JPH09101804A - Method for setting parameter in adaptive control system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To initialize a parameter such as the width of a dead zone and to set an optimum parameter by allowing an input signal to be controlled to pass a filter to obtain a signal and adopting the signal as dead zone width. SOLUTION: A position controlling hydraulic servo device having an electromagnetic proportional solenoid valve and a hydraulic cylinder is adopted as a device to be controlled and the position of the hydraulic cylinder is detected by a position sensor and controlled by adaptive control. In this case, an input signal Ω to be controlled is allowed to pass a filter Df (q) to obtain a signal and the obtained signal is adopted as dead zone width do (k). The dead zone width d(k) is obtained by do (k)=|α<-1> Df (q)Nn (q)| (provided that the α is adaptive gain and the Nn (q) is a molecular polynomial for a transmission function to be controlled which is estimated by an input signal). Thereby the dead zone width can be simply, quickly and suitably set.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a parameter setting method in an adaptive control system, and more specifically, it adopts a dead zone method to adaptively control a controlled object by estimating the dynamic characteristics of the controlled object online. The present invention relates to a method suitable for initial parameter setting in an adaptive control system.

[0002]

2. Description of the Related Art Conventionally, there has been proposed an adaptive control system adopting a dead zone method so as to adaptively control a controlled object by estimating a dynamic characteristic of the controlled object online.
FIG. 1 is a block diagram showing the configuration of the adaptive control system. This adaptive control system inputs the target value ym (k) and outputs the output y (k) of the controlled object 1 to the target value ym (k).
Input signal u (k) for the controlled object 1 to match
And the input / output signals of the controlled object 1 as input, the parameters of the transfer function P (q) of the controlled object 1 (generally, since there are a plurality of parameters, all of these parameters are It has an estimator 2 which outputs a vector display by a parameter vector θ (k), and a designer 3 which sets a coefficient (for example, gain) of a controller 4 with a parameter as an input. The output signal from the controlled object 1 is fed back to the controller 4.

Therefore, the output signal of the controlled object 1 is made to follow the target value by inputting the input / output signal for the controlled object 1 to obtain the parameters by the estimator 2 and setting the coefficient of the controller 4 by the designing device 3. be able to. Here, as an estimation rule applied to the adaptive control system, an estimation rule (dead zone method) according to Formula 1 has been proposed. Here, k is the number of samples, θ (k) is an estimated parameter vector, φ (k) is a signal vector, μ is an adaptive gain, and d
(K) is the disturbance, e (k) {= y (k) −φ ^T (k) θ
(K)} is the equation error, and max [] is the maximum value in [].

[0004]

(Equation 1)

[0005]

The estimation rule of the equation 1 is robust against disturbance when the parameters such as the dead zone width are appropriately set, and the dead zone width becomes larger than necessary. It is possible to prevent the inconvenience that the range in which the output signal of the controlled object cannot follow the target value is expanded. However, the parameters to be initialized include, for example, the adaptive gain as the estimation rule parameter in the estimator 2, the dead band width, and the numerator polynomial of the transfer function representing the controller and the denominator polynomial as the control rule parameter in the controller 4. There are coefficients, and it is extremely difficult to initialize them with proper values.

Therefore, any value is adopted as the parameter to be initialized. In this case, since the adopted value is likely to be far from the proper value, it is impossible to make the output signal of the controlled object follow the target value at the beginning of the adaptive control. Will end up. There is also a disadvantage that the time required from the start of the adaptive control until the output signal of the controlled object can be made to follow the target value becomes long. Furthermore, if a fixedly used parameter, such as the dead band width, is far from an appropriate value, the output signal of the controlled object may greatly vary due to the influence of disturbance (if the dead band width is too small. ), The output signal of the controlled object continues to be held at 0 even if the input signal becomes large to some extent (when the dead band width is too large).

[0007]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and in an adaptive control system that is robust against external disturbances and that can quickly adapt to changes in the controlled object, the parameters such as dead band width It is an object of the present invention to provide a parameter setting method that can automatically achieve the initial setting of and set optimal parameters.

