JPS60153504A - Feedback control device having learning function - Google Patents

Abstract

PURPOSE:To improve the accuracy of decision and to improve the accuracy of control moreover in accordance with the decision accuracy by deciding the advancing degree of learning in accordance with real feedback control state in addition to the number of times of learning. CONSTITUTION:A feedback correcting variable deciding means decides a feedback correcting variable to be added/substracted to/from a fixed control variable on the basis of proportional integrating control to make the control value of a controlled system follow up an objective value. A leaning means correctively learns the fixed controlled variable so that the deviation between the control center value of the feedback correcting variable and a reference value is contracted. A learning advancing degree deciding means decides the learning advancing degree in accordance with the number of times of learning of the learning means. A control constant reduction correcting means corrects the reduction of a control constant under the proportional integrating control in accordance with the learning advancing degree by the deciding means. A learning advancing degree reduction correcting means corrects the reduction of the learning advancing degree decided by the learning advancing degree deciding means in accordance with the deviation value between the control center value and the reference value.

Description

[Detailed Description of the Invention] Technical Field> The present invention performs learning to reduce the deviation amount between the control center value of the feedback correction amount determined by proportional-integral control and the reference value, and also adjusts the progress of the learning. The present invention relates to a feedback control device with a learning function in which a control constant in proportional component control is reduced accordingly, with the aim of improving the accuracy of determining the degree of learning progress.

BACKGROUND ART As this type of feed bank control device, there is one used, for example, in air-fuel ratio feedback control of an electronically controlled fuel injection type internal combustion engine.

To explain this, in an electronically controlled fuel injection type internal combustion engine, the fuel injection amount Ti (valve opening pulse width for driving the fuel injection valve) is determined by the following equation.

T'i=TpxCOEF×α+Ts Here, 'rp is the basic injection amount, expressed as Tp=KxQ/N, K is a constant, Q is the intake air flow rate, and N is the engine rotation speed. C0EF is a correction coefficient determined by various operating conditions,
α is a feedbank correction coefficient for air-fuel ratio feedback control (hereinafter referred to as λ control), which will be described later. Ts is a voltage correction amount, which is used to correct changes in the injection amount of the electromagnetic fuel injection valve due to fluctuations in battery voltage.

λ control installs an 02 sensor in the exhaust system to detect the actual air-fuel ratio, determines whether the air-fuel ratio is richer or leaner than the stoichiometric air-fuel ratio based on the slice level, and then adjusts the air-fuel mixture supplied to the cylinder to the stoichiometric air-fuel ratio. This is to control the amount of fuel injected so that be.

Here, the value of the air-fuel ratio feedback correction coefficient α is changed by proportional-integral control ('PI control) to achieve stable control.

In other words, the output of the 02 sensor is compared with the slice level, and the output of the 02 sensor is larger than the slice level.
(small) When the air-fuel ratio is rich (lean), do not suddenly make the air-fuel ratio lean (rich).
゛ (raise) by the integral (
(hereinafter referred to as 1 minute), the air-fuel ratio is controlled to be lean <l<).

However, in a region where λ control is not performed, the desired air-fuel ratio is obtained by clamping to α-1 and setting various correction coefficients COEF.

By the way, the air-fuel ratio obtained when α is set to 1 in the λ control region, that is, the base air-fuel ratio, is the stoichiometric air-fuel ratio (λ = 1
), feedbank control would not be necessary, but in reality, there are variations in components (e.g. air flow meter, fuel injection valve, pressure regulator, control unit), changes over time, and pulse width vs. flow rate of the fuel injector. The base air-fuel ratio deviates from λ = 1 due to factors such as non-linearity of
Feed bank control is performed so that the value becomes 1.

However, the deviation of the base air-fuel ratio from λ-1 differs depending on the operating region. In other words, the value of α to obtain λ-1 differs depending on the operating region, so if this difference is large, the operating region When the air-fuel ratio changes, λ is controlled by feedback control.
It takes time to settle to =1.

