fuel injection amount) input from the basic operation amount calculating unit 24. The amount of fuel directly entering the combustion chamber of the cylinder 5 from 4 is calculated. Adhered fuel calculation unit 22b
Is based on the fuel adhesion rate x obtained from the fuel adhesion rate / evaporation time constant estimating unit 22e and once adhered to the wall surface in the basic operation amount Mfn (basic fuel injection amount) input from the basic operation amount calculating unit 24. The amount of fuel entering the cylinder 5 later is calculated. The amount of fuel obtained from the non-adhered fuel calculating unit 22a and the amount of fuel obtained from the adhering fuel calculating unit 22b are respectively the first-order delay unit
2c and 22d, the fuel adhesion rate / evaporation time constant estimation unit 2
2e, is approximated by a first-order lag system based on the evaporation time constant τf and the constant τv, and is then added to obtain the estimated fuel amount F.
It is output from the fuel system forward model 22 as v. Normally, when modeling the behavior of the fuel injected from the fuel injection device 4 in the engine 1, the dead time until the injected fuel enters the cylinder 5 from the fuel injection device 4 is considered in FIG. Although it is necessary to provide a dead time phase delay unit 22g for delaying the phase by the dead time as shown by, the fuel system forward model 22 in the present embodiment advances the phase of the fuel system forward model by the dead time. This eliminates the need to provide the dead time phase delay unit 22g. As a result, the fuel system forward model 22 becomes a simple first-order lag system, so that when performing feedback control using the output of the fuel system forward model 22, it is possible to increase the feedback gain, and the proper It constitutes an accurate inverse model that can obtain a large basic operation amount.

The fuel system forward model 22 configured as described above converts the correction signals A1 and A2 output from the shift correction amount determination unit 60 for detecting a shift between the fuel system forward model 22 and the actual engine. Learning is performed as teacher data to correct model deviation. A method of detecting a deviation between the fuel system forward model 22 and the actual engine by the deviation correction amount determining unit 60 will be described later in detail.

FIG. 6 is a schematic block diagram showing the configuration of the air system forward model 21. The air system forward model 21 includes a throttle opening phase advance section 21a, an air amount calculation section 21b, a pressure conversion section 21
c, an intake negative pressure calculating unit 21d, a volume efficiency estimating unit 21e, and an engine speed phase lead unit 21f.

(Regarding each phase advance section 21a and 21f) The phase advance section 21a for the throttle opening and the phase advance section 21f for the engine speed are provided by the fuel system forward model 2
2, the phase of the throttle opening α and the phase of the engine speed n that are input are advanced by the removed dead time (that is, the time from when the injected fuel is injected from the fuel injection device 4 to when it enters the cylinder 5). . Specifically, each of the phase advance units 21a and 21f includes a neural network in which a change pattern of the engine speed or the throttle opening with respect to time is previously learned. The phase is advanced by obtaining a future value of the engine speed or the throttle opening based on the engine speed or the throttle opening at. As described above, in the air system forward model 21, the phases of the throttle opening and the engine speed are advanced by the dead time, so that the phases of both the fuel system forward model 22 and the air system forward model 21 are advanced by the dead time. In other words, the phase difference between the estimated fuel amount Fv and the estimated air amount Av is eliminated by removing the dead time in the fuel system forward model 22,
The estimated air-fuel ratio calculator 23 can estimate an appropriate estimated air-fuel ratio. Further, for example, in the case of an in-cylinder injection type engine in which there is no dead time until the injected fuel enters the cylinder from the fuel injection device, or in a case where the dead time is so small as to be negligible, the injected fuel is When the behavior is modeled, there is no need to provide a phase delay portion for dead time from the beginning, so that there is no need to provide each phase lead portion 21a and 21f in the air system forward model 21. The method of advancing each phase is not limited to a method using a neural network, but may be any method, for example, a least square method or the like.

The calculation units 21b and 21d for the air amount Av and the intake negative pressure Pman perform modeling using hydrodynamic equations (1) and (2). Here, Ct is the flow rate count at the throttle, D is the diameter of the throttle, Pamb is the atmospheric pressure, k is the specific heat of air, Tam
b is the atmospheric temperature, R is the gas constant, Mao is the correction coefficient, Tma
n is the intake pipe temperature, V is the intake pipe volume, β1 is a count that depends on the throttle opening, and β2 is a count that depends on the intake pipe pressure. In addition, since it is difficult to model the volume efficiency η using mathematical formulas, the volume efficiency estimating unit 21e has learned the relationship between the engine speed signal n and the throttle opening α and the volume efficiency η in advance by using a fuzzy neural network ( Alternatively, modeling is performed using a neural network, CMAC, or the like (not shown).