[0008]

According to a first aspect of the present invention, there is provided a parameter setting method for an adaptive control system, wherein an input signal to be controlled is filtered to obtain a signal, and the obtained signal is adopted as a dead band width. The parameter setting method in the adaptive control system according to claim 2 is the dead zone width d0.
(K) is represented by d0 (k) = | α ⁻¹ Df (q) Nn (q) |
{However, α is an adaptive gain, Df (q) is a filter, N
n (q) is a method obtained by the numerator polynomial of the transfer function of the controlled object estimated by the acquired input / output signal}.

A parameter setting method in the adaptive control system according to a third aspect is a method of adjusting a coefficient of the dead zone width by using an input / output signal of a control target which is acquired in advance.
The parameter setting method in the adaptive control system according to claim 4 is a method of setting each parameter of the transfer function of the controller using the filter element Df (q). In the first to fourth aspects, the control target may be a hydraulic device or the like.

[0010]

According to the parameter setting method in the adaptive control system of claim 1, the dead zone method is adopted in the adaptive control system to adaptively control the control target by online estimating the dynamic characteristics of the control target. Then, the input signal to be controlled is filtered to obtain a signal, and the obtained signal is adopted as the dead band width, so the dead band width can be set easily and in a short time, and the dead band width can be set. It can be set appropriately. As a result, it is possible to achieve good estimation of the dynamic characteristics of the control target for adaptive control from the beginning of adaptive control.

According to the parameter setting method in the adaptive control system of claim 2, the dead zone width d0 (k) is changed to d0.
(K) = | α ⁻¹ Df (q) Nn (q) | {where α is an adaptive gain, Df (q) is a filter, and Nn (q) is a control target estimated by the acquired input / output signals. Since the numerator polynomial of the transfer function of is obtained, the same operation as that of claim 1 can be achieved.

According to the parameter setting method in the adaptive control system of claim 3, the dead band width is adjusted more appropriately because the coefficient of the dead band width is adjusted by using the input / output signal of the control target acquired in advance. Can be set.
According to the parameter setting method in the adaptive control system of claim 4, each parameter of the transfer function of the controller is set by using the element Df (q) of the filter. Therefore, each parameter of the transfer function of the controller is simplified. In addition, it can be set in a short time and each parameter can be set appropriately. As a result, good adaptive control can be achieved from the beginning of adaptive control.

This will be described in more detail. Let us assume that the delay operator is q, and consider a single input / output linear system shown in Equations 1 and 2.
However, y (k) and u (k) are outputs and inputs of the controlled object, yt (k) is an output when there is no disturbance, and φ (k) and θ (k) are signal vectors, respectively. Parameter vectors, Nt (q) and Dt (q) are the numerator polynomial and denominator polynomial of the transfer function Pt (q) that is the true control object, and Δt (q) is the transfer function Pt.
It is the uncertainty of (q).

[0014]

(Equation 2)

[0015]

(Equation 3)

Further, the complementary sensitivity function, the sensitivity function, and the feedback controller are expressed by equations 4, 5, and 6, respectively. However, Smn (q), Tmn (q), R (q), X
(Q) and Y (q) are polynomials of q, R (q) is stable (the absolute value of the root is smaller than 1), and Smn
Both (q) and R (q) have no common divisor, and Tmn
Neither (q) nor R (q) has a common divisor.

[0017]

(Equation 4)

[0018]

(Equation 5)

[0019]

(Equation 6)

Here, the configuration of the system of equation 2 is as shown in FIG. 2, and the configuration of the system of equation 3 is as shown in FIG. θu is the convergence value of the parameter vector, Equation 7 is the nominal system after convergence, and Δu
(Q) is the uncertainty of the transfer function Pu (q) that satisfies Eq. 8, θn is the ultimate value of the parameter vector estimated from the previously measured input / output signal Ω, and the control target estimated by adaptive control Has a transfer function of, and Δn
(Q) is the uncertainty of the transfer function Pn (q).

