JPH05173605A - Gain control device for rotating speed pid controller of internal combustion engine - Google Patents

Abstract

PURPOSE:To perform proper control over a controlled system such as the internal combustion engine for which many evaluation items needs to be set by performing proper gain control matching the thinking of a human being. CONSTITUTION:A storage means 22 stores feature quantities corresponding to respective groups when plural feature quantities are divided into plural groups, a rule group showing the relation of the feature quantities with gain control quantities, and the relation between the value of the feature quantity and the acceptance of the gain adjustment quantity for each group. An antecedent corrector 24 when judging that it is impossible to suppress each feature quantity within a permissible range corrects the rule group so that the permissible range is relaxed. A fuzzy inference part 20 infers gain control quantities of P, I, and D by the rule groups according to the detected feature quantities and rule group to find the acceptance according to the values of the feature quantities and finds the weighted mean value of the inference result based upon the acceptance as weight to determine the gain control quantity of PID. The determined gain control quantity is set in a PID arithmetic part 12 and respective gains of PID are controlled.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a gain adjusting device for a rotational speed PID controller of an internal combustion engine which sets a gain for a PID controller which adjusts the rotational speed of the internal combustion engine.

[0002]

2. Description of the Related Art P which is a kind of feedback control device
In an ID control device, a PID control device that uses knowledge and know-how that humans use to adjust the PID gain, sets a gain adjustment amount by using fuzzy reasoning, and automatically adjusts the gain. It is known (see, for example, JP-A-62-241006 and JP-A-1-258003). In the PID control device using these fuzzy inferences, a rule base including a large number of rules describing knowledge and know-how for gain adjustment is stored in advance.

A plurality of characteristic quantities for evaluating the control state are set for the controlled object controlled by the PID control device. Each of the plurality of rules is created by using the knowledge and know-how that humans use for gain adjustment. For example, when the target arrival time, the overshoot amount, and the vibration damping are set as the feature values, the “IF target value arrival” is set. Time = slow and overshoot amount = large, and vibration damping = bad. Increase the THENP (proportional gain) value to I
The value of (integral gain) is made large and the value of D (differential gain) is made large ”is described by a compound proposition of the IF-THEN-type in which all the feature quantities to be evaluated are connected. Further, as shown in FIGS. 14A to 14D as an example, the PID control device uses the membership function as an antecedent part for evaluating the detected feature amount and obtaining the degree of conformity to each rule. I remember it in shape. As an example, as shown in FIGS. 15A to 15C, the consequent part for weighting the gain adjustment amount indicated by each rule according to the degree of conformity to each rule is stored in the form of a membership function. is doing.

When the gain adjustment amount is actually set by using fuzzy inference, the plurality of feature amounts are detected and, as an example shown in FIG. 16, the degree of conformity of the detected feature amount to each rule is detected. Is calculated using the membership function of the antecedent part. Next, in the membership function of the consequent part, for each rule, a figure obtained by cutting the top of the triangle corresponding to the amount of gain adjustment indicated by each rule with the goodness of fit for each rule was obtained and obtained for each rule. The center of gravity is obtained by superimposing the figures, and the gain adjustment amount is set. This process corresponds to a process of weighting the gain adjustment amount designated by each rule according to the degree of conformity of each rule and calculating an average value. Since the gain adjustment amount set by the above processing reflects the knowledge and know-how used by humans for gain adjustment, the PID to which fuzzy inference is not applied.
As compared with the gain adjusting device, it is possible to perform an appropriate gain adjustment suitable for human thinking.

[0005]

However, the above-mentioned P
The setting of the gain adjustment amount in the ID control device is performed using a rule base in which rules are described in a compound proposition that connects all the feature amounts to be evaluated as described above, and the number of divisions of the evaluation in the antecedent part If m is the number of evaluations such as “slightly large” and “small” for the feature amount of n and the number of feature amounts to be evaluated is n, the number of rule combinations is an enormous number of m to the nth power. For example, when the number of divisions is 4 and the number of feature quantities is 5, the number of rule combinations is 1000 or more.

[0006] In order to finely set the demand for the control state, it is necessary to increase the number of divisions and the number of feature quantities to evaluate the control state. However, if the number of divisions and the number of feature quantities are increased as described above, With the above gain adjustment amount setting method, the number of rule combinations becomes enormous, which makes it difficult to construct rules. Therefore, the conventional PID gain adjusting device based on the fuzzy inference method cannot be applied to the rotational speed control of an internal combustion engine which requires setting a large number of evaluation items.

Further, in the internal combustion engine, the rotation becomes unstable especially in a low rotation range where the output shaft rotates at a relatively low speed, and it becomes impossible to suppress each characteristic amount within an allowable range. When the rotation speed of the internal combustion engine is adjusted by the amount setting method, the gain is changed to the high gain side in an attempt to suppress the fluctuation of the rotation speed in the low rotation range, and the control becomes oscillating rather than controlling, and this vibration is suppressed. As a result, there was a problem that a situation in which the gain did not converge, such as decreasing the gain, occurred next time.

The present invention has been made in consideration of the above facts, and is suitable for an internal combustion engine that requires a large number of evaluation items to be set, and that can perform an appropriate gain adjustment that suits human thinking. An object of the present invention is to obtain a gain adjusting device for a rotational speed PID controller of an internal combustion engine that can perform various controls.

