JPH10184431A - Engine control system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the control accuracy of an engine by detecting a physical quantity showing a engine performance in a specified operating state of the engine, evaluating the engine performance on the basis of the detected physical quantity, and using the information on this evaluated result as educator data. SOLUTION: In time of driving an engine 1, a physical quantity showing the engine performance such as a torque variation conforming to the differentiated value of throttle opening, is detected by a control unit 10 and evaluated on the basis of the physical quantity to find out the manipulated variable of operating parameters to secure the requested engine performance including those of throttle opening, engine speed or the like. This control unit 10, for example, is constituted so as to perform learnable feedforward control used with a neural network, a fuzzy neural network, or a map and so on, and information on the evaluated result of the engine performance is learned as educator data, while it is controlled so as to find the manipulated variable of the operating parameters whose evaluation become optimized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an engine control system for controlling an engine using physical quantities representing engine performance.

[0002]

2. Description of the Related Art Conventionally, an engine control method for obtaining an operation amount of an operation parameter of a fuel injection device or the like based on an operation state of an engine and controlling the operation parameter based on the operation amount to obtain a desired engine performance. Is there. In this engine control method, the operation amount of the operation parameter with respect to the operation state of the engine when the engine is in a so-called normal operation state, such as during a steady operation, is operated by using a map or the like obtained in advance by experiment or the like and stored. Determine the amount and
For a predetermined operation state different from a so-called normal operation state at the time of acceleration, startup, warm-up, or the like, a correction map storing correction amounts of the operation amounts for these predetermined operation states is prepared. It corresponds.

[0003]

However, in the above-described conventional engine control method, information (for example, information indicating the engine performance directly at the time of acceleration, startup, warm-up, etc.) is provided.
(In the case of acceleration, the rate of change in torque, etc.), but the operation amount of the operation parameter is determined based on information representing the operating state of the engine at that time, so it is not possible to execute highly accurate control, There is a problem that satisfactory engine performance may not be obtained depending on the operating condition of the vehicle. An object of the present invention is to solve the above-mentioned problems and to provide an engine control system capable of obtaining required engine performance with high accuracy.

[0004]

In order to achieve the above object, an engine control system according to claim 1 of the present invention comprises:
When the engine is in a predetermined operating state, a physical quantity representing the engine performance related to the operating state is detected, and the feed-forward control with a learning function evaluates the engine performance based on the detected physical quantity. The present invention is characterized in that information on a result is used as teacher data for the learning, and an operation amount of a predetermined operation parameter is obtained. In the engine control method according to claim 9 of the present invention, a basic operation amount of a predetermined operation parameter is obtained based on an operation state of the engine, and a physical quantity representing engine performance related to the predetermined operation state of the engine is detected. Then, by feedforward control with a learning function, an engine performance is evaluated based on the detected physical quantity, and information on the evaluation result is used as teacher data of the learning, and a correction amount of the basic operation amount is obtained. It is a feature.

[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 some embodiments shown in the accompanying drawings. The basic concept of the engine control system according to the present invention will be described based on two embodiments shown in FIGS.

FIG. 1 is a control block diagram for executing an engine control system according to a first embodiment of the present invention.
As shown in the drawing, the control device detects a physical quantity representing the engine performance from the engine, and calculates and outputs an operation amount of an operation parameter for obtaining the required engine performance while performing an evaluation based on the physical quantity. The control device is, for example, a neural network, a fuzzy neural network, C
It is configured to perform a feedforward control that can be learned using a MAC, a map, or the like, evaluates the engine performance based on the physical quantity representing the engine performance, and learns information on the evaluation result as teacher data. At the same time, the operation amount of the operation parameter for which the evaluation is optimal is obtained. As described above, it is possible to perform a highly accurate engine control by detecting a physical quantity representing a required engine performance, performing an evaluation based on the physical quantity, and obtaining an operation amount of an operation parameter that optimizes the evaluation. ,
Further, by learning the operation amount obtained based on the evaluation, it becomes possible to execute control with good responsiveness corresponding to a temporal change or the like. Further, the control device described above inputs a physical quantity representing the engine performance as shown by a virtual line in FIG. 1 and evaluates the engine performance based on this physical quantity,
An operation amount of an operation parameter for which an evaluation is optimal may be obtained, and thereafter, the operation amount may be learned as teacher data.