[0021]

(Equation 7)

[0022]

(Equation 8)

[0023]

(Equation 9)

The input signal of Ω has the bandwidth ωb of the target signal, and the polynomial suffixes + and − represent a stable polynomial and an unstable polynomial, respectively. The adaptive gain α and the dead zone element Df (q) are initial setting parameters of the estimation rule, and Tmn ⁺ (q) and Smn ⁺ (q) are initial setting parameters of the control rule. FIG. 4 is a diagram schematically showing a data flow in the case of automatically initializing the initial setting parameters of adaptive control. β is an estimated value difference limit {limit of θ (k) −θ (k−1)} that the engineer must initialize, but β has a clear role, so it can be easily initialized. it can.

Here, the following eight assumptions are made. (A1) The numerator polynomial Nt (m) of the true control target, the orders m and n of the denominator polynomial Dt (n), and the dead time d of the control target are known. (A2) The bandwidth ωb of the target signal is known. (A3) The input / output signal Ω measured in advance is bounded and P. S. The condition is satisfied and the size of Ω is sufficiently large. (A4) The input / output signal (during control) is bounded and P. S.
Meet the conditions. (A5) The polynomials Nt (q) and Dt (q) are time invariant. (A6) Uncertainties Δu (q), Δn (q), Δt
(Q) is stable and proper. (A7) Eq. 10 (In many cases, the range of the coefficient of the transfer function to be controlled is actually known. And if the polynomial with the smallest norm of the transfer function is Nn (q), then Eq.
Is always satisfied. )

[0026]

(Equation 10)

Under these assumptions, the adjustment parameter for adaptive control is initialized using the input / output signal Ω obtained in advance, and the following object is achieved. (O1) The magnitude of the estimated error number 11 is asymptotically reduced to within the dead band width.

[0028]

[Equation 11]

(O2) If the magnitude of the estimation error e (k) is within the dead band width, the number of parameter vectors under estimation is 12
Satisfies the inequality of Eq. Here, β is the limiting rate of Equation 14, and is the only parameter that the technician must initialize. Note that max [] is the maximum value of [].

[0030]

(Equation 12)

[0031]

(Equation 13)

[0032]

[Equation 14]

(O3) Adaptability is improved by constructing a dead zone depending on the input signal. (O4) Obtain the stability condition using the dead zone element Df (q) constructed in O3 based on the small gain theorem. (O5) Construct a control law that satisfies the stability condition derived by O4.

Next, the online least-squares method that achieves O1 {Shinji Shinnaka, Naoki Kawamata, Mikio Sato Research on Discrete-Time Adaptive Algorithms Measurement and Control Vol. 28,
No. 12, pp1082-1097, (198
9) reference} and the dead zone are expressed by Equations 15, 16, and 17 under A1.

[0035]

(Equation 15)

[0036]

(Equation 16)

[0037]

[Equation 17]

Here, the weighting factors λ1 (k) and λ2 (k)
Is 0 <λ1 (k) ≦ 1, 0 ≦ λ2 (k), and F
(K) {F (0) = F ^T (0)> 0} is a gain matrix. Also, the weighting factor λ2 (k) is determined as in Eq. However, α is an adaptive gain, and α> 1.

[0039]

(Equation 18)

Here, the equations 2 and 3 are rewritten as the equation 19, and the dead zone width d0 (k) is set as the equation 20.

[0041]

[Equation 19]

[0042]

(Equation 20)

Further, Equation 21, V (k), and Equation 22 are substituted into Equation 2
It is defined as 3, 3, 24 and 25.

[0044]

(Equation 21)

[0045]

(Equation 22)

[0046]

(Equation 23)

[0047]

(Equation 24)

[0048]

(Equation 25)

From Equations 15 to 18, Equation 23, and Equation 24, Equation 25 can be rewritten as Equation 26.

[0050]

(Equation 26)

By applying the equations 18 to 20 to the equation 26, the equation 27 is obtained.

[0052]

[Equation 27]

Further, by considering V (k) ≧ 0 in Expression 27, Expression 28 is obtained.

[0054]

[Equation 28]

By applying Equation 26 to Equation 28, Equation 2 is obtained.
9 is obtained.

[0056]

(Equation 29)

By applying equation 18 to equation 29, equation 3 is obtained.
0 is obtained.