The air system forward model 21 configured as described above also has an air system forward model 2 like the fuel system forward model 22.
The volume efficiency η output from the volume efficiency estimator 21e is adjusted to the actual engine state by the correction signal A3 output from the shift correction amount determiner 60 that detects the shift between the engine 1 and the actual engine. The volume efficiency estimating unit 21e calculates the volume efficiency η after the correction based on the correction signal A3.
Is used as teacher data to correct model deviation. The method of detecting the deviation between the fuel system forward model 22 and the actual engine by the correction amount determining unit 60 and determining the correction will be described later in detail.

(Target Air-Fuel Ratio Calculating Unit 30) The target air-fuel ratio calculating unit 30 inputs the throttle opening α and the engine speed n, and determines the optimal target air-fuel ratio at each time based on these information. And outputs it to the model base control unit 20.

(Regarding Correction Amount Determining Unit 60) Next, a method of detecting a deviation between the model and the actual engine and a method of determining the correction amount by the correction amount determining unit 60 will be described. The correction amount determining unit 60 includes a linear sensor 15 in addition to the throttle opening signal α obtained from the throttle opening detecting means 12.
And the target air-fuel ratio calculation unit 30
The target air-fuel ratio Ep obtained from the above is input, and it is determined whether the operating state of the engine is in a steady state or a transient state based on the throttle opening signal α, and the actual measured value E of the air-fuel ratio and the target air-fuel ratio Ep are determined. From this, the deviation between the model and the actual engine is detected, and the correction amount as teacher data for learning the corresponding model (the fuel system forward model 22 or the air system forward model 21) is determined.

(Regarding the Correction of the Displacement of the Air-Forward Model 21) The displacement of the air-system forward model 21 (specifically, the displacement of the fuzzy neural network constituting the volumetric efficiency estimator 21e) causes the engine to operate in a steady state At one time, detection is performed based on the difference between the actual measured value E of the air-fuel ratio and the target air-fuel ratio Ep. Based on this difference, for example, when the actual measured value E is richer than the target air-fuel ratio Ep, the air system forward model 21 So that the estimated volumetric efficiency η at
Is leaner than the target air-fuel ratio Ep, the correction signal A3 is determined so that the estimated volumetric efficiency η in the air system forward model 21 becomes smaller. The correction signal A3 is input to the volume efficiency estimator 21e of the air system forward model 21 as teacher data as described above, and the volume efficiency estimator 21e performs learning based on the teacher data by a back pro vacation method or the like. , Update the coupling coefficient of the fuzzy neural network to correct the model deviation. The correction of the deviation of the air system forward model 21 is started at a predetermined timing, for example, when the operation state changes, and is repeatedly performed until the deviation becomes equal to or less than a predetermined value. The method of correcting the deviation of the air system forward model is disclosed in detail in Japanese Patent Application No. 9-20463 previously filed by the applicant.