[0009]

In order to achieve the above object, a gain adjusting device for a rotational speed PID controller of an internal combustion engine according to the present invention is based on a deviation and a gain between the rotational speed of the internal combustion engine and a target value. The rotation speed P of the internal combustion engine that sets the gain for the PID controller that controls the operation amount of the rotation speed so that the rotation speed matches the target value.
A gain adjusting device for an ID controller, which corresponds to a feature amount detecting means for detecting a plurality of feature amounts relating to the rotation speed used for inference and each set when the plurality of feature amounts are divided into a plurality of sets. A first storage unit that stores a rule group that represents the relationship between the feature amount and the gain adjustment amount, and a second storage that stores the relationship between the feature amount size and the gain adjustment amount adoption level corresponding to each set. And a correction unit that corrects the relationship between the feature amount and the gain adjustment amount of the rule group so that the allowable range is relaxed when it is determined that each feature amount cannot be suppressed within the allowable range. , Infering a gain adjustment amount for each set based on the detected feature amount and the rule group, and determining an adoption degree of the gain adjustment amount for each set according to the size of the feature amount, Recruitment based on inference result of gain adjustment amount and recruitment Has a inference means for determining an adjustment amount of the gain to be set in the PID controller calculates the weighted average value of the inference result of the gain adjustment amount as a weight, a.

[0010]

In the present invention, the rule group is configured to represent the relationship between the feature amount and the gain adjustment amount corresponding to each set when the plurality of feature amounts are divided into a plurality of sets. For example, when the number of divisions is 4 and the number of feature quantities is 4, conventionally, each rule has 4
It is a compound proposition that connects two features, and the number of rules is 4
⁴ = 256, but if four feature quantities are divided into, for example, two sets according to the present invention to form a rule group, each rule becomes a compound proposition that connects two feature quantities.
The number of rules is 4 ² +4 ² = 32. Further, when the rule group is divided into four groups, each rule becomes a single proposition, and the number of rules is 4 × 4 = 16. Therefore, since the number of rules can be reduced, it becomes easier to construct and adjust the rule group, and especially when applied to a control target that needs to set a large number of evaluation items, the number of rules can be reduced more than before. It can be greatly reduced, and the rule base can be easily constructed and adjusted. In this way, it can be easily applied to a PID controller for controlling the rotation speed of an internal combustion engine, which requires setting a large number of evaluation items.

Further, since each rule group of the rule base can be created by using the knowledge and know-how that humans use for gain adjustment, the rotational speed PI of the internal combustion engine of the present invention.
The amount of gain adjustment set by the gain adjuster for the D controller can be made to reflect human knowledge and know-how, and appropriate gain adjustment suitable for human thinking can be performed as in the conventional case.

Further, the relationship between the size of the feature amount and the adoption degree of the gain adjustment amount for each set is stored, and the adoption amount of the gain adjustment amount for each set is obtained according to the size of the feature amount. Since the weighted average value of the inference result of the gain adjustment amount with the adoption degree as a weight is obtained from the inference result of the gain adjustment amount and the adoption degree, the importance of each inference result can be changed depending on the adoption degree. .. This is when humans make adjustments,
This is equivalent to paying particular attention to the evaluation items that do not meet the requirements and paying little attention to the evaluation items that satisfy the requirements. Therefore, the feature amount is divided into groups so as to correspond to the evaluation items, and when the requirements of a particular evaluation item are not satisfied, the importance of the inference result of the group corresponding to the evaluation item, that is, the adoption degree becomes high. By changing the degree of adoption according to the change in the feature amount, it is possible to perform an appropriate gain adjustment more suited to human thinking.

Further, when it is impossible to suppress the feature amount within the allowable range, the correction means corrects the relationship between the characteristic amount of the rule group and the gain adjustment amount so as to relax the allowable range. Therefore, for example, even if an irremovable rotation speed variation in the low rotation speed range is detected, the rule group is modified so that the allowable range is relaxed to allow this variation. Gain is not changed and proper gain is not set. Therefore, appropriate control can be performed on the internal combustion engine in which there is a characteristic amount that cannot be suppressed within the allowable range.

The number of sets may be plural. However, the number of rules for each set is m ^p , where m is the number of evaluation divisions and p is the number of feature quantities corresponding to each set. Therefore, in order to reduce the number of rules and facilitate the construction and adjustment of the rule group, it is preferable to divide the plurality of feature amounts into as many sets as possible. In particular, when the number of feature quantities is the same as the number of sets, each rule of the rule group becomes a single proposition as described above, and therefore the inference by the inference means becomes easy.

[0015]

【Example】

[First Embodiment] A first embodiment of the present invention will now be described in detail with reference to the drawings. FIG. 1 shows a PID control device 10 with a gain adjuster according to the first embodiment.

The PID controller 10 with a gain adjuster controls the rotational speed of the internal combustion engine 14 as a control target.
The PID controller 10 with the gain adjuster includes the PID calculator 1
Equipped with 2. The PID calculator 12 is connected to the internal combustion engine 14 via the intake air amount adjuster 13. The PID calculator 12 applies a predetermined manipulated variable consisting of P (proportional gain), I (integral gain), and D (differential gain) to the intake air amount controller 13. The intake air amount adjuster 13 is provided with a damper (not shown) arranged in the intake pipe of the internal combustion engine 14, and adjusts the intake air amount of the internal combustion engine 14 by moving the damper according to the applied operation amount. The result of adjusting the intake air amount appears in the rotation speed as the control amount of the internal combustion engine 14. The deviation between the rotation speed and the rotation speed target value is input to the PID calculation unit 12, and the well-known PID control for changing the operation amount according to the input deviation is performed. In addition, PI
An automatic tuning unit 16 corresponding to the gain adjusting device of the present invention is connected to the D calculation unit 12. The PID calculation unit 12 is set with gain adjustment amounts for each of P, I, and D from the auto-tuning unit 16, and the P, I, and D gains are adjusted based on the set gain adjustment amounts.