As a result, operation is performed at an air-fuel ratio with poor conversion efficiency for the ternary n, which not only increases costs due to the increase in precious metals in the catalyst, but also makes it necessary to replace the catalyst due to further deterioration of conversion efficiency due to deterioration of the catalyst. The problem arises that

Therefore, a learning correction coefficient αl is introduced so that the air-fuel ratio when α-1, that is, the base air-fuel ratio, is λ-1, and this
Learning correction is performed so that l becomes equal to λ=1 when α=1 in each operating state. This reduces the level difference in the base air-fuel ratio due to changes in operating conditions, shortens the period in which the air-fuel ratio deviates from λ-1, and in turn, P in proportional-integral control for determining the value of α A learning control device for the base air-fuel ratio has been devised that makes it possible to reduce the value of /1 minute to improve controllability and thereby reduce the cost of the catalyst.

That is, for example, a map of the learning correction coefficient αl having a value such that the base air-fuel ratio corresponding to the engine operating conditions such as engine speed and load is λ-1 is provided on a data storage means such as a RAM of a microcomputer; Injection amount T
When calculating i, the injection amount 'rp is corrected as shown in the following equation.

Ti = TpXCOEFXαXαl +Ts where α
The learning of l proceeds in the following steps.

i) Detect the engine operating conditions and α in a steady state.

ii) Searching for αl that has been learned and stored up to now in response to the engine operating conditions.

iii) Find the value of αl+Δα/M from this α and αl,
The memory is updated using the result as a new αl.

Note that Δα indicates the deviation amount of the control center value (average value) αC of α that increases or decreases due to PI control from the reference value α1, and Δα
−αC−αl, and the reference value α1 is set to 1.0. Moreover, M is a constant.

By the way, when adopting such a learning control device,
If the P/I component for determining the value of α is made small from the beginning, the transient response will deteriorate before learning has progressed, that is, before the base air-fuel ratio reaches λ-1. Become. On the other hand, learning progresses and the base air-fuel ratio becomes λ-1.
If λ control is performed with a large P/I when , the air-fuel ratio will fluctuate around λ-1, resulting in rotational speed fluctuations and the like due to air-fuel ratio fluctuations.

An invention to avoid these problems has been made by the applicant (Japanese Patent Application No. 1976-076224).
This will be explained using the flowchart shown in the figure.

That is, the learning control is performed in 5101, the updated learning correction coefficient αl is stored in the RAM in 5102, and the number of updates (learning number) C of α1 is counted and amplified in S103. Then, the next time this program is activated, in 8104 of the learning control routine 3101, it is determined whether the number of updates of αl has reached a predetermined value, and if it is greater than or equal to the predetermined value, learning is progressing. (
It is determined that the air-fuel ratio λ is close to 1), and the process proceeds to 5105.
Decrease the value of P/I of PI control for determining . On the other hand, if the number of updates C of αl is less than the predetermined value, it is determined that learning is not progressing, and the reduction process 5105 of P'/1 is bypassed. In other words, the learning correction coefficient αl
At the same time, the value of P/1 is also corrected by learning according to the progress of the control.

In this way, controllability is improved by reducing the value of the air-fuel ratio feedbank correction coefficient α of 271 in PI control according to the degree of progress of learning.

However, while monitoring the learning progress,
The following problems occur in the learning control device that corrects the decrease by /1 minute.

A. The count value of the learning counter will increase over the long term even in operating regions where learning occurs very rarely, but if the progress of learning is slower than the change in the state of the system, it cannot be said that learning is actually occurring. , reducing P/1 minute according to the count value only worsens the transient response.

B. When the state of the system suddenly changes, even if the count value is large, learning is not performed according to the suddenly changed state of the system. Therefore, in this case as well, it is preferable to decrease the P/I amount according to the count value. do not have.

On the other hand, the learning counter is used to monitor the progress of learning, but in reality, the learning progress, that is, the learning count value is small (also known as the amount of deviation between the feedbank control center value and the reference value (the conventional In the case of the example, if Δα) is small, it is the same as learning as a result, and therefore it can be said that P/I can be reduced.