FIG. 7 shows the deviation of the fuel system forward model 22 (specifically, the deviation of the fuzzy neural network constituting the fuel adhesion rate / evaporation time constant estimating unit 22e). As shown, the measured value E and the target air-fuel ratio Ep while the engine operating state is in the transient state are shown.
7 is determined based on the displacement area S, and specifically, the displacement area S is reduced so that the displacement area S decreases.
As shown in the figure, when the value is larger on the lean side, the correction signals A1 and A2 are determined so that the fuel adhesion rate x and the evaporation time constant τf in the fuel system forward model 21 become larger.
When the shift area is large on the rich side, the correction signals A1 and A2 are determined and output so that the fuel adhesion rate x and the evaporation time constant τf in the fuel system forward model 21 are reduced.
As described above, these correction signals A1 and A2 are input to the fuel adhesion rate / evaporation time constant estimating unit 22e of the forward fuel system model 22 as teacher data, and the fuel adhesion rate / evaporation time constant estimating unit 22e is based on the teacher data. Then, learning by the back pro vacation method or the like is performed, and the coupling coefficient of the fuzzy neural network is updated to correct the model deviation. By this learning, for example, the correction signal A as teacher data
When 1 and A2 are positive, the estimated output of the fuzzy neural network increases by ΔV, and when the correction signals A1 and A2 as teacher data are negative, the estimated output of the fuzzy neural network is ΔV Only decrease. The detection of the shift area S is performed every time the operating state of the engine is in a transition state. When the shift area S is smaller than a predetermined value, the model is not corrected. Next, FIG.
With reference to the flowchart of FIG. 7, the method of detecting the deviation and determining the correction amount for the above-described forward model of the fuel system in the correction amount determination unit 60 will be described more specifically. When the engine is started, the deviation correction amount determination unit 60 first determines from the transient flag whether the transition state is currently being monitored or the deviation area is being detected (step 1). In the case of "", the process proceeds to step 2 for monitoring the transient state, and in the case where the transient flag is "1", the process proceeds to step 4 for detecting the deviation area S. Deviation correction amount determination unit 6
0 monitors whether the operating state has entered a transient state based on the throttle opening signal α (step 2), and sets the transient flag to “1” when the throttle opening signal α changes by a predetermined rate or more. Set (step 3). When the operating state of the engine enters a transient state, the measured value E of the air-fuel ratio is compared with the target air-fuel ratio Ep, and whether the measured value E deviates from the target air-fuel ratio Ep on the rich side or lean side is determined. Is determined (step 5), and if it is off the lean side, the absolute value of the deviation of the measured value E from the target air-fuel ratio Ep is stored as the lean area (step 6), and if it is off the rich side The absolute value of the deviation of the measured value E from the target air-fuel ratio Ep is stored as a rich area (step 7). In the above-described processing, it is determined whether or not the measured value E deviates from the target air-fuel ratio Ep in the same direction (step 4).
The measurement is repeatedly performed while the measured value E continues to deviate in the same direction on either the rich side or the lean side with respect to the target air-fuel ratio Ep (that is, while the operating state is in a transient state), and the actual measurement for the target air-fuel ratio Ep is performed. The absolute value of the deviation of the value E is cumulatively added one after another as a lean area or a rich area (step 6 or 7). If the measured value E deviates from the target air-fuel ratio Ep in a direction different from the previous time, it is determined that the transient state has ended (step 4), and then it is determined whether the lean area is greater than 0 (step 8). When the lean area is larger than 0, it is determined that the measured value E is on the lean side with respect to the target value air-fuel ratio Ep during the transient state, and predetermined coefficients a and b are added to the cumulatively added lean area. The fuel adhesion rate correction amount A1 and the evaporation time constant correction amount A2 are calculated by multiplication (step 9). If the lean area is smaller than 0, it is determined that the measured value E is on the rich side with respect to the target value air-fuel ratio Ep during the transient state (step 8), and a predetermined value is added to the accumulated area. Is multiplied by the coefficients a and b to calculate the fuel adhesion rate correction amount A1 and the evaporation time constant correction amount A2 (step 11). After outputting the fuel adhesion rate correction amount A1 and the evaporation time constant correction amount A2, the stored rich area and lean area are cleared, the transient flag is cleared, and the engine operation state again enters the transient state. Return to the monitoring state of whether or not. As described above, the correction amount detection unit 60
Then, it is determined whether or not the operating state of the engine has entered the transient state based on the change in the throttle opening signal α, and the transient state has been terminated based on the direction of deviation of the measured value E of the air-fuel ratio from the target air-fuel ratio Ep. After the engine operating state has entered the transient state, the absolute value of the deviation while the measured value E of the air-fuel ratio continues to deviate from the target air-fuel ratio Ep in the same direction is cumulatively added. The detected value is detected as a deviation area (that is, a rich area when deviating on the rich side, and a lean area when deviating on the lean side), and after the transient state ends, the deviation area S decreases. Thus, the correction signals A1 and A2 for the fuel system forward model 22 are determined.
In the fuel system forward model 22, as described above, the correction signal A1
And the fuel adhesion rate / evaporation time constant estimating unit 22e using A2
However, since learning is performed and the coupling coefficient of the fuzzy neural network is adjusted to correct the model deviation, when the engine enters a transient state again after learning, the deviation area becomes small, and the engine operating state becomes transient. Responsiveness when entering the state is improved. The above-described processing is performed every time the operating state of the engine enters the transient state, and even if the shift area during the transient becomes large again due to the aging of the engine or the like,
Since the correction and learning are performed immediately, a stable responsiveness can always be obtained when the operating state of the engine enters a transient state. Note that the detection processing of the shift area is always performed during the operation of the engine, but the correction signal output processing based on the shift area is performed only when the shift area is larger than a predetermined value, and the shift area is smaller than the predetermined value. If not, do not do it. This is because when the displacement area is small,
When the learning is performed, due to the effect of the dead time between the change in the throttle opening and the change in the actual air-fuel ratio, for example, the actual measurement value E of the air-fuel ratio slightly deviates from the target air-fuel ratio Ep toward the lean side. In this case, the fuel adhesion rate x and the evaporation time constant τf after the correction / learning become too large, and conversely swing to the rich side, and the actual measured value E of the air-fuel ratio is different from the target air-fuel ratio Ep. If it is slightly deviated to the rich side, the fuel adhesion rate x and the evaporation time constant τf after correction / learning become too small, and conversely swing to the lean side, resulting in an oscillating air-fuel ratio. Is to prevent that