The auto tuning unit 16 is composed of a microcomputer and the like. Auto tuning unit 16
Includes a feature amount detection unit 18. The operation amount applied to the internal combustion engine 14 via the intake air amount controller 13, the rotational speed of the internal combustion engine 14, the rotational speed target value, and the deviation are input to the characteristic amount detection unit 18. The feature amount detector 18 detects 12 feature amounts from the input information. The feature quantity to be detected is "deviation between target value arrival time and specified value", "overshoot amount", "vibration damping", "overshoot convergence time", "operation fluctuation during settling", "before reaching target value". Control amount fluctuation ”,“ _TP (95% arrival time ÷ 60% arrival time) ”,“ Control amount fluctuation when disturbance occurs ”,“ Control amount fluctuation recovery time when disturbance occurs ”,“ Auto Excited vibration ",
“Vibration of control amount” and “vibration of operation amount”. A fuzzy inference unit 20 is connected to the feature amount detection unit 18, and outputs the detected feature amount to the fuzzy inference unit 20. The fuzzy inference unit 20 is connected to a storage unit 22 that stores an adjustment rule base. The storage unit 22 is composed of, for example, a non-volatile memory or the like.

In this embodiment, in order to evaluate the control state of the internal combustion engine 14, eleven evaluation items, which will be described later, are set, and the twelve feature quantities are divided into eleven sets corresponding to the evaluation items. Has been. In the present embodiment, the adjustment rule base is configured by providing a rule group representing the relationship between the feature amount corresponding to a specific evaluation item and the gain adjustment amount corresponding to each of the 11 evaluation items. .. Each rule group reflects the knowledge and know-how that humans use to adjust the gain. Hereinafter, the evaluation items and the rule group corresponding to the evaluation items in this embodiment will be shown.

Evaluation item 1: Is the deviation between the target value arrival time and the specified value small? Rule group 1: Gain adjustment is performed so as to maintain the target value arrival time at the specified value. IF Target value arrival time is fast enough THEN Gain adjustment is not performed. IF Target value reaching time is a little late. Increase THEN proportional gain by 1%.

Increase the integral gain by 2%. Increase the differential gain by 1%. IF Target value arrival time is slow THEN Proportional gain is increased by 35%.

Increase the integral gain by 35%. Increase the differential gain by 35%. The IF target value arrival time is very slow. Increase THEN proportional gain by 50%.

Increase the integral gain by 50%. Increase the differential gain by 50%.

Evaluation item 2: Is the amount of overshoot small and vibration damping good? Rule group 2: The gain is adjusted so that the overshoot amount is small and the vibration damping is improved. The evaluation item 2 is intended to stabilize the control, and there are two related characteristic quantities, “overshoot quantity” and “vibration damping”. The configurations of the rules of the rule group 2 are shown in Tables 1 to 3 below.

[0024]

[Table 1]

[0025]

[Table 2]

[0026]

[Table 3]

Evaluation item 3: Overshoot convergence time T
Is ₀ short? Rule group 3: Even if the overshoot amount is small, it is not preferable that the control deviation remains indefinitely. Therefore, the gain adjustment is performed so as to shorten the overshoot convergence time T ₀ . IF T ₀ is short enough THEN Gain adjustment is not performed. IF T ₀ is a little long THEN Reduce the differential gain by 5%. IF T ₀ is long THEN Reduce the differential gain by 10%. Very long IF T ₀ THEN Reduce the differential gain by 20%.

Evaluation item 4: Is the change in operation during settling small? Rule group 4: Even if the controlled variable is stable (during settling), the manipulated variable may fluctuate due to noise, etc. Therefore, the gain adjustment is performed so as to reduce the fluctuation in the manipulated variable during settling. The manipulated variable change at IF setting is sufficiently small. THEN Gain adjustment is not performed. The manipulated variable fluctuates slightly during IF setting. THEN Proportional gain adjustment is not performed.

The integral gain adjustment is not performed. Reduce the derivative gain by 7%. The manipulated variable during IF setting is large. THEN Proportional gain is reduced by 2%.

The integral gain is reduced by 2%. Reduce the differential gain by 15%. If the manipulated variable fluctuates during IF setting, the THEN proportional gain is reduced by 4%.

Very large integral gain is reduced by 4%. Reduce the differential gain by 30%.

Evaluation item 5: Is there a change in the control amount before the target value is reached? Rule group 5: If the value of the proportional gain is large, the control amount may return before reaching the target value. Therefore, when this phenomenon occurs, the gain adjustment is performed so that the control amount does not return. Control amount fluctuation before reaching the IF target value does not adjust THEN gain.

The variation of the control amount before reaching the non-IF target value reduces the THEN proportional gain by 15%.

A little larger integral gain is increased by 1%. Reduce the derivative gain by 3%. The control amount fluctuation before reaching the IF target value reduces the THEN proportional gain by 30%.

Large Increase the integral gain by 2%. Reduce the differential gain by 5%. The control amount fluctuation before reaching the IF target value reduces the THEN proportional gain by 40%.

Increase the very large integral gain by 5%. Decrease the differential gain by 10%.