[0058]

[Equation 30]

By applying Equation 16 to Equation 30, Equation 3 is obtained.
1 is obtained.

[0060]

(Equation 31)

Under the assumption of A4, the following equations (18) and (30) are obtained.

[0062]

(Equation 32)

Therefore, it is understood that the objective O1 can be achieved by the estimation rule under the assumption of A1 and A4. Also,
Equation 16 to Equation 33 are obtained.

[0064]

[Equation 33]

Here, φ (k) is composed of the input / output signal Ω obtained in advance, and c is given by the equation 34.

[0066]

(Equation 34)

From the case of | e (k) | ≧ αd0 (k) in Expression 18, Expression 35 is obtained.

[0068]

(Equation 35)

From the equations (13), (33) and (35), the adaptive gain α for achieving the object O2 satisfies the equation (36).

[0070]

[Equation 36]

However, d1 = max [e (k)]. Therefore, the object O2 can be achieved by selecting α that satisfies Expression 37 or Expression 38.

[0072]

(37)

[0073]

(38)

In order to construct the dead zone, d0
(K) is determined as in Expression 39.

[0075]

[Equation 39]

Here, Df (q) is given by equation 40.

[0077]

(Equation 40)

Further, Dfn (q) and Dfd (q) are given by the equations 41 and 42, respectively.

[0079]

[Equation 41]

[0080]

(Equation 42)

However, Dfn (q) and Dfd (q) must be selected so that the order of Dfn (q) is larger than that of Dfd (q). Based on the input / output signal Ω obtained in advance, the coefficient of Df (q) is obtained by minimizing the expression 43 by the restriction conditions of the expression 20 and the restriction conditions of the expressions 44 and 45.

[0082]

[Equation 43]

[0083]

[Equation 44]

[0084]

[Equation 45]

The constraint condition of the equation 45 is Rough-Hur.
witz stability criterion {Parks, P. C. :
A new proof of the Route-
Hurwitz stability criteri
on using the second method
of Lyapunov. , In Process
ings of the Cambridge Phi
l. Soc. 58: 694-702, (196
2) Reference} is used to represent the inequality.
Therefore, the objective 03 is achieved.

Further, by applying the small gain theorem to the systems of the equations (7) and (8), there is a sufficient condition that the closed loop system achieves the input / output stabilization under the assumption of A6.
Given by

[0087]

[Equation 46]

Here, T (q) is a complementary sensitivity function, and |
Δu ′ (q) | <1 and Δu ′ (q) is stable. From Expressions 7 and 8, Expression 47 is obtained.

[0089]

[Equation 47]

The estimation errors shown in equations 47 to 48 are obtained.

[0091]

[Equation 48]

From Expression 32 and Expression 39, Expression 49 is obtained.

[0093]

[Equation 49]

From Expression 49, Expression 50 is established.

[0095]

[Equation 50]

However, 0 <ρ <1. Equation 51 is obtained from the assumption of A7.

[0097]

(Equation 51)

However, || || ₂ is the square root of the value obtained by squaring and integrating. The number 51 is obtained from the number 51.

[0099]

[Equation 52]

Therefore, the equation 53 is obtained.

[0101]

(Equation 53)

Here, | Δu ′ (q) | <1. Therefore, when the complementary sensitivity function T (q) satisfies the equation 54, the control loop is stable and the objective O4 is achieved by applying the small gain theorem to the system.

[0103]

(Equation 54)

Then, the parameter Tmn of the control law
A procedure for initializing ⁺ (q) and Smn ⁺ (q) will be described. Tmn (q) and Smn (q) are dead zone elements Df
Obtained by using H ∞ suboptimization for (q). Using the nominal system Pn (q), a controller satisfying the stability condition of Expression 54 is designed in advance. Then, the complementary sensitivity function T (q) and the sensitivity function S (q) are set as a standard complementary sensitivity function Tm (q) and a standard sensitivity function Sm (q), respectively. Tm (q) and Sm (q) are obtained by using H∞ suboptimization.

The control law is defined as equations 55 and 56.