FIG. 2 is a control block diagram for executing an engine control system according to a second embodiment of the present invention.
As shown in the drawing, the control device inputs an operation state of the engine, calculates an operation amount of an operation parameter, detects an operation amount of the operation parameter, detects a physical amount representing engine performance, and performs an evaluation based on the physical amount to perform a request. And a correction value calculation unit for calculating a correction value for the operation amount of the operation parameter for obtaining the engine performance. The operation amount calculation unit includes a map in which operation amounts of operation parameters corresponding to the operating state of the engine are stored, an inverse model that inputs a target value and outputs an operation amount of the operation parameter, and the like. The correction amount calculation unit is configured to perform a feedforward control that can be learned using, for example, a neural network, a fuzzy neural network, a CMAC, or a map, and the evaluation based on a physical quantity representing the engine performance is performed. In addition to calculating the correction amount for the operation amount of the optimal operation parameter,
Learn that information. As described above, the basic operation amount is obtained by a map, an inverse model, or the like, and the correction amount for the operation amount is obtained so as to optimize the evaluation while performing the evaluation based on the physical quantity representing the engine performance. By correcting the operation amount, it becomes possible to perform high-precision engine control, and the correction amount calculation unit learns the correction amount obtained based on the evaluation so that the correction amount calculation unit can also perform a change over time. Corresponding control with good responsiveness can be executed. Further, as described above, the basic operation amount determination method is complicated by obtaining the final operation amount separately from the determination of the basic operation amount of the operation parameter and the calculation of the correction amount of the operation amount. Instead, the same method as the conventional operation amount determination method can be used. The correction amount calculation unit inputs a physical quantity representing the engine performance as shown by a virtual line in FIG. 2, evaluates the engine performance based on the physical quantity, and learns information on the evaluation result as teacher data. In addition, it may be configured to obtain and output a correction amount at which the evaluation is optimal.

Next, an example in which the engine control system according to the present invention is applied to air-fuel ratio control of an engine will be described with reference to FIGS. FIG. 3 is a schematic diagram illustrating a relationship between the engine 1 and a control device 10 that executes the engine control method according to the present invention. As shown in FIG. 3, the control device 10 obtains information α regarding the throttle opening obtained from a throttle opening detecting means 12 provided in the throttle 2.
And the information r on the crank angle obtained from the crank angle detecting means 13 provided on the crankcase 3 and the rate of change in the combustion pressure obtained from the combustion pressure detecting means 14 provided on the cylinder head 4 (that is, on the torque).
Information p1, information t1 relating to the temperature of the cooling water obtained from the water temperature detecting means 15 and oil temperature detecting means 16 provided in the cylinder block 5, and information t2 relating to the temperature of the lubricating oil.
And an intake air temperature detecting means 17 provided in the air cleaner 6.
, Information t3 ′ on the intake pipe wall surface temperature obtained from the intake pipe wall temperature detecting means 17 ′ provided on the intake pipe wall surface, and information on the atmospheric pressure obtained from the atmospheric pressure detecting means 18. p2 as well as an actual control amount E (information on the air-fuel ratio A / F) obtained from the air-fuel ratio detecting means 19 provided in the exhaust pipe 7, and based on the input information, the engine 1
The air-fuel ratio of the engine (i.e., the control amount E) is not only in a so-called normal operation state, but also in a special operation state such as start-up, acceleration / deceleration, warm-up operation, or asynchronous injection.
Is adjusted to a value suitable for these operating conditions.
And determines and outputs the operation amount M (that is, the fuel injection amount) of the fuel injection device 9 provided in the first embodiment.

(Internal Configuration of Control Device 10) The internal configuration of the control device 10 will be described below in detail with reference to FIGS. FIG. 4 is a schematic block diagram showing the internal configuration of the control device 10. As shown in the drawing, the control device 10 controls the basic operation amount M of the fuel injection device 9 based on a target air-fuel ratio Ep corresponding to the operating state of the engine.
f, a target air-fuel ratio control unit 30 that determines the target air-fuel ratio Ep corresponding to the operating state of the engine, and outputs the target air-fuel ratio Ep to the basic operation control unit 20. For example, in a special operation state requiring a rapid torque change such as a sudden acceleration or the like, an asynchronous operation amount Ma for determining the asynchronous injection operation amount Ma for performing the asynchronous injection so that a required torque change rate is obtained. At the time of the injection operation control unit 40 and the special operation state of starting the engine,
A start operation control unit 50 that determines the start operation amount Ms of the fuel injection device 9 instead of the basic operation control unit 20;
A switching unit 60 for switching between the basic operation control unit 20 and the start-time operation control unit 50; and an engine speed n calculated based on a crank angle signal r obtained from the crank angle detection unit 13. And an engine speed calculation section 70 for outputting the output to the engine.

(Basic operation control unit 20) The basic operation control unit 20 includes a target control amount (target air-fuel ratio) in addition to the throttle opening signal α, the engine speed signal n, the atmospheric pressure signal p2, and the intake air temperature signal t3. ) Ep and an actual control amount (air-fuel ratio) E are input, and a basic operation amount Mf of the fuel injection device 9 corresponding to the operating state of the engine is determined and output based on the input information. FIG. 5 is a block diagram showing the internal configuration of the basic operation control unit 20 in FIG. FIG.
As shown in FIG. 2, the basic operation control unit 20 includes an air system forward model 21 that models the behavior of air in the intake pipe 8 and a fuel system forward model that models the behavior of fuel injected from the fuel injection device 9. 22, an estimated control amount calculation unit 23, and a basic operation amount calculation unit 24.
1, a forward model of the engine 1 is constituted by the fuel system forward model 22 and the estimated control amount calculation unit 23, and the output of the estimated control amount calculation unit 23 is fed back to the basic operation amount calculation unit 24 to obtain the basic operation amount Mf. Constitutes an inverse model of the engine that calculates.