(Second Embodiment) Next, a second embodiment of the engine control system according to the present invention will be described with reference to FIGS. FIG. 9 is a schematic block diagram of a control device 100 that executes the engine control method according to the present invention. The control device 100 determines a fuel injection amount so as to maintain an air-fuel ratio at a stoichiometric air-fuel ratio by an inverse model of an engine using an output of a forward model of the engine modeled using a feedforward control logic with a learning function. A correction amount for correcting a deviation between an engine model and an actual engine in the model base control unit 120 based on an output of the oxygen sensor 115 and a model base control unit 120 and outputting the correction amount to the model base control unit 120 O based on the output from the oxygen sensor 115
And an O2 feedback control unit 170 that performs two-feedback control, and controls the air-fuel ratio along the stoichiometric air-fuel ratio. The model base control unit 120 includes a fuel system forward model and an air system forward model, and determines the basic fuel injection amount along the stoichiometric air-fuel ratio based on the virtual air-fuel ratio obtained from these forward models. The internal configuration of the model-based control unit 120, that is, the configuration of each forward model, the method of estimating the air-fuel ratio based on the output of the model, and the like are the same as those of the model-based control unit 20 of the first embodiment. Detailed description is omitted. Further, the configuration of a conversion unit that converts the output of the model-based control unit 120 into a fuel injection cycle, a detection unit that detects each input information, and a calculation unit are the same as those of the control device 10 of the first embodiment. Here, detailed description is omitted. The O2 feedback control unit 170 performs so-called O2 feedback control.

The shift correction amount determining section 160 receives a detection signal E2 obtained from the oxygen sensor 115 in addition to the throttle opening signal α obtained from the throttle opening detection means, and outputs an engine based on the throttle opening signal α. It is determined whether the operating state of the engine is in a steady state or a transient state, and based on the detection signal E2 of the oxygen sensor 115, a deviation between the model and the actual engine is detected.
20 and determines and outputs the correction amount of the model (fuel system forward model or air system forward model), and outputs an O2 feedback start signal and an end signal to the O2 feedback control unit 170 as necessary, so that the engine is in a transient state. The O2 feedback control is performed only for a predetermined period after the entry. The deviation of the air system forward model is determined by the deviation correction amount determination unit 1
At 60, at a predetermined time during steady operation, the oxygen sensor 115
This is performed by performing a so-called O2 feedback control based on the detection signal from. The method of correcting the deviation of the air system forward model is disclosed in detail in Japanese Patent Application No. 9-20463 previously filed by the applicant.