Evaluation item 6: Is T _P (95% arrival time ÷ 60% arrival time) small? Rule group 6: Even if the control amount rises quickly, the change becomes blunt immediately before the target value and the control deviation may remain forever. Therefore, the gain adjustment is performed to improve this. 95% arrival time ÷ 60% arrival time is T _P. IF T _P is sufficiently small. THEN gain adjustment is not performed. IF T _P is a little larger THEN Proportional gain is reduced by 30%.

The integral gain is reduced by 5%. Reduce the differential gain by 30%. IF T _P is large THEN Proportional gain is reduced by 50%.

The integral gain is reduced by 10%. Reduce the differential gain by 50%. IF T _P is very large THEN Proportional gain is reduced by 70%.

The integral gain is reduced by 20%. Reduce the differential gain by 70%.

Evaluation item 7: Is the fluctuation of the control amount small when a disturbance occurs? Rule group 7: The gain adjustment is performed so as to suppress fluctuations in the control amount when disturbance such as load fluctuations occurs. IF The amount of control fluctuation during disturbance is sufficiently small. THEN Gain adjustment is not performed. IF The amount of control fluctuation during disturbance is a little large. Increase the THEN proportional gain by 6%.

Increase the integral gain by 4%. Do not adjust differential gain. IF The amount of control fluctuation during disturbance is large THEN Proportional gain is increased by 12%.

Increase the integral gain by 8%. Increase the derivative gain by 2%. IF fluctuation during disturbance increases THEN proportional gain by 24%.

Increase the very large integral gain by 16%. Increase the derivative gain by 4%.

Evaluation item 8: Is the recovery time T _r of the control amount fluctuation in the case of a disturbance being short? Rule group 8: Gain adjustment is performed so as to shorten the time T _r required for recovery of control amount fluctuation when disturbance such as load fluctuation occurs. IF T _r is sufficiently small THEN Gain adjustment is not performed. IF T _r is a little larger THEN Increase proportional gain by 2%.

Increase the integration gain by 5%. Reduce the differential gain by 8%. IF T _r is large Increase the THEN proportional gain by 4%.

Increase the integral gain by 15%. Do not adjust differential gain. IF T _r is very large Increases THEN proportional gain by 8%.

Increase the integral gain by 30%. Increase the differential gain by 5%.

Evaluation item 9: Is self-excited vibration attenuated? Rule group 9: Since self-excited vibration may occur due to characteristic fluctuation of the controlled object, etc., gain adjustment is performed so as to suppress vibration at the initial stage of self-excited vibration. IF vibration is attenuated. THEN Gain adjustment is not performed. IF vibration is slightly amplified THEN Proportional gain is reduced by 10%. IF vibration is amplified THEN Proportional gain is reduced by 20%. IF vibration is greatly amplified. THEN Proportional gain is reduced by 50%.

Evaluation item 10: Is the vibration of the controlled variable damped? Rule group 10: When the vibration of the control amount remains large and remains forever, the control gain is reduced to avoid danger. IF controlled variable vibration is sufficiently damped. THEN Gain adjustment is not performed. The magnitude of IF controlled variable vibration reduces THEN total gain by 10%.

The magnitude of the remaining IF controlled variable vibration reduces THEN total gain by 20%.

The magnitude of the remaining IF control amount vibration reduces the THEN total gain by 50%.

Remaining evaluation item 11: Is the vibration of the manipulated variable damped?

Rule group 11: When the vibration of the operation amount remains large and remains forever, an operation for reducing the control gain is performed to avoid danger. IF manipulated variable vibration is sufficiently damped. THEN Gain adjustment is not performed. The magnitude of IF manipulated variable vibration reduces THEN total gain by 10%.

The magnitude of the remaining IF manipulated variable vibration reduces the THEN total gain by 20%.

The magnitude of the remaining IF manipulated variable vibration reduces the THEN total gain by 50%.

As described above, since the number of divisions of evaluation is “4” and the number of detected feature amounts is “12” in the present embodiment, the number of rule combinations is 4 in the conventional rule base. ^The number is ¹² = 16777216, and it is almost impossible to build and adjust the rule base. However, in this embodiment, 12 feature quantities are divided into groups corresponding to 11 evaluation items, and a rule group is set for each group, and rules other than the rule group 2 are configured by a single proposition. Since the number of rules is 4 × 10 + 4 ² = 56, rules can be easily constructed and adjusted.

The 11 rule groups include a membership function for inferring the gain adjustment amount for each rule group by obtaining the goodness of fit for each rule of the rule group, and the storage means 22 stores these. I remember. As an example,
2A to 2D show membership functions expressing the rule group 1. Note that (A) is a membership function of the antecedent part, and (B) to (D) are membership functions of the antecedent part, which correspond to a proportional gain, an integral gain, and a differential gain, respectively. The storage unit 22 also stores the relationship between the size of the feature amount and the adoption degree for adopting the inference result of the adjustment amount for each rule group. As shown in FIG. 6, the adoption degree in this embodiment is adapted to be changed according to the change of the feature amount except the adoption degree corresponding to the rule group 2, and the relation is the form of the membership function. Remembered in.

The overshoot amount and the vibration damping which are the characteristic amounts of the rule group 2 are the characteristic amounts relating to the stability of control,
Since the importance is high, these two feature quantities are combined to provide detailed rules as shown in Table 1, Table 2 and Table 3 above. Further, regarding the degree of adoption, the degree of adoption of the rule group 2 is maximized, and adjustment is performed with emphasis on control stability.