(Regarding Air System Forward Model 21) The air system forward model 21 models, for example, the amount of air with a fluidological formula using a throttle opening and an intake negative pressure. The volumetric efficiency is modeled by a hydrodynamic formula using the air amount, the engine speed, and the volumetric efficiency, and the volumetric efficiency is further modeled using a fuzzy neural network (learnable modeling) using the intake negative pressure and the engine speed. Any means may be used as long as it is a means, a simple neural network or a CMAC may be used), and the estimated air amount Av may be obtained using these models. The fuzzy neural network modeling the volumetric efficiency inputs the actual measured value E of the control amount, that is, the actual air-fuel ratio, and the volumetric efficiency such that "the larger the volumetric efficiency of the model, the smaller the actual air-fuel ratio becomes". Based on the relationship between the control amount and the air-fuel ratio, learning may be performed to reduce the error between the actual control amount E and the target control amount Ep during engine operation. Further, since the mathematical model of the air amount and the mathematical model of the intake negative pressure mutually require the air amount and the intake negative pressure as parameters, for example, only the intake negative pressure is measured in advance by a pressure sensor or the like. It is necessary to input an appropriate value such as a value as an initial value. This makes it possible to obtain the estimated air amount Av at that time from the engine speed signal n and the throttle opening signal r.

(For fuel system forward model 22) For example, the fuel system forward model 22 calculates the evaporation time constant of the fuel injected from the fuel injection device 9 by an engine speed signal n, a throttle opening signal r, and an intake pipe wall surface. The model is modeled by a neural network (an arbitrary means may be used as long as it is a modeling means capable of learning, a fuzzy neural network or a CMAC) which receives the temperature signal t3 ', and the intake of the injected fuel is performed. A neural network that uses the engine speed signal n and the throttle opening r as inputs for the fuel adhesion rate to the inner wall of the pipe and the intake valve, etc.
In addition, CMAC) may be used to estimate the evaporation time constant and the fuel adhesion rate of the injected fuel, and the basic operation amount M may be calculated using the evaporation time constant and the fuel adhesion rate.
It is configured to obtain an estimated fuel injection amount Fv corresponding to f. The neural network relating to the evaporation time constant and the neural network relating to the fuel adhesion rate each input the measured value E of the control amount, that is, the actual air-fuel ratio, and when the evaporation time constant of the model is large, the actual air-fuel ratio is The actual control amount is determined based on the relationship between the evaporation time constant and the air-fuel ratio, which is "smaller", and the relationship between the fuel adhesion ratio and the air-fuel ratio, which is "the larger the model fuel deposition ratio is, the smaller the actual air-fuel ratio is" The learning may be performed so as to reduce the error between E and the target control amount Ep.

(Estimated control amount calculation unit 23 and basic operation amount calculation unit 24) The estimated control amount calculation unit 20 includes an estimated air amount Av and an estimated fuel obtained by the air system forward model 21 and the fuel system forward model 22. Input an amount Fv, and calculate an estimated air-fuel ratio, that is, an estimated control amount Ev based on these values,
The estimated control amount Ev is fed back to the basic operation amount calculation unit 24. The basic manipulated variable calculator 24 calculates the target air-fuel ratio Ep based on the target air-fuel ratio Ep output from the target air-fuel ratio controller 30 and the estimated control amount Ev fed back from the estimated control amount calculator 23. The basic operation amount Mf is calculated so that the difference from the estimated air-fuel ratio Ev becomes small. The basic manipulated variable Mf is output as an output of the basic manipulated variable controller 20 and is also input to the fuel system forward model 22.

(Target air-fuel ratio control unit 30) The target air-fuel ratio control unit 30 includes a throttle opening signal α, an engine speed signal n, a crank angle signal r, and a combustion pressure change rate signal p.
1. A water temperature signal t1 and an oil temperature signal t2 are input, and a target air-fuel ratio Ep corresponding to the operating state of the engine is determined based on these input information and output to the basic operation control unit 20. FIG. 6 is a schematic block diagram showing the internal configuration of the target air-fuel ratio control unit 30 in FIG. The target air-fuel ratio controller 30 includes a basic target air-fuel ratio determiner 31, a rotation fluctuation detector 32, a fluctuation allowable value determiner 33, a comparator 34, a correction amount calculator 35, an acceleration / deceleration correction amount determiner 36, and a throttle. An opening degree differential value calculation unit 37 and an additional correction amount calculation unit 38 are provided. The basic target air-fuel ratio determination unit 31 determines a basic target control amount Epf corresponding to the operating state of the engine, and performs other processing. The units 32 to 38 calculate a correction amount of the basic target control amount Epf corresponding to a special operation state, and based on the basic target control amount Epf and each correction amount, a target control amount Epf.
It is configured to determine and output p.

(Basic target air-fuel ratio determining unit 31) The basic target air-fuel ratio determining unit 31 receives the throttle opening α, the engine speed n, the water temperature t1, and the oil temperature signal t2, and receives the basic values corresponding to these temperatures. It comprises a neural network that outputs the target control amount Epf (see FIG. 7). For example, when the water temperature and the oil temperature are lower than predetermined values, the neural network is configured to operate the engine 1 because the engine 1 is in a warm-up operation state.
The engine 1 to be controlled, such as making the basic target air-fuel ratio rich so as to stabilize the rotation of the engine, and making the basic target air-fuel ratio lean when the water temperature and the oil temperature are higher than predetermined values.
The information of the optimal air-fuel ratio corresponding to various water temperatures and oil temperatures is previously learned.