The present invention has been made in view of the conventional situation, and improves the accuracy of the judgment by judging the progress of learning not only according to the number of learning times but also according to the actual feedbank control state, thereby improving control. It is an object of the present invention to provide a feedbank control device with a learning function that can further improve accuracy.

(Structure of the Invention) Therefore, as shown in FIG. 2, the present invention uses proportional-integral control to determine the feed bank correction amount that is added to or subtracted from the fixed control amount in order to make the control value of the controlled object follow the target value. means for correcting and learning the fixed control amount in order to reduce the deviation amount between the control center value of the feed bank correction amount and the reference value; and means for determining the progress of learning according to the number of times of learning by the learning means. and means for decreasing the control constant in the proportional-integral control according to the degree of progress of learning by the determining means. A configuration is provided in which means is provided for reducing the learning progress determined by the learning progress determining means accordingly.

<Example> The present invention will be described below based on an example shown in FIG. Note that this is an application of the present invention to the air-fuel ratio feedback control of the aforementioned electronically controlled fuel injection type internal combustion engine.

That is, in the figure, 1 is the CPU, 2 is the P-ROM, and 3 is the CPU.
CM'03-RAM for learning control, and 4 an address decoder. Note that a bank amplifier power supply circuit is used for the RAM 3 in order to retain the stored contents even after the key switch is turned off.

Analog input signals to the CPUI for fuel injection amount control include an intake air flow rate signal from the hot wire airflow make 5, a throttle opening signal from the throttle sensor 6, a water temperature signal from the water temperature sensor 7, and the 02 sensor. There are an exhaust oxygen concentration signal from 8 and a battery voltage from battery 9, which are inputted via an analog input interface 10 and an A/D converter 11.

12 is an A/D conversion timing controller.

Digital input signals include ON-OFF signals from the idle switch 13, start switch 14, and neutral switch 15, and these are input via the digital input interface 16.

In addition, a reference signal every 180 degrees and a position signal every 1 degree from the crank angle sensor 17 are inputted via a one-shot multi-circuit 18. Further, a vehicle speed signal from a vehicle speed sensor I9 is inputted via a waveform shaping circuit 20.

The output signal (driving pulse signal to the fuel injection valve) from C, P U 1 is sent to the fuel injection valve 22 via the current waveform control circuit 21.

Next, the operation of this system will be explained according to the flowchart shown in FIG.

The basic injection amount Tp (
=KXQ/N).

In S2, a correction coefficient C0EF determined from various operating conditions is set.

In S3, as described later, in 316, the count value C of the learning counter determined by the number of updates of the learning correction coefficient αl and the reduction correction amount X is compared with a predetermined value, and if it is equal to or greater than the predetermined value, control is progressing. After it is determined that P/1 minute for determining α is decreased by a predetermined amount in S4, the process proceeds to S5. On the other hand, if it is less than the predetermined value, without changing the P/I,
Proceed directly to S5. That is, step S4 corresponds to the control constant reduction correction means shown in FIG.

In S5, the output slice level from the 02 sensor 8 is compared and the air-fuel ratio feedback correction coefficient α is set by proportional component control based on the P/1 minute. Process S5 corresponds to the feedback correction amount determining means in FIG.

In S6, a voltage correction numerator s is set based on the mesoteric voltage from the battery 9.

In S7, a learning correction coefficient αl is searched from the engine rotational speed N and the basic injection amount (load) Tp. Note that the map of the learning correction coefficient αl for the rotational speed N and the basic injection amount Tp is stored in the rewritable RAM 3, and is all set to αt=1 at the time when learning has not started. Further, the map has N=8 lattices and T p = about 4 lattices.