The antecedent correction unit 24 is stored in the storage unit 22.
Are connected. In the present embodiment, the rotational speed range in which the characteristic amount of the internal combustion engine 14 cannot be suppressed within the allowable range is observed in advance. In this embodiment, the rotation speed range is 1000 rpm or less. The antecedent part corrector 24 receives the rotational speed as a control amount, and determines whether or not the internal combustion engine 14 is within the rotational speed range based on the input rotational speed. When it is determined that the rotation speed is within the rotation range, the relationship between the feature amount of the rule group and the gain adjustment amount is corrected so as to allow the change in the rotation speed.

Specifically, as shown in FIG. 3, the rule group 7
The membership function of the antecedent part for evaluating the "rotation speed (control amount) fluctuation amount at the time of" is corrected so that the triangle is shifted to the larger side (right side). This increases the weight for the evaluation “good” for the same rotation speed fluctuation amount, and thus the evaluation for the rotation speed fluctuation amount at the time of disturbance becomes unsatisfactory. The same applies to the membership functions in the antecedent part for evaluating the “overshoot amount”, “overshoot convergence time”, and “control amount fluctuation recovery time at disturbance” of rule groups 2, 3, and 8 To fix. Further, the membership function of the antecedent part for evaluating the "vibration damping" of the rule group 2 is corrected as shown in FIG. That is, regarding the triangle corresponding to the evaluation of the membership function as "bad", the lower right vertex of the triangle is fixed in consideration of the relation with the triangle corresponding to the evaluation of "very bad", and the other 2
Modify the vertex to shift it to the right. The triangles corresponding to the evaluations "slightly bad" and "good enough" are corrected so that they are shifted to the right as a whole.

On the other hand, the fuzzy inference unit 20 performs fuzzy inference using the feature amount input from the feature amount detection unit 18, the rule base and the adoption degree stored in the storage unit 22, and sets the gain adjustment amounts P and I. , D for each and output to the PID calculator.

Next, the operation of the first embodiment will be described with reference to the flowchart of FIG. The flowchart of FIG. 5 is executed when the deviation exceeds a specified value.

In step 100, it is determined whether the rotation speed of the internal combustion engine 14 is 1000 rpm or less. If the determination in step 100 is affirmative, in step 102 the antecedent part corrector 24 determines the membership functions of the antecedent parts of the rule groups 2, 3, 7, and 8 as shown in FIGS. Correction is made and the process proceeds to step 104. As a result, the evaluations of “vibration damping”, “overshoot amount”, “overshoot convergence time”, “control amount fluctuation amount during disturbance”, and “control amount fluctuation recovery time during disturbance” are weakened, and the internal combustion engine 1
The deterioration of each of the above-mentioned characteristic amounts due to the occurrence of the fluctuation of the rotation speed of No. 4 is tolerated. When the determination in step 100 is negative, the process proceeds to step 104 without executing the process of step 102.

In step 104, the above-mentioned twelve feature quantities are detected based on various information input to the feature quantity detecting section 18. In the next step 102, the storage means 22
A rule group corresponding to a specific evaluation item (for example, rule group 1 in the case of evaluation item 1) is read from the adjustment rule base stored in. At this time, rule group 2,
For steps 3, 7, and 8, when the process of step 102 is executed, the rule group in which the membership function of the antecedent part is corrected is read. In step 108, a feature amount related to the evaluation item (for example, evaluation item 1
In the case of, "the deviation between the target value arrival time and the specified value"),
Using the membership function of the antecedent part corresponding to the read rule group, the degree of conformity of the feature quantity to each rule is calculated.

As an example, when the value of the feature value “deviation between the target value arrival time and the designated value” is the value indicated by the arrow A in FIG. 2A, the “target value arrival time of the first rule is sufficient. Fast
Is calculated to be 0.25, and the goodness of fit for the second rule, "the target value arrival time is a little slow," is 0.75,
It is determined that the degree of conformity with the third rule and the fourth rule is zero.

In the next step 110, in the membership function of the consequent part, for each rule, a figure obtained by cutting the apex of the triangle corresponding to the gain adjustment amount indicated by each rule by the goodness of fit for each rule, The center of gravity is obtained by superposing the obtained figures. As an example, the adjustment amount of the proportional gain when the goodness of fit to the first rule is 0.25 and the goodness of fit to the second rule is 0.75 will be described with reference to FIG. 2B. The amount of gain adjustment specified by one rule is "No gain adjustment"
Therefore, the figure B obtained by cutting the top of the figure corresponding to “without change” with the matching degree 0.25 for the first rule is obtained.
Regarding the second rule, the gain adjustment amount instructed by the second rule is “to increase the proportional gain by 1%”.
A figure C obtained by cutting the top of the figure corresponding to "1% increase" with a matching degree of 0.75 to the second rule is obtained. Here, figure B
The center of gravity of the figure obtained by superposing the figure C and the figure C is obtained. The center of gravity corresponds to an average value obtained by weighting the gain adjustment amount indicated by each rule according to the suitability of each rule, and is the result of inferring the adjustment amount of the proportional gain by the rule group 1. By performing the above-described processing in the same manner for the adjustment amount of the integral gain and the adjustment amount of the differential gain, the gain adjustment amount for a single evaluation item (group) is inferred.