[0106]

[Equation 55]

[0107]

[Equation 56]

Here, equations 57 and 58 are stability factors of the numerator polynomial and denominator polynomial of the estimated transfer function, respectively.

[0109]

[Equation 57]

[0110]

[Equation 58]

From the equations (55) and (56), the complementary sensitivity function T (q) and the sensitivity function S (q) based on the control law match Tm (q) and Sm (q) regardless of the variation of the parameter of the controlled object. To do. Therefore, the stability condition of Equation 54 and the objective O5
To achieve.

[0112]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of an adaptive control system in which the parameter setting method of the present invention is implemented. This adaptive control system receives the target value ym (k) as an input and outputs y (k) of the controlled object 1.
To output the input signal u (k) to the controlled object 1 so as to match the target value ym (k) with the controlled object 1
Input / output signal of
The parameter (q) {generally, since there are a plurality of parameters, all of these parameters are displayed as a vector by the parameter vector θ (k)}, and the estimator 2 that outputs the parameters and the controller 4 that receives the parameters as input. And a designer 3 for setting the coefficient (for example, gain) of. The output signal from the controlled object 1 is fed back to the controller 4. Further, the controlled object 1 may be a hydraulic device or the like.

While performing adaptive control, the estimator 2 updates the estimation parameter θ (k) by the equation 16 and the gain matrix F (k) by the equation 17 and sets the dead band width by the equation 39. It is a thing. Further, in performing the initial setting for performing the adaptive control, the input / output signal Ω obtained in advance, the estimated value difference limit β, the order of Df (q), and the bandwidth ωb of the target input ym (k). , Complementary sensitivity T (q)
Adaptive gain α and dead band width d0 (k)
And the transfer function K (q) of the controller 4 = X (q) / Y
Calculate (q). Specifically, the adaptive gain α is
7, the dead zone width d is initialized so as to satisfy Expression 38.
0 (k) is initialized based on the equation 39, and the transfer function K (q) of the controller 4 is initialized based on the equations 55 and 56.

FIG. 5 is a flow chart for explaining the adaptive control method. In step SP1, controlled object 1 to y
(K) is acquired, ul (k) and yl (k) are acquired, and the stability factor numbers 57 and 58 are created from the parameter vector number 12 in step SP2, and the numerator polynomial number 55 and the denominator polynomial number of the controller 4 are obtained. Create the number 56, controller 4
The transfer function number 6 of is constructed. Next, in step SP3, u (k) = K (q) {ym (k) -y (k)}.
(However, ym (k) is a target signal), the input signal u (k) is calculated, and the input signal u (k) is input to the controlled object 1 in step SP4.

Then, in step SP5, the dead band width d0 (k) is set based on the equation 39, the estimation error e (k) is calculated based on the equation 15, and the weighting factor λ2 (k) is calculated based on the equation 18. Then, the estimated parameter vector is set based on Equation 16, and the gain matrix is set based on Equation 17. After that, the process of step SP1 is performed again.

FIG. 6 is a data flow for explaining the parameter setting method of the present invention in detail. In this data flow, the estimated value difference limit β is a value to be set by the engineer, the input / output signal Ω is obtained in advance, and the limit of the complementary sensitivity T (q) is predetermined. And the target input bandwidth ωb is known. The adaptive gain α and the dead zone width d0 (k) are the initial setting parameters of the estimation rule, and the transfer function K (q) of the controller 4 is the initial setting parameter of the control law.

As is clear from this data flow, the signal vector φ (k) and the parameter vector θn are obtained from the input / output signal Ω. And the signal vector φ (k)
And the parameter vector θn, the disturbance d (k) = y
(K) −φ ^T (k) θn is obtained, and the signal vector φ
From (k), c = max [| φ ^T (k) θn |] is obtained. An adaptive gain α is obtained from the estimated value difference limit β and c, and the adaptive gain α is given to the control loop as an initial setting parameter of the estimation rule.