S8 to Sll are provided to detect a steady state, and S8 determines a change in vehicle speed based on the signal from the vehicle speed sensor 19, and S9 determines the gear position based on the signal from the neutral switch 15. , SIO determines the change in throttle opening based on the signal from the throttle sensor 6, and S11 determines whether a predetermined time has elapsed. If the predetermined time has elapsed, the process returns to S8. In this way, if the change in vehicle speed is less than a predetermined value within a predetermined time and the gear is engaged,
However, if the change in the throttle opening is less than or equal to a predetermined value, it is determined that the state is steady, and the process proceeds to step S12. At Sl3, the learning correction coefficient αl is corrected.

Also, if any of the determinations from 88 to 310 is NO,
That is, if an unsteady state is determined, 312, which will be described later,
The process proceeds to 816 without performing the learning in 313 and the learning count in 314 and S15.

The correction of the learning correction coefficient αl in step 313 when the steady state is determined is αl−αl as in the conventional method described above.
This is done based on the formula +Δα/M.

In S13, a new learning correction coefficient αl is written to the corresponding engine rotational speed N and basic injection amount Tp in the RAM 3. In other words, the contents in RAM3 are updated. Process 313.
514 corresponds to the learning means in FIG.

In S14, the learning counter decrement correction amount X, which is preset according to the deviation amount Δα, is read from the map stored in the ROM 2. Here, the reduction correction amount X becomes a larger value as the deviation amount Δα becomes larger, and is set so that when the system state suddenly changes and the deviation amount Δα increases by more than a predetermined amount, X becomes a value larger than 1.

In step 315, the count value C of the learning counter is updated as C4-C-X+1 using the X set in Sll.

Process 315. S, 16 corresponds to the learning progress reduction correcting means and the learning progress determining means shown in FIG.

In this way, even if the operating region that is only rarely learned or the state of the system suddenly changes, the learning progress can be monitored in light of the actual system state by correcting the learning count value according to Δα. Since the P/1 minute subtraction correction is performed according to the learning count value, control adapted to the true learning progress status can be performed.

That is, even if the number of updates of the learning correction coefficient αl increases, if the deviation between the base air-fuel ratio and the stoichiometric air-fuel ratio is large and learning is not substantially progressing, X is increased to decrease the learning count value. This stops the P/1 minute reduction correction in S4, keeping the transient response thick (and maintaining good followability to the stoichiometric air-fuel ratio.Conversely, even if the number of updates of the learning correction coefficient α is small, the base air-fuel ratio If the deviation between
'JP/1 minute reduction correction frequency is increased, thereby reducing the amount of deviation around λ-1 of the air-fuel ratio and suppressing rotational speed fluctuations.

Finally, in step 316, the injection amount Ti is calculated using the following equation.

Ti = Tp A drive pulse signal is applied to the fuel injection valve 22 via the current waveform control circuit 21 at a predetermined timing.

<Effects of the Invention> As explained above, according to the present invention, since the learning progress level is determined in consideration of the actual feedback control state, the accuracy of the determination is improved, and the control accuracy can also be improved. can.

[Brief explanation of the drawing]

Fig. 1 is a flowchart showing a conventional example, Fig. 2 is a block diagram showing the configuration of the present invention, Fig. 3 is a hardware configuration diagram of an embodiment of the present invention, and Fig. 4 shows the operation process of the same. FIG. 1...CPU 3...CMO3-RAM for learning control
5...Air flow meter 8...02 sensor 17.
...Crank angle sensor patent applicant: Japan Electronics Co., Ltd. Representative: Fujio Sasashima, patent attorney Figure 1

Claims (1)

[Claims] means for determining, by proportional-integral control, a feedthrough correction amount that is added or subtracted from a fixed control amount in order to cause a control value of a controlled object to follow a target value; and a deviation amount between a control center value of the feedback correction amount and a reference value. means for correcting and learning the fixed control amount in order to reduce the fixed control amount; means for determining the degree of progress of learning according to the number of times of learning by the learning means; and means for learning the proportional integral according to the degree of progress of learning determined by the determining means. In the feedbank control device with a learning function, the learning progress level determined by the learning progress level determining means is reduced in accordance with the deviation amount between the control center value and the reference value. A feedback control device with a learning function, characterized in that a correction means is provided.