In step 112, it is determined whether the inference processing has been completed for all the evaluation items. Step 112
If the determination of No is denied, the process returns to Step 102, and Steps 102 to 112 are repeated until the determination of Step 112 is affirmed to read the rule group corresponding to the evaluation item for which the inference processing is not completed, Process as in. Thereby, the gain adjustment amount is inferred for all the evaluation items. In the case where the rotation speed of the internal combustion engine 14 is in the rotation speed range in which the characteristic amount cannot be suppressed within the permissible range, the membership function of the rule group is corrected as described above. The gain adjustment amount inferred in step 1 does not include an adjustment amount for suppressing the irremovable rotation speed fluctuation, and an appropriate gain adjustment amount is inferred.

In step 114, the adoption degree of the inference result of each rule group is calculated using the membership function of the adoption degree and the feature amount stored in the storage unit 22. The membership function of the adoption degree of each rule group except the rule group 2 is set so that the adoption degree becomes high when the control state of the controlled object 14 does not satisfy the requirement of the evaluation item. As a result, a larger weight is given to the inference result of the rule group corresponding to the evaluation item that does not satisfy the request, so that it is possible to perform an appropriate gain adjustment more suitable for human thinking. Note that rule group 2 is closely related to control stability, so considering the control stability, the adoption level is set to "1.0" (maximum value) regardless of changes in the feature values (overshoot amount and damping ratio). ) (See FIG. 6).

In step 116, the inference result for each rule group is weighted using the adoption degree calculated above, and the average value is calculated. As a result, each inference result is treated in parallel, and the gain adjustment amount reflecting a large number of feature amounts is proportional gain,
It is determined for each integral gain and differential gain. Step 11
In 8, the adjustment amount of each gain of the PID determined above is set to PI.
It is output to the D calculation unit 12 and set. Since each rule group of the rule base is created using human knowledge and know-how, the gain adjustment amount reflects human knowledge and know-how. Therefore, each gain is adjusted so that appropriate control suitable for human thought is performed.

In the next step 120, step 102
If the membership function in the antecedent part has been modified by the processing of, the function is restored and restored. As a result, when the rotation speed of the internal combustion engine 14 is higher than 1000 rpm when the routine is next executed, the gain adjustment amount is inferred by the normal rule group and the rotation speed fluctuates. , The gain adjustment amount is determined so as to remove this.

Next, P with a gain adjuster of the first embodiment
The result of actually adjusting the gain of the internal combustion engine 14 by the ID control device 10 will be described. 7A and 7B show the process of auto-tuning when the target value is changed when the internal combustion engine 14 has no load. You can see how the auto-tuning improves the vibrational changes in engine speed. FIGS. 8A and 8B show an auto tuning process when the load applied to the internal combustion engine 14 is changed by the hydraulic pump.
The peak value of the fluctuation of the engine speed when the load is applied (the moment when the load is applied) and when the load is removed (the moment when the applied load is removed) gradually decreases. In this way, the gain adjustment is performed so as to achieve both target value tracking and disturbance suppression by alternately performing tuning by changing the target value and tuning by changing the load. The response after the tuning is shown in FIGS. 9 (A) and 9 (B).

As described above, in the first embodiment, 12 feature quantities are divided into 11 groups corresponding to the evaluation items, and the relationship between the corresponding feature quantity and the gain adjustment quantity is determined by membership function for each group. Since a rule base is configured by setting a rule group expressed by using, and each rule of the rule group other than the rule group 2 is configured by a single proposition, the number of rules can be reduced and a large number of evaluation items can be obtained. Internal combustion engine 1 that needs to be set
It can also be applied to controlled objects such as 4.

Further, the membership function of the adoption degree of each rule group except the rule group 2 becomes high when the control state of the controlled object 14 does not satisfy the requirement of the evaluation item corresponding to the rule group. Since the setting is made as described above, a larger weight is given to the inference result of the rule group corresponding to the evaluation item that does not satisfy the request, and it is possible to perform an appropriate gain adjustment more suitable for human thinking.

Further, in the first embodiment, when the rotation speed of the internal combustion engine 14 is 1000 rpm or less, the membership function of the antecedent part of the rule group is modified so as to relax the allowable range of each feature amount. Therefore, the gain is changed in an attempt to suppress the fluctuation of the rotation speed, and the proper gain is not set. Therefore, for a control target such as the internal combustion engine 14 that needs to set a large number of evaluation items,
More appropriate control can be performed.

In this embodiment, the rule group 2 is composed of detailed rules in consideration of the two feature quantities as shown in Tables 1, 2 and 3 above. However, the rule group 2 is not limited to this. Can be divided into two sets and the number of sets can be made equal to the number of feature quantities. FIG. 17 is a conceptual diagram similar to FIG. 6 in this case.
, And the rules are shown below. In this case, the convergence time may be extended by about 30% depending on the operating conditions, but there is no practical problem.

Evaluation item 2: Is the amount of overshoot small? Rule group 2: Gain adjustment is performed to reduce the amount of overshoot. In addition, the adoption rate of the rule group is always 1.0
Is. IF Overshoot amount is small enough THEN Gain adjustment is not performed. IF Overshoot amount is a little larger THEN Proportional gain is increased by 3%.

The integral gain is reduced by 3%. IF Overshoot amount is large Increase the THEN proportional gain by 16%.

The integral gain is reduced by 6%. IF The amount of overshoot is very large. Increase the THEN proportional gain by 30%.

The integral gain is reduced by 12%. Evaluation item 3: Is vibration damping good?

Rule group 3: Gain adjustment is performed so that vibration damping is improved. The degree of adoption of the rule group is always 1.0. IF Vibration damping is good enough. THEN Gain adjustment is not performed. IF Vibration damping is a little bad THEN Proportional gain is reduced by 3%.