(Regarding rotation fluctuation detecting section, fluctuation allowable value determining section, comparing section, and correction amount calculating section) Rotation fluctuation detecting section 32
Inputs, as feedback information, two predetermined crank angle signals r when the engine 1 is in the expansion stroke, which is obtained from the crank angle detecting means 13, and outputs the crank angle signal r
, An angular velocity v between the two crank angles is calculated for each cycle, and an angular acceleration (acc) is calculated from the angular velocities (v1 and v2) between the two crank angles. Each time the calculation of the angular acceleration is repeated, that is, for each cycle, the average value of the angular acceleration (acc_ave) is obtained, and the angular acceleration (acc) at that time and the average angular acceleration (a) are obtained.
cc_ave) is calculated as the angular acceleration variation value (f
1), and the angular acceleration fluctuation value (f1) is cumulatively added one after another as a combustion deterioration index (pnt). Then, the combustion deterioration index (pnt) obtained by repeating the above processing for 100 cycles is output as rotation fluctuation. On the other hand, the fluctuation allowable value determination unit 33 previously obtains an allowable value (lim) of the rotation fluctuation corresponding to the engine speed and the throttle opening of the engine 1 by an experiment or the like, and stores it in the form of a map, a mathematical expression, or the like. Engine speed signal n and throttle opening signal α
And outputs the permissible value (lim) of the rotation fluctuation at that time. The comparison unit 34 compares the actual rotation fluctuation (pnt) obtained from the rotation fluctuation detection unit 32 with the fluctuation allowable value (lim) obtained from the fluctuation allowable value determination unit 33,
The comparison result is output to the correction value calculation unit 35. When the actual rotation fluctuation (pnt) is larger than the fluctuation allowable value (lim), the correction value calculation unit 35 determines that the rotation of the engine 1 is not stable, and proportionally determines the target air-fuel ratio. The basic target air-fuel ratio E is set so that Ep becomes rich, that is, the amount of fuel injected from the fuel injection device 9 increases.
A correction value Ar for correcting pf is output. If the actual rotation fluctuation (pnt) is smaller than the fluctuation allowable value (lim), it is determined that the rotation of the engine 1 is stable, and the correction value Ar is determined. Ar is set to zero. FIG. 8 is a series of flowcharts showing the processing of the above-described rotation fluctuation detection unit 32, fluctuation allowable value determination unit 33, comparison unit 34, and correction amount calculation unit 35. The rotation fluctuation of the engine is information directly indicating the performance related to the rotation stability of the engine 1. In particular, in a special operation state such as a warm-up operation, if the rotation fluctuation is large, the rotation of the engine becomes stable. And the engine is about to stop, so as described above, a variation allowable value corresponding to the engine speed and the throttle opening is obtained in advance, and the actual rotation variation is larger than the variation allowable value. Then, the basic target air-fuel ratio Epf is corrected and the target air-fuel ratio E
By making p rich, as shown in FIG. 9, the basic operation amount Mf of the fuel injection device 9 output by the basic operation amount control unit 20 increases, and the fuel injection amount increases (that is, the air-fuel ratio increases). The measured value E becomes rich), and the rotation of the engine 1 becomes stable. The neural network constituting the basic target air-fuel ratio determining unit 31 includes a correction amount calculating unit 35 obtained by feeding back the actual rotation fluctuation described above.
The learning is performed in correspondence with the throttle opening, the engine speed, the water temperature, and the oil temperature at that time using the basic operation amount Mf obtained by adding the correction value Ar from the above as teacher data. As a result, after learning, the controllability in an operating state closely related to the rotation fluctuation of the engine 1 such as during a warm-up operation is improved. As described above, in a special operation state such as a warm-up operation, the fuel injection amount is controlled while correcting the basic control amount Mf by feeding back the engine rotation fluctuation directly representing the engine performance during the warm-up operation. Further, by learning the results, the responsiveness is better than the control based on other factors, and it is possible to cope with aging, so that the rotational stability of the engine is remarkably improved. Become.

(Acceleration / Deceleration Correction Amount Determining Unit 36) The acceleration / deceleration correction amount determining unit 36 includes a throttle opening α, a throttle opening differential dα / dt obtained from a throttle opening differential calculating unit 37, and It is composed of a neural network that inputs the engine speed n and outputs an acceleration / deceleration correction value As for the basic target control amount Epf corresponding to each of these conditions (see FIG. 10). This neural network, for example,
When the throttle is suddenly opened, the driver seeks rapid acceleration, so the basic target control amount Epf is corrected so that the target air-fuel ratio Ep becomes rich and the torque of the engine 1 rises rapidly. The acceleration / deceleration correction value As for the basic target control amount Epf corresponding to each condition such as the throttle opening degree, such as outputting the deceleration correction value As, is learned in advance.