Further, the limit of the complementary sensitivity T (q) and the disturbance d
From (k), the constraint conditions (20), (44), and (45) are obtained. From these constraint conditions and the target input bandwidth ωb, (4) is obtained.
Optimization is performed by minimizing 3 and a filter Df (q) that forms the dead band width d0 (k) is obtained based on the result of this optimization. From this filter Df (q), the dead band width d0 ( k) is obtained. This dead zone width d0 (k)
Is also given to the control loop as an initial setting parameter of the estimation rule.

Further, by using | Df (q) T (q) | <1 as an evaluation, H∞ suboptimization is performed to obtain a complementary sensitivity function T (q) and a sensitivity function S (q). Tm (q) to Tm
(Q) = T (q) and the reference sensitivity function S
m (q) is obtained by Sm (q) = 1-Tm (q). Then, the standard complementary sensitivity function Tm (q) and the standard sensitivity function Sm
A factor Tmn having a stable root of the numerator polynomial of (q)
⁺ (Q), Smn ⁺ (q) are obtained, and these factors Tmn
The transfer function of the controller is initialized by giving ⁺ (q) and Smn ⁺ (q) to the control loop as initialization parameters of the control law.

After the initial setting is performed as described above, adaptive control can be performed based on the estimation rule and the control rule.

[0121]

EXAMPLE A position control hydraulic servo device having an electromagnetic proportional solenoid valve and a hydraulic cylinder is adopted as a control target 1, a position of a hydraulic cylinder is detected by a position sensor (not shown), and a hydraulic pressure is controlled by adaptive control. The position of the cylinder is controlled.

First, after measuring the input / output signal Ω,
7, the adaptive gain α was obtained by the formula 38. The adaptive gain α at this time was 1.033. Also, the number 20 and the number 4
A filter Df (q) was obtained by 3, 3, 44 and 45. The filter Df (q) at this time is 5.8-5.4q.
Met. When the hydraulic servo device is adaptively controlled by initializing the estimation rule with the adaptive gain α = 1.033 and the filter Df (q) = 5.8−5.54q, the estimation error | e (k) | (see solid line) and dead zone width d0 (k) (see broken line) were obtained. Further, as shown in FIG. 8, the difference vector V (k) between the true parameter vector θn and the estimated parameter vector number 12 has changed. The difference vector V (k) is unitless.

Also, the estimation error | e (k) | (see the solid line) and the dead zone width d0 (k) (see the broken line) when the estimation rule is initialized by changing only the adaptive gain α to 1.5 are shown in FIG. 9
10 shows the difference vector V (k). 11 shows the estimation error | e (k) | (see the solid line) and the dead zone width d0 (k) (see the broken line) when the estimation rule is initialized by changing only the adaptive gain α to 1.001.
12 shows the difference vector V (k), respectively. Furthermore, only the filter Df (q) is set to 5.8-4.4.
Estimated error when changing to q and initializing the estimation rule | e
(K) | (see solid line) and dead zone width d0 (k) (see broken line) are shown in FIG. 13, and the difference vector V (k) is shown in FIG. 14, respectively.

Comparing these figures, it can be seen that the estimation error e (k) can be reduced by properly initializing the adaptive gain α and the filter Df (q). Further, the difference vector is also smoothly reduced, and the estimated parameter vector number 12 is the true parameter vector θn.
You can see that is approaching. Also, the adaptive gain α is initialized to 1.033, and the filter Df (q) is set to 5.8.
The position of the hydraulic cylinder when the hydraulic servo device is adaptively controlled by initial setting to -5.4q is as shown by the solid line in FIG. 15, and almost overlaps with the target position (see the broken line).
It can be seen that there is a good match with the target position. Further, the relationship between the dead zone width d0 (k) (see the solid line) and the estimation error e (k) (see the broken line) is as shown in FIG. 16, and the estimation error e (k) is greater than the dead zone width d0 (k). It turns out to be small.

[0125]

According to the invention of claim 1, the dead zone width can be easily
In addition, it can be set in a short time, the dead band width can be set appropriately, and by extension, it is possible to achieve a good estimation of the dynamic characteristics of the control target for adaptive control from the beginning of adaptive control. Has a unique effect.