The deviation of the forward model of the fuel system (specifically, the deviation of the fuzzy neural network constituting the fuel adhesion rate / evaporation time constant estimating unit) is shown in FIG. 10, where the operating state of the engine is in a transient state. The shift time of the oxygen sensor 115 during, for example, is detected based on the time when the output signal of the oxygen sensor 115 is substantially maintained at either the rich signal or the lean signal, and the shift time is shortened.
More specifically, when the detection signal E2 of the oxygen sensor 115 substantially keeps the lean signal when the operating state of the engine enters the transient state, the fuel adhesion rate x and the evaporation time constant in the fuel system forward model When the correction signals A1 and A2 are determined so that τf becomes large, and when the detection signal E2 of the oxygen sensor 115 keeps substantially maintaining the rich signal, the fuel adhesion rate x and the evaporation time The correction signals A1 and A2 are determined and output so that the constant τf becomes small. Further, if the air-fuel ratio when the operating state is in the steady state is slightly shifted in the same direction as the shift direction of the transient state (that is, the rich side or the lean side), the oxygen sensor 115 will continue to operate even after the transient state ends. Is fixed in the same direction as the shift direction in the transient state, and the end of the transient state cannot be determined.
A value of 0 outputs a control start signal to the O2 feedback control section 170 when the operating state of the engine enters a transient state when detecting a shift time of the oxygen sensor 115.
As a result, when the transient state ends, the output of the oxygen sensor 115 starts oscillating between rich and lean as shown in FIG. 10 by the output from the O2 feedback control unit 170, so that the deviation time can be reliably detected. become. In this case, when the transient state ends and the output of the oxygen sensor 115 starts oscillating between rich and lean, the deviation correction amount determination unit 160 outputs a control end signal to the O2 feedback control unit 170 to stop the O2 feedback control. Let it. If the output signal of the oxygen sensor 115 changes due to other causes such as noise regardless of the change in the air-fuel ratio, it is preferable not to use the change to determine the deviation time. By not using the change in the cause for the determination of the shift time, the accuracy of detecting the shift time is improved, and the shift of the model can be detected more accurately. The above-mentioned correction signals A1 and A2 are used as teacher data of the fuzzy neural network in the fuel adhesion rate / evaporation time constant estimating unit of the fuel system forward model, similarly to the control device of the first embodiment. The detection of the shift time is performed every time the operating state of the engine becomes a transient state. Next, with reference to the flowchart of FIG. 11, a method of detecting the deviation and determining the correction amount for the above-described forward model of the fuel system in the correction amount determination unit 160 will be described more specifically. When the engine is started, the correction amount determining unit 160 first determines from the transient flag whether the transient state is currently being monitored or the deviation time is being detected (step 1), and the transient flag is set to “0”. In this case, the process proceeds to step 2 for monitoring the transient state, and to the step 4 for detecting the deviation time when the transient flag is "1". The correction amount determination unit monitors whether or not the operating state has entered a transient state based on the throttle opening signal α (step 2), and the throttle opening signal α
Changes by more than a predetermined change rate, the transient flag is set to “1”.
(Step 3). When the operating state of the engine enters a transient state, it is determined whether the signal from the oxygen sensor 115 is a rich signal or a lean signal (step 5). If the signal is a lean signal, a lean time counter is started (step 5). 6) In the case of a rich signal, a rich time counter is started (step 7).
In the above-described processing, the fluctuation of the signal of the oxygen sensor 115 is determined. Specifically, the fluctuation of the signal from the rich signal to the lean signal or the fluctuation from the lean signal to the rich signal is determined (step 4). (The signal of either the lean signal or the rich signal) is continuously maintained while the lean time counter or the rich time counter measures the shift time (step 6 or 7). When the signal of the oxygen sensor 115 substantially fluctuates,
It is determined that the transient state has ended (step 4), and then it is determined whether the lean time counter is greater than 0 (step 8). If the lean time is greater than 0, the air-fuel ratio becomes lean during the transient state. It is determined that it is off the side, and the shift time measured by the lean time counter is multiplied by predetermined coefficients a and b to calculate the fuel adhesion rate correction amount A1 and the evaporation time constant correction amount A2 (step 9). If the lean time is smaller than 0, it is determined that the air-fuel ratio has deviated to the rich side during the transient state, and the difference time measured by the rich time counter is multiplied by predetermined coefficients a and b to obtain the fuel. An adhesion rate correction amount A1 and an evaporation time constant correction amount A2 are calculated (step 11). After outputting the fuel adhesion rate correction amount A1 and the evaporation time constant correction amount A2, the stored rich time counter and lean time counter are cleared, the transient flag is cleared, and the operating state of the engine returns to the transient state. It returns to the monitoring state of whether or not it has entered. As described above, the deviation correction amount detection unit 160 determines whether the operating state of the engine has entered a transient state based on the change in the throttle opening signal α, and determines the actual measured value E of the air-fuel ratio with respect to the target air-fuel ratio Ep. It is determined whether or not the transient state has ended based on the deviation direction, and after the engine operating state has entered the transient state, the measured value E of the air-fuel ratio continues to deviate in the same direction with respect to the target air-fuel ratio Ep. Is detected as a lag time (that is, a rich time when deviating on the rich side, a lean time when deviating on the lean side), and based on the detected lag time after the transient state ends. Thus, the correction signals A1 and A2 for the fuel system forward model 22 are determined. In the above-described embodiment, the O2 feedback control is performed at the same time as the start of the transient state, so that the air-fuel ratio when the operating state is in the steady state is the same as the shift direction of the transient state (that is, the rich side or the lean side). Although the inconvenience when the vehicle is slightly deviated in the direction is solved, instead of the O2 feedback control, the air-fuel ratio slightly fluctuates near the stoichiometric air-fuel ratio, for example, after it is determined that the operation state of the engine has entered a transient state. Thus, the output of the model may be varied to such an extent that the operating state of the engine is not affected. Further, when the O2 feedback control is started when the transient state is entered, the O2 feedback control can be continued even after the transient state ends, and the control for correcting the deviation of the air system forward model can be executed as it is. . The model base control unit 120 learns the fuzzy neural network in the fuel system forward model based on the correction signals A1 and A2, and corrects the model deviation. Since this processing is the same as that of the control device in the first embodiment, a detailed description is omitted here. still,
The above-described processing for detecting the shift time is always performed during the operation of the engine, but the correction signal output processing based on the shift time is performed only when the shift time is larger than a predetermined value, and the shift time becomes smaller than the predetermined value. Do not do in case. this is,
When the deviation time is small, if the correction and learning are further performed, due to the effect of the dead time between the change in the throttle opening and the actual fluctuation in the air-fuel ratio, the air-fuel ratio swings in the direction opposite to the direction in which the air-fuel ratio is deviated. This is to prevent vibration as a result (see FIG. 12).