(Additional Correction Amount Calculator 38) The additional correction amount calculator 38 obtains an optimum combustion pressure change rate ps1 (that is, a torque change rate) corresponding to the throttle opening differential value by experimentation or the like in advance. The optimum combustion pressure change rate ps1 at that time is obtained from the actual throttle opening degree differential value dα / dt obtained from the throttle opening degree differential value calculation unit 37. ps1 is compared with the actual combustion pressure change rate p1 fed back from the combustion pressure detecting means 14, and an additional correction value Asa for the acceleration / deceleration correction value As is calculated and output so that the difference becomes smaller. I do. That is, for example, when the actual combustion pressure change rate p1 is smaller than the optimum combustion pressure change rate ps1 when the throttle is rapidly opened, the engine 1 does not perform the acceleration required by the driver. And the acceleration / deceleration correction value A for the basic target control amount Epf so that the torque change rate increases.
An additional correction value Asa for correcting s is calculated and output. This additional correction value Asa is equal to the acceleration / deceleration correction value As.
The neural network constituting the acceleration / deceleration correction amount determination unit 36 uses the acceleration / deceleration correction value As to which the additional correction value Asa has been added as teacher data, and sets the throttle opening α and the throttle Opening degree differential value dα / dt,
And learning in association with the engine speed n. The rate of change of the combustion pressure of the engine 1, that is, the rate of change of the torque, is information directly indicating the acceleration and deceleration performance of the engine 1. If this is low, the acceleration or deceleration is poor, and if it is high, the acceleration or deceleration is high. Can be said to be intense. Therefore, as shown in FIG. 11, when the torque change rate actually obtained with respect to the throttle opening does not reach the optimum torque change rate, the acceleration / deceleration correction value As is further increased by the additional correction value Asa as described above. After the correction, the target control amount Ep is enriched, that is, the fuel injection amount is increased, and this value is learned by the acceleration / deceleration correction amount determination unit 36. , The optimum torque change required, that is, acceleration / deceleration can be obtained.
As shown in FIG. 11, the operation parameters for controlling the acceleration / deceleration characteristics of the engine are not limited to the air-fuel ratio, that is, the fuel injection amount, but may be other operation parameters such as the ignition timing. These may be controlled simultaneously. For example, in the case of the ignition timing, in order to obtain a sufficient rise of the torque, the ignition timing may be controlled to be advanced.
Although not shown, when the electronic throttle opening is controlled as an operation parameter, the electronic throttle opening can be controlled so as to give the most efficient opening change for the engine, thereby preventing breathing and sudden change of excessive torque. .

(Asynchronous injection operation control unit 40) Next, the asynchronous injection operation control unit 40 in FIG. 4 will be described with reference to FIG. FIG. 12 is a schematic block diagram showing the internal structure of the asynchronous injection operation control unit 40. The asynchronous injection operation control unit 40 includes a basic asynchronous injection operation amount determination unit 41, a target torque change rate determination unit 42, a torque change rate calculation unit 43, a comparison unit 44, a correction value calculation unit 45, and a throttle opening degree differential value calculation. A portion 46 is provided.

(Regarding Basic Asynchronous Injection Manipulation Determining Unit 41) The basic asynchronous injection manipulating amount determining unit 41 includes an engine speed signal n, a throttle opening signal α, and a throttle opening obtained by a throttle opening differential value calculating unit 46. Degree derivative dα
/ Dt, and a fuzzy neural network for outputting the basic asynchronous injection manipulated variable Maf (see FIG. 13). In this fuzzy neural network, for example, an asynchronous injection operation amount for obtaining an asynchronous injection amount corresponding to a throttle opening, an engine speed, and a differential value of the throttle opening is learned in advance.

(Regarding the target torque change rate determining section 42, the torque change rate calculating section 43, the comparing section 44, and the correction value calculating section 45) The target torque change rate determining section 42 determines the throttle opening degree differential value and the engine speed. Then, an optimum torque differential value corresponding to these pieces of information, that is, a target torque change rate hp is obtained. The torque change rate calculation unit 43 feeds back the actual combustion pressure change rate p1 from the combustion pressure detecting means 14, and obtains the actual torque change rate h based on the combustion pressure change rate p1. The comparison unit 44 compares the target torque change rate hp with the actual torque change rate h, and outputs the result to the correction value calculation unit 45. In the correction value calculation unit 45,
A correction value Am for correcting the basic asynchronous injection operation amount Maf is calculated based on the comparison result from the comparison unit 44 so that the actual torque change rate h becomes the same value as the target torque change rate hp, and the basic asynchronous injection operation amount is calculated. It is added to Maf and output from the asynchronous injection operation control unit 40 as the asynchronous injection operation amount Ma. At this time, the fuzzy neural network constituting the basic asynchronous injection operation amount determination unit 41 uses the asynchronous injection operation amount Ma corrected by the correction value Am as teacher data to indicate that “the engine speed is high, the throttle opening is large, When the throttle opening differential value is large, the asynchronous injection amount increases. "" When the engine speed is low, the throttle opening is small, and when the throttle opening differential value is small, the asynchronous injection amount decreases. " Also, the asynchronous injection amount such as "when the engine speed is large, the throttle opening is large, and the throttle opening differential value is small, the asynchronous injection amount is slightly large" and each input such as the engine speed. Learning based on the relationship. Since the torque change rate is information directly representing the acceleration performance of the engine 1, this torque change rate is fed back to correct the basic asynchronous injection operation amount Maf so as to obtain the optimum torque change rate, and to perform the correction after the correction. The learning is performed by the fuzzy neural network constituting the basic asynchronous injection amount determination unit 41 using the asynchronous injection operation amount Ma as the teacher data, and after the learning, an optimal amount of asynchronous injection as shown in FIG. 14 is obtained. Thus, the torque change rate required by the driver can be obtained. The basic asynchronous injection manipulated variable determiner 41 is also configured to learn by directly inputting the information obtained by the correction value calculator 45 as learning information (teacher data), as indicated by a virtual line in FIG. obtain.