The invention of claim 2 has the same effect as that of claim 1. According to the invention of claim 3, the dead zone width can be set more appropriately. According to the invention of claim 4, the initial values of the respective parameters of the transfer function of the controller can be set easily and in a short time, and further, the initial values of the respective parameters can be properly set. There is a unique effect that good adaptive control can be achieved from the beginning of control.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing the configuration of an adaptive control system.

FIG. 2 is a diagram showing a configuration of a system of Expression 1.

FIG. 3 is a diagram showing a configuration of a system of Expression 2.

FIG. 4 is a diagram schematically showing a data flow in the case of automatically initializing parameters of adaptive control.

FIG. 5 is a flowchart illustrating adaptive control.

FIG. 6 is a diagram showing in detail a data flow in the case where parameters of adaptive control are automatically initialized.

FIG. 7: Adaptive gain α = 1.033, filter Df
It is a figure which shows the estimation error | error (e) (k) | and dead zone width d0 (k) at the time of carrying out the adaptive control of the hydraulic servo apparatus by setting (q) = 5.8-5.4q.

FIG. 8: Adaptive gain α = 1.033, filter Df
It is a figure which shows the transition of the difference vector V (k) at the time of carrying out the adaptive control of the hydraulic servo device by setting (q) = 5.8-5.4q.

FIG. 9: Adaptive gain α = 1.5, filter Df (q) =
Estimated error | e (k) | and dead band width d0 when the hydraulic servo device is adaptively controlled by setting 5.8-5.4q.
It is a figure which shows (k).

FIG. 10: Adaptive gain α = 1.5, filter Df (q)
7 is a diagram showing a transition of a difference vector V (k) when the hydraulic servo device is adaptively controlled by setting the value to be 5.8-5.4q.

FIG. 11: Adaptive gain α = 1.001, filter Df
It is a figure which shows the estimation error | error (e) (k) | and dead zone width d0 (k) at the time of carrying out the adaptive control of the hydraulic servo apparatus by setting (q) = 5.8-5.4q.

FIG. 12: Adaptive gain α = 1.001, filter Df
It is a figure which shows the transition of the difference vector V (k) at the time of carrying out the adaptive control of the hydraulic servo device by setting (q) = 5.8-5.4q.

FIG. 13: Adaptive gain α = 1.033, filter Df
It is a figure which shows the estimation error | error (e) (k) | and dead zone width d0 (k) at the time of carrying out the adaptive control of the hydraulic servo device by setting (q) = 5.8-4.4q.

FIG. 14: Adaptive gain α = 1.033, filter Df
It is a figure which shows the transition of the difference vector V (k) at the time of carrying out the adaptive control of the hydraulic servo apparatus by setting (q) = 5.8-4.4q.

FIG. 15: Adaptive gain α = 1.033, filter Df
It is a figure which shows an actual cylinder position and target position at the time of carrying out the adaptive control of the hydraulic servo system by setting (q) = 5.8-5.4q.

FIG. 16: Adaptive gain α = 1.033, filter Df
(Q) = 5.8-5.4q and the dead zone width d0 (k) and estimation error | e when the hydraulic servo device is adaptively controlled.
It is a figure which shows (k) |.

[Explanation of symbols]

 1 controlled object 4 controller

Claims (4)

[Claims] 1. An adaptive control system adopting a dead zone method for adaptively controlling a controlled object (1) by online estimation of a dynamic characteristic of the controlled object (1), wherein an input signal of the controlled object (1). To get the signal,
A parameter setting method in an adaptive control system, characterized in that the obtained signal is adopted as a dead band width. 2. The dead zone width d0 (k) is d0 (k) = |
α ⁻¹ Df (q) Nn (q) | {where α is an adaptive gain, Df (q) is a filter, and Nn (q) is the transfer of the controlled object (1) estimated by the acquired input / output signals. The numerator polynomial of the function}. The parameter setting method in the adaptive control system according to claim 1. 3. The parameter setting method in an adaptive control system according to claim 1, wherein the dead band width coefficient is adjusted using the input / output signal of the control target (1) acquired in advance. 4. The parameters of the transfer function of the controller (4) are set by using the element Df (q) of the filter.
A method for setting parameters in the adaptive control system according to claim 1.