The integration gain is reduced by 3%. Increase the derivative gain by 4%. IF Vibration damping is poor THEN Proportional gain is reduced by 6%.

The integral gain is reduced by 6%. Increase the differential gain by 10%. IF Vibration damping is very bad THEN Proportional gain is reduced by 12%.

The integral gain is reduced by 12%. Increase the differential gain by 20%.

[Second Embodiment] Next, a second embodiment of the present invention will be described. The second embodiment has almost the same configuration as the first embodiment, but after the gain adjustment amount is determined by the fuzzy inference unit 20 as in the first embodiment, the gain adjustment amount and the current gain are adjusted. Calculate the gain and adjust the gain to P
The ID calculation unit 12 is set. The PID calculator 12 determines the operation amount to be applied to the internal combustion engine 14 based on the set gain and the input deviation.

As described above, even when the fuzzy inference unit 20 is configured to calculate the gain and set the gain in the PID calculating unit 12, as in the first embodiment, it is more suitable for human thinking. The effect that the gain can be adjusted and can be applied to the controlled object such as the internal combustion engine 14 that needs to set a large number of evaluation items is obtained. As described above, the gain adjusting device of the present invention is not limited to the one in which the gain adjustment amount is set in the PID controller, but may be one in which the gain is calculated and set as described above.

[Third Embodiment] Next, a third embodiment of the present invention will be described. The same parts as those in the above-mentioned respective embodiments are designated by the same reference numerals, and the description thereof will be omitted.

As shown in FIG. 10, the PID control device with a gain adjuster according to the third embodiment has a PID control unit 11 having a PID calculation unit 12, a feature amount detection unit 18, a fuzzy inference unit 20 and a memory. And an auto tuning unit 16 having means 22. The PID control unit 11 has a communication device 32. The operation amount applied to the internal combustion engine 14, the rotation speed of the internal combustion engine 14, the rotation speed target value, and the deviation between the rotation speed target value and the rotation speed are input to the communication device 32. The communication device 32 transmits the input information to the communication device 31 of the auto tuning unit 16.

The communication device 31 is connected to the characteristic amount detecting unit 18, and supplies the received respective information to the characteristic amount detecting unit 18. The feature amount detection unit 18 detects the feature amount based on the supplied information as in the first embodiment. Further, the communication device 31 is connected to the fuzzy inference unit 20, and the gain adjustment amount for each PID determined by the fuzzy inference unit 20 is input. The communication device 31 transmits the input gain adjustment amount for each PID to the communication device 32 described above. Communication device 3
2 sets the received PID gain adjustment amount in the PID calculator 12. The PID calculator 12 adjusts the gain for each PID according to the set gain adjustment amount, and adds a predetermined operation amount to the internal combustion engine 14.

As described above, the auto tuning unit 16
Even when the PID control unit 11 is configured to set the gain adjustment amount via communication means or the like, as in the first embodiment,
Internal combustion engine 1 that can perform more appropriate gain adjustment more suited to human thinking and needs to set many evaluation items
It is possible to obtain the effect that it can be applied to controlled objects such as 4. As described above, the present invention provides the gain adjusting device and the PI.
The D controller need not be directly connected, but may be connected via a transmission device such as a communication device.

[Fourth Embodiment] Next, a fourth embodiment of the present invention will be described. The fourth embodiment has substantially the same configuration as the third embodiment, but after the gain adjustment amount is determined by the fuzzy inference unit 20 as in the first embodiment, the gain adjustment amount and the current gain are adjusted. Is calculated, and the adjusted gain is output to the communication device 31. The output gain is set in the PID calculator 12 via the communication devices 31 and 32. The PID calculator 12 determines the operation amount based on the set gain and the input deviation.

As described above, even when the fuzzy inference unit 20 calculates the gain and the gain is set in the PID calculation unit 12 via the transmission means such as a communication device, it is more human thought. The effect that suitable gain adjustment can be performed and it can be applied to a controlled object such as the internal combustion engine 14 that needs to set a large number of evaluation items is obtained.

In the above-described embodiment, a rotation range in which each feature amount cannot be suppressed within an allowable range is obtained in advance, and when the rotation speed of the internal combustion engine 14 is within this rotation range, the antecedent correction unit Although the membership function of the antecedent part was corrected by 24, instead of this, the control amount related to the fuel supply is compared with the rotation speed of the internal combustion engine 14 to determine the operation amount related to the fuel supply. When the rotation speed is changing more than the allowable amount even though it is not changed, the membership function of the antecedent part may be modified as described above to allow the change of the rotation speed.

Further, in the above embodiment, the membership function of the antecedent part is corrected in the rotation range where it is impossible to suppress each feature amount within the allowable range. Without modification, the membership function of the consequent part may be modified so that the gain adjustment amount for the evaluation result of the membership function of the antecedent part becomes small.

Further, in the above-mentioned embodiment, the example in which the control target is the internal combustion engine 14 has been described, but the present invention is not limited to this, and it is applied to a control target in which a large number of evaluation items need to be set. Even in this case, appropriate control can be performed. As an example, the result when the present invention is applied to the gain adjustment of the analog computer will be described. FIGS. 11A and 11B show the process of auto-tuning by gain adjustment when the controlled object 14 is a second-order lag system of an analog computer, and the waveform recognition is performed every time the target value is changed. To adjust the gain. It can be seen that the change in the control amount, which was oscillatory, is improved by the auto tuning. In FIGS. 12 (A) and 12 (B), a second-order delay system of an analog computer is controlled,
The gain readjustment process when the response characteristic of the controlled object changes from a fast system to a slow system is shown. It can be seen that although the characteristics have changed and vibrations have occurred in the response of the controlled variable, a good response waveform can be obtained by auto tuning without lowering the response speed. 13 (A) and 13 (B) show an auto-tuning process when a fourth-order delay system of an analog computer is controlled. It can be understood that automatic tuning by gain adjustment is possible even if the order of the controlled object changes. As described above, the present invention has high versatility and can be applied to control of various controlled objects.