(Regarding the starting operation control unit 50)
Referring to FIG. 15, start operation control unit 50 in FIG.
Will be described. FIG. 15 is a schematic block diagram showing the internal structure of the starting operation control unit 50. As shown in the drawing, the starting manipulated variable control unit 50 includes a starting basic manipulated variable determining unit 51, a starting time calculating unit 52, a target starting time storing unit 53, a comparing unit 54, and a correction value calculating unit 55. ing. The start-time basic operation amount determination unit 51 includes a neural network that inputs the water temperature t1, the intake air temperature t3, the atmospheric pressure p2, the throttle opening α, and the engine speed n, and outputs the start-time basic operation amount Msf ( See FIG. 14). In this neural network, the optimal operation amount Msf at the start corresponding to each condition such as the engine speed n is learned in advance. The start time calculation unit 52 feeds back the engine speed signal n, and the time Tr until the engine 1 is actually started is Tr.
Is obtained by measuring the time from when the engine is started to when the engine speed reaches or exceeds a predetermined speed. The target start time storage unit 53 stores in advance information such as an optimal time Tp until the engine starts to be started, for example, one second, and the comparison unit 54 stores the actual time Tr until the engine is started and the above information. The result is compared with the target time Tp, and the result is sent to the correction value calculation unit 55. The correction value calculation unit 55 calculates and outputs a correction value An of the starting basic operation amount Msf such that the actual time Tr becomes the target time Tp. That is, for example, when the time Tr actually required before the engine starts exceeds the target time Tp, the basic operation amount M at the start is increased so that the fuel injection amount increases.
A correction value An is determined so as to correct sf. The start-time operation control unit 50 outputs a value obtained by adding the correction value An to the start-time basic operation amount Msf as a start-time operation amount Ms, and at the same time, the neural network constituting the start-time basic operation amount determination unit 51 is: Then, learning is performed using the corrected starting operation amount Ms as teacher data. Since the time Tr until the engine actually starts is information that directly indicates the engine performance in the starting operation state of the engine, the start-time basic operation amount determination unit 51 constituted by a neural network determines the actual time. If learning is performed using the starting operation amount Ms corrected by the correction value Am obtained by feeding back the Tr as teacher data, the engine starting performance is significantly improved after learning. It should be noted that the start-time manipulated variable calculator 50 can also be configured to learn by directly inputting the information obtained by the correction value calculator 54 as learning information (teacher data), as indicated by a virtual line in FIG. .

(Regarding the Switching Unit 60) The switching unit 60 switches between the basic operation control unit 20 and the starting operation control unit 50 described above. The switching unit 60 is always connected to the starting operation control unit 50 at the time of starting the engine. When the engine speed exceeds a predetermined value, the switching unit 60 determines that the engine 1 has started, and the connection is switched to the basic operation control unit. After that, the state connected to the basic operation control unit 20 is maintained until the engine is restarted.

In the present embodiment described above, the basic target air-fuel ratio determining unit 31, the acceleration / deceleration correction amount determining unit 36, and the starting basic operation amount determining unit 51 are configured by a neural network. At the same time, the basic asynchronous injection manipulated variable determiner 41 is configured by a fuzzy neural network,
The means for configuring these units is not limited to the present embodiment, and any means may be used as long as the means can be learned.For example, a fuzzy neural network may be used instead of the neural network. Also, a neural network may be used instead of the fuzzy neural network,
Further, CMAC may be used instead of the neural network and the fuzzy neural network. In addition, in the present embodiment, in addition to the time when the engine is in the normal operation state, at the time of starting, acceleration / deceleration, warm-up operation, and in the special operation state such as the asynchronous injection operation, Although the operation amount of the fuel injection device 9, that is, the fuel injection amount is used as the operation parameter for controlling the engine, the operation parameter for controlling the engine is not limited to this embodiment, and may be any parameter. Electronic throttle opening,
The ignition timing or the valve timing may be controlled.
In this case, for example, the electronic throttle opening can be controlled so as to increase the torque change rate when the operating state of the engine is in an accelerating state, and to stabilize the rotation of the engine at the time of starting or warming up. The ignition timing may be controlled to be advanced to increase the rate of change of torque when the operating state of the engine is in an accelerating state, and the valve timing may be controlled when the operating state of the engine is at a high load and a high speed. It can be controlled to increase the overlap and reduce the valve overlap at low load and low speed. Further, the various operation parameters described above may be individually controlled, or at least two types of operation parameters may be controlled simultaneously. When controlling two or more types of operation parameters, the processing for each operation parameter may be performed sequentially in series, or may be performed simultaneously in parallel.