In the embodiment described above, the fuzzy neural network is used as the feedforward control logic with the learning function. However, the present invention is not limited to this embodiment. For example, a neural network, a CMAC, or the like is used. You may. Further, in the present embodiment, the timing at which the operating state of the engine enters the transient state is determined based on the change in the throttle opening. However, this is not limited to the present embodiment, and other parameters, for example, The determination may be made using the engine speed or the like.

[0022]

According to the engine control system according to the present invention described above, at least the feed-forward control logic with the learning function is used to model the parameters representing the dynamic behavior of the fuel injected from the fuel injection device. In an engine control system that controls an exhaust air-fuel ratio of an engine according to a target air-fuel ratio by a forward model of an engine having a system-order model, an exhaust air-fuel ratio is detected using an air-fuel ratio sensor, and the exhaust air-fuel ratio and the target Since the correction amount of the parameter is determined based on the deviation from the air-fuel ratio, and the correction result is obtained as teacher data of the control logic, the parameter representing the dynamic behavior of the fuel of the fuel system forward model is accurately calculated. It is possible to execute more accurate control, and the correction result is used as teacher data. Since learning control can be performed that following the change in the aging and operating conditions of the engine. According to the engine control method according to claims 2 and 3,
Since the correction amount is determined based on the shift area and the shift time, it is possible to easily detect the shift between the forward model of the engine and the actual engine. Claim 3
According to the engine control method according to the above, it is possible to use an inexpensive oxygen sensor or the like. Furthermore, claims 4 to 6
According to the engine control method according to the first aspect, since the lower limit is set for the correction of the parameter, the parameter is not excessively corrected. According to the engine control method of the eighth aspect, the start of the transient state is determined based on a change in the operating state of the engine, and the end of the transient state is determined based on the detection result of the air-fuel ratio sensor. Therefore, a transient state in the operating state of the engine can be accurately determined.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a relationship between an engine 1 and a control device 10 capable of executing an engine control method according to the present invention.

FIG. 2 is a schematic block diagram illustrating a configuration of a control device 10.

FIG. 3 is a schematic block diagram showing a configuration of a model base control unit 20 in FIG.

FIG. 4 is a schematic block diagram showing a configuration of a fuel system forward model 22.