Further, although the triangle type membership function is used in the above embodiment, a bell type membership function may be used.

[0097]

As described above, according to the present invention, the relationship between the feature amount corresponding to each set and the gain adjustment amount of the controlled object when the plurality of feature amounts are divided into a plurality of sets by the rule group is shown. , When it is determined that each feature amount cannot be suppressed within the allowable range, the relationship between the feature amount of the rule group and the gain adjustment amount is modified so that the allowable range is relaxed. It is possible to obtain an excellent effect that an appropriate gain adjustment suitable for a thought can be performed, and an appropriate control can be performed on a control target such as an internal combustion engine that needs to set a large number of evaluation items.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram showing a PID control device with a gain adjuster including an auto tuning unit according to a first embodiment.

2A to 2D are conceptual diagrams showing membership functions corresponding to rule group 1. FIG.

[Fig. 3] Members of the antecedent part corresponding to "the amount of control amount fluctuation at the time of disturbance", "overshoot amount", "overshoot convergence time", and "time to recover the amount of control amount fluctuation at the time of disturbance" by the antecedent part modifier It is a conceptual diagram for demonstrating correction of a ship function.

FIG. 4 is a conceptual diagram for explaining the modification of the antecedent part membership function corresponding to “vibration damping” by the antecedent part modifier.

FIG. 5 is a flowchart illustrating the operation of the first embodiment.

FIG. 6 is a conceptual diagram illustrating a process after inferring a gain adjustment amount for each rule group.

7 (A) and 7 (B) are waveform diagrams showing the process of auto-tuning when the target value is changed when the internal combustion engine is unloaded.

8A and 8B are waveform charts showing an auto tuning process when the load applied to the internal combustion engine is changed by the hydraulic pump.

9A and 9B are waveform charts showing the response of the internal combustion engine after the tuning is completed.

FIG. 10 is a schematic configuration diagram showing a PID control device with a gain adjuster including an auto-tuning unit according to a third embodiment.

11A and 11B are waveform diagrams showing an auto tuning process when the present invention is applied to gain adjustment of a second-order delay system of an analog computer.

12 (A) and 12 (B) are diagrams showing a gain readjustment process when the present invention is applied to gain adjustment of a second-order lag system of an analog computer and the response characteristic of a controlled object changes from a fast system to a slow system. It is a waveform diagram shown.

13A and 13B are waveform diagrams showing an auto tuning process when the present invention is applied to gain adjustment of a fourth-order delay system of an analog computer.

14A to 14D are conceptual diagrams for explaining the membership function of the antecedent part in the conventional PID control device to which fuzzy inference is applied.

15A to 15C are conceptual diagrams for explaining the membership function of the consequent part in the conventional PID control device to which fuzzy inference is applied.

FIG. 16 is an explanatory diagram for explaining, as an example of the operation of a conventional PID control device to which fuzzy inference is applied, a process of calculating an adjustment amount of a proportional gain.

17 is a conceptual diagram similar to FIG. 6 when the rule group 2 is divided into two sets.

[Explanation of symbols]

 10 PID control device with gain adjuster 12 PID calculation unit 14 Internal combustion engine 16 Auto tuning unit 18 Feature amount detection unit 20 Fuzzy inference unit 22 Storage means 24 Antecedent correction unit

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification number Reference number within the agency FI Technical display location G05D 13/62 E 7361-3H (72) Inventor Masataka Osawa Nagakute-cho, Aichi-gun, Aichi Prefecture No. 41 1 Toyota Central Research Institute Co., Ltd. (72) Inventor Kota Kota Aichi 2-1, Toyota-cho, Kariya-shi Aichi Prefecture Toyota Industries Corporation (72) Inventor Tomohiro Iwai Toyota-cho, Kariya-shi, Aichi 2-chome 1 Toyota Industries Corp.

Claims (1)

[Claims] 1. A PID controller for controlling a manipulated variable of the rotational speed such that the rotational speed matches the target value based on a deviation and a gain between the rotational speed of the internal combustion engine and the target value. Rotational speed P of internal combustion engine for setting gain
A gain adjusting device for an ID controller, which corresponds to a feature amount detecting means for detecting a plurality of feature amounts relating to the rotation speed used for inference, and each set when the plurality of feature amounts are divided into a plurality of sets. A first storage unit that stores a rule group that represents the relationship between the feature amount and the gain adjustment amount, and a second storage that stores the relationship between the feature amount size and the gain adjustment amount adoption level corresponding to each set. And correction means for correcting the relationship between the feature amount of the rule group and the gain adjustment amount so as to relax the allowable range when it is determined that each of the feature amounts cannot be suppressed within the allowable range. , Inferring the gain adjustment amount for each set based on the detected feature amount and the rule group, and determining the adoption degree of the gain adjustment amount for each set according to the size of the feature amount, Based on the inference result of the gain adjustment amount and the adoption level And a gain adjusting device for calculating a weighted average value of the inference result of the gain adjustment amount with the degree as a weight to determine the gain adjustment amount to be set in the PID controller. ..