[0025]

According to the above-described engine control method according to the first aspect of the present invention, when the engine is in a predetermined operating state, a physical quantity representing the engine performance related to the operating state is detected and learned. With the feedforward control with function, the engine performance is evaluated based on the detected physical quantity, and information on the evaluation result is used as the learning teacher data, and the operation amount of a predetermined operation parameter is obtained. Compared to a control method based on irrelevant information, control accuracy is improved, and more optimal operating conditions can be obtained. In addition, learning is performed using operation amounts obtained from physical quantities representing engine performance as teacher data. Therefore, the responsiveness of the control after learning is improved, and
This has the effect of being able to respond to changes over time. Further, according to the engine control method of the ninth aspect of the present invention, a basic operation amount of a predetermined operation parameter is obtained based on an operation state of the engine, and a physical quantity representing an engine performance related to the predetermined operation state of the engine. And the feedforward control with a learning function evaluates the engine performance based on the detected physical quantity, and uses information on the evaluation result as the learning teacher data and obtains the correction amount of the basic operation quantity. Therefore, in addition to the effects described above, there is an effect that the same method as the conventional operation amount determination method can be used without complicating the method of obtaining the basic operation amount of the operation parameter.

[Brief description of the drawings]

FIG. 1 is a control block diagram for executing an engine control method according to a first embodiment of the present invention.

FIG. 2 is a control block diagram for executing an engine control method according to a second embodiment of the present invention.

FIG. 3 is a schematic diagram showing a relationship between an engine 1 and a control device 10 that executes an engine control method according to the present invention.

FIG. 4 is a schematic block diagram showing an internal configuration of the control device 10.

FIG. 5 is a block diagram showing an internal configuration of a basic operation control unit 20 in FIG.

FIG. 6 is a schematic block diagram showing an internal configuration of a target air-fuel ratio control unit 30 in FIG.

FIG. 7 is a schematic configuration diagram of a neural network constituting the basic target air-fuel ratio determination unit 31.

FIG. 8 shows a rotation fluctuation detecting unit 32 and a fluctuation allowable value determining unit 3
3 is a flowchart illustrating a series of processes of a comparison unit, a correction unit, and a correction amount calculation unit.

FIG. 9 is a diagram showing a relationship among an angular acceleration, a target air-fuel ratio, and an actual air-fuel ratio.

FIG. 10 is a schematic configuration diagram of a neural network constituting the acceleration / deceleration correction amount determination unit 36;

FIG. 11 is a diagram showing the relationship between the throttle opening, the target air-fuel ratio, and the torque separately before and after learning.

FIG. 12 is a schematic block diagram showing an internal structure of an asynchronous injection operation control unit 40.

FIG. 13 is a schematic configuration diagram of a fuzzy neural network constituting the basic asynchronous injection manipulated variable determination unit 41.

FIG. 14 is a diagram showing the relationship between the throttle opening, the synchronous injection pulse, the asynchronous injection pulse, and the torque separately before and after learning.

FIG. 15 is a schematic block diagram showing the internal structure of a start-up operation control unit 50.

FIG. 16 is a schematic configuration diagram of a neural network constituting the starting basic manipulated variable determination unit 51;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Engine 2 Throttle 3 Crankcase 4 Cylinder head 5 Cylinder block 6 Air cleaner 7 Exhaust pipe 8 Intake pipe 9 Fuel injection device 10 Control device 12 Throttle opening detection means 13 Crank angle detection means 14 Combustion pressure detection means 15 Water temperature detection means 16 Oil Temperature detection means 17 Intake temperature detection means 17 'Intake pipe wall surface temperature detection means 18 Atmospheric pressure detection means 19 Air-fuel ratio detection means 20 Basic operation control unit 21 Air system forward model 22 Fuel system forward model 23 Estimated control amount calculation unit 24 Basic operation Amount calculation unit 30 Target air-fuel ratio control unit 31 Basic target air-fuel ratio determination unit 32 Rotation fluctuation detection unit 33 Allowable fluctuation value determination unit 34 Comparison unit 35 Correction value calculation unit 36 Acceleration / deceleration correction amount determination unit 37 Throttle opening degree differential value calculation Unit 38 Additional correction amount calculation unit 40 Asynchronous injection operation control unit 41 Basic Synchronous injection operation amount determination unit 42 Target torque change rate determination unit 43 Torque change rate calculation unit 44 Comparison unit 45 Correction value calculation unit 46 Throttle opening degree differential value calculation unit 50 Start-up operation control unit 51 Start-up basic operation amount determination unit 52 Start time calculation unit 53 Target start time storage unit 54 Comparison unit 55 Correction amount calculation unit 60 Switching unit 70 Engine speed calculation unit α Throttle opening signal r Crank angle signal p1 Combustion pressure change rate signal t1 Water temperature signal t2 Oil temperature signal t3 Intake air temperature signal t3 'Intake pipe wall surface temperature signal p2 Atmospheric pressure signal n Engine speed dα / dt Throttle opening differential value * Control amount = Air-fuel ratio E Control amount actual measurement value Ep Target control amount Epf Basic target control amount Ev Estimated control amount Ar (for rotation fluctuation) correction value As (for acceleration / deceleration) correction value Asa Additional correction value Am (asynchronous) correction value * Manipulated variable = fuel Injection device operation amount M Operation amount Mf Basic operation amount Maf Basic asynchronous injection operation amount Ma Asynchronous injection operation amount Msf Basic operation amount at start Ms Operation amount at start Av Estimated air amount Fv Estimated fuel injection amount v Angular velocity acc Angular acceleration acc_ave angle Acceleration average f1 Angular acceleration fluctuation value pnt Combustion deterioration index lim Rotation fluctuation allowable value h Torque change rate hp Target torque change rate