FIG. 5 is a schematic configuration diagram of a fuzzy neural network constituting the fuel adhesion rate / evaporation time constant estimating unit 22e.

FIG. 6 is a schematic block diagram illustrating a configuration of an air system forward model 21.

FIG. 7 is a diagram comparing changes in the air-fuel ratio deviation area S and the fuel injection amount in a transient state before and after learning.

FIG. 8 is a flowchart of a shift correction process in a transient state in the shift correction amount determination unit 60;

FIG. 9 is a schematic block diagram illustrating a configuration of a control device 100 according to a second embodiment that can execute the engine control method according to the present invention.

FIG. 10 is a diagram comparing a time difference of an air-fuel ratio and a change in a fuel injection amount in a transient state before and after learning.

FIG. 11 is a flowchart of a shift correction process in a transient state in the shift correction amount determination unit 160.

FIG. 12 is a diagram showing an air-fuel ratio, an oxygen sensor output signal, and a fuel injection amount showing a state in which the air-fuel ratio oscillates due to excessive correction.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Engine 2 Intake pipe 3 Air cleaner 4 Fuel injection device 5 Cylinder 6 Exhaust pipe 7 Crankcase 8 Throttle valve 10 Control device 12 Throttle opening detection means 13 Crank angle detection means 15 Oxygen sensor 20 Model base control unit 21 Air system forward model 21a Throttle opening phase advance section 21b Air amount operation section 21c Pressure conversion section 21d Intake negative pressure operation section 21e Volume efficiency estimation section 21f Engine speed phase advance section 22 Fuel system forward model 22a Non-adhered fuel operation section 22b Adhered fuel operation Unit 22c first-order lag unit 22d first-order lag unit 22e fuel adhesion rate / evaporation time constant estimation unit 23 estimated air-fuel ratio calculation unit 24 basic operation amount calculation unit 30 target air-fuel ratio calculation unit 40 engine speed calculation unit 50 conversion unit 60 shift correction amount Determiner α Throttle opening r Crank angle n Engine rotational speed Mf operating amount Mfn basic operation value E detected signals of the actual air-fuel ratio Ep target air-fuel ratio Ev estimated air-fuel ratio Av estimated air quantity Fv estimated fuel amount A1~3 correction signal

Claims (9)

[Claims] An engine using a forward model of an engine having a fuel system forward model in which a parameter representing a dynamic behavior of fuel injected from a fuel injection device is modeled using at least a feedforward control logic with a learning function. In an engine control system that controls an exhaust air-fuel ratio in accordance with a target air-fuel ratio, an exhaust air-fuel ratio is detected using an air-fuel ratio sensor, and the forward model of the forward model is determined based on a difference between the exhaust air-fuel ratio and a target air-fuel ratio. An engine control method characterized by acquiring teacher data for correcting a shift. 2. The air-fuel ratio sensor is a sensor that linearly detects an exhaust air-fuel ratio, and corrects the parameter using a deviation area between an exhaust air-fuel ratio obtained from the air-fuel ratio sensor and a target air-fuel ratio. The engine control method according to claim 1, wherein the control is performed. 3. An air-fuel ratio sensor for detecting whether an exhaust air-fuel ratio is rich or lean with respect to a target air-fuel ratio, wherein an invariable time of an output signal obtained from the air-fuel ratio sensor is determined. The engine control method according to claim 1, wherein the parameter is corrected using the parameter. 4. The engine control system according to claim 1, wherein the parameter is corrected until the deviation becomes equal to or less than a preset value. 5. The engine control method according to claim 1, wherein the parameter is corrected until the correction amount of the operation parameter becomes equal to or less than a preset value. 6. The engine control method according to claim 1, wherein the parameter is corrected until the output from the air-fuel ratio sensor becomes oscillating between rich and lean. . 7. The engine control method according to claim 1, wherein the acquisition of the teacher data is performed while the operating state of the engine is in a transient state. 8. The system according to claim 1, wherein a start of the transient state is determined based on a change in an operating state of the engine, and an end of the transient state is determined based on a detection result of the air-fuel ratio sensor. The engine control system according to any one of claims 7 to 13. 9. The engine control system according to claim 8, wherein a change in the operating state of the engine is determined based on a change in a throttle opening and / or an engine speed.