Claims (16)

[Claims] 1. When the engine is in a predetermined operating state,
A physical quantity representing the engine performance related to the operating state is detected, the engine performance is evaluated based on the detected physical quantity by feedforward control with a learning function, and information on the evaluation result is used as teacher data for the learning. An engine control method which is used and obtains an operation amount of a predetermined operation parameter. 2. The feedforward control with a learning function is performed by a neural network, a fuzzy neural network,
The engine control method according to claim 1, wherein the control is performed using at least one of a CMAC and a map. 3. The engine control method according to claim 1, wherein the operation parameter is at least one of a fuel injection amount, an ignition timing, a valve timing, and an opening of an electronic throttle. 4. The predetermined operating state is an acceleration state, a physical quantity representing engine performance related to the operating state is a torque change rate corresponding to a differential value of a throttle opening, and the feedforward control with learning function 4. An operation amount of the operation parameter is determined so that the torque change rate becomes a target torque change rate corresponding to a preset differential value of the throttle opening.
The engine control method according to any one of claims 1 to 4. 5. The learning function according to claim 1, wherein the predetermined operating state is an operating state at the time of starting the engine, a physical quantity representing engine performance related to the operating state is a time until the engine speed starts to increase, and The engine control method according to any one of claims 1 to 3, wherein an operation amount relating to a fuel injection amount is obtained by forward control such that the time becomes shorter than a preset upper limit time. 6. The predetermined operating state is an operating state during a warm-up operation, a physical quantity representing engine performance related to the operating state is torque fluctuation, and the feed-forward control with learning function causes the torque fluctuation to: The engine control method according to any one of claims 1 to 3, wherein an operation amount relating to the fuel injection amount is obtained so as to fall within a range of a preset threshold value. 7. The engine control system according to claim 6, wherein the torque fluctuation is obtained from a rotation fluctuation. 8. The feed with learning function, wherein the predetermined operation state is an operation state during asynchronous injection, a physical quantity representing engine performance related to the operation state is a torque change rate corresponding to a differential value of a throttle opening, By the forward control, at least one of the asynchronous injection amount, the ignition timing, the valve timing, or the electronic throttle opening is set so that the torque change rate becomes a target torque change rate corresponding to a differential value of a preset throttle opening. The engine control method according to any one of claims 1 to 3, wherein an operation amount of the engine control is calculated. 9. A basic operation amount of a predetermined operation parameter is obtained based on an operation state of the engine, a physical quantity representing engine performance related to the predetermined operation state of the engine is detected, and feedforward control with a learning function is performed. An engine control method which evaluates engine performance based on the detected physical quantity, uses information on the evaluation result as the learning teacher data, and calculates a correction amount of the basic operation amount. 10. The engine control method according to claim 9, wherein the feedforward control with the learning function is performed using at least one of a neural network, a fuzzy neural network, a CMAC, and a map. 11. The operation parameter is a fuel injection amount,
11. The engine control system according to claim 9, wherein the engine control method is at least one of an ignition timing, a valve timing, and an opening degree of an electronic throttle. 12. The predetermined operation state is an acceleration state, the physical quantity representing engine performance related to the operation state is a torque change rate corresponding to a differential value of a throttle opening, and the feedforward control with learning function is 12. A correction amount for a basic operation amount of the operation parameter is determined so that the torque change rate becomes a target torque change rate corresponding to a preset differential value of the throttle opening. An engine control method according to any one of the preceding claims. 13. The feed with learning function, wherein the predetermined operating state is an operating state at the time of starting the engine, a physical quantity representing engine performance related to the operating state is a time until an engine speed starts to increase, The engine control according to any one of claims 9 to 11, wherein a correction amount for the basic operation amount of the fuel injection amount is calculated so that the time becomes shorter than a preset upper limit time by the forward control. method. 14. The predetermined operation state is an operation state at the time of a warm-up operation, a physical quantity representing engine performance related to the operation state is a torque fluctuation, and the torque fluctuation is obtained by the feedforward control with the learning function. The engine control method according to any one of claims 9 to 11, wherein an operation amount with respect to the basic operation amount of the fuel injection amount is determined so as to fall within a preset threshold value range. 15. The engine control method according to claim 14, wherein the torque fluctuation is obtained from a rotation fluctuation. 16. The learning function according to claim 16, wherein the predetermined operating state is an operating state during asynchronous injection, a physical quantity representing engine performance related to the operating state is a torque change rate corresponding to a differential value of a throttle opening, By the forward control, at least one of the asynchronous injection amount, the ignition timing, the valve timing, or the electronic throttle opening is set so that the torque change rate becomes a target torque change rate corresponding to a preset differential value of the throttle opening. The engine control method according to claim 9, wherein a correction amount for the basic operation amount is obtained.