JPH0626841B2 - Molded product thickness control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Use) The present invention is applied to film or sheet manufacturing equipment such as extrusion or cast molding, automatic control of die lips, control of moisture content of paper in paper machine, coating. The present invention relates to a molded product thickness control device that can be used for controlling the coating amount thickness of a machine.

(Prior Art) Here, the thickness control of a conventional resin film will be described as an example. FIG. 2 shows a process of manufacturing a general film. First, the extruder 1a melts resin particles by a screw to make them liquid. The melted resin has a small gap in the heat-insulating die 2a.
Extruded from 3a. Since this gap 3a is kept at the same width perpendicularly to the paper surface, a thin plate-shaped resin liquid 4a is discharged from the die 2a.
Flows down. The plate-shaped resin liquid 4a comes into contact with the cooled rotating roller 5a and solidifies to form a thick film 6a, and the film 6a is wound by the winder 9a. The thickness gauge 10 measures the thickness of the film 6a.

By the way, the film 6a is required to have the same predetermined thickness in the width direction, but in reality, it is difficult for the resin liquid to pass through the narrow gap 3a of the die 2a in the width direction at the same speed,
The thickness of the original film 6a is not necessarily the same in the width direction.

Therefore, conventionally, the heater 12a is provided on both sides of the gap 3a of the die 2a.
As shown in FIG. 2 and FIG. 3, a large number are distributed in the width direction in a form of being embedded in the die 2a, and the flow in the width direction in the gap 3a is made uniform. That is, if the heat generated by the heater 12a at a portion where the thickness of the original film 6a is too thick is reduced,
Since the temperature of the resin in contact with the die 2a decreases and the viscous resistance increases, the resin speed at that portion decreases. Therefore, the heater 12
The thickness of the original film 6a at the portion where the heat generated is reduced, and the portion where the film 6a is too thick is corrected. On the contrary, if the thickness of the original film 6a is too thin, by raising the heat generated by the heater 12a at that location, the resin speed at that portion is increased, and the thickness of the original film 6a at that location is increased to correct the thickness. Is done.

Based on the above principle, it has been conventionally considered to automatically control the thickness of the original film 6a by the heater 12a. FIG. 4 is a block diagram showing the principle of thickness control.
The difference between the original thickness measured in 10 and its set value is the controller 1
Input to 3. The controller 13 outputs, for example, a heat generation command for the heater 12a to change the heat generated by the heater 12a. When the heat generated by the heater changes, the resin speed in the die 2a changes,
It is possible to control the thickness by changing the original thickness of the part where the heat generated by the heater is changed.

(Problems to be Solved by the Invention) However, there are the following two reasons that the conventional control system of FIG. 4 cannot perform good control.

(1) After the thickness of the web changes at the die exit,
There is a dead time due to movement of the web or the like from the die exit to the thickness gauge 10 until the thickness gauge 10 detects the change.

(2) There is an interference phenomenon that when the operation end of a certain part of the die lip adjusting mechanism is operated, the thickness of the web or the like corresponding to the adjacent part of the lip adjusting mechanism changes.

As a result, the following problems occur. That is, since the phase delay due to the dead time is large, even if the phase compensation is performed to secure the stability of the control system, the gain of the controller cannot be increased. Therefore, there are problems that the quick response of the control system and the steady-state accuracy of the control system are deteriorated, and that the web thickness is always subjected to the disturbance due to the change of the adjacent die lip adjusting mechanism.

The present invention has been proposed to solve the above conventional problems.

(Means and Actions for Solving Problems) Therefore, the present invention has a die having a mechanism for adjusting the discharge amount of the molten resin along the width direction, and a molded product is provided between the die position and the thickness gauge position. Time L required for the fluid to flow
In an extrusion molding and casting molding apparatus having a thickness gauge for detecting a thickness change, a plurality of die for changing a discharge amount at a predetermined position along the width direction of the molded product in the die. The operation ends 1 to N are attached, and at a discrete time t = t _{k + 1} , a plurality of original fabric thicknesses y ₁ (k + 1), y at positions corresponding to the operation end positions along the molded product width direction.
₂ (k + 1) ... y _N (k + 1) thickness gauge,
Thickness detection value y corresponding to a predetermined position in the width direction of the molded product
_i (k + 1) and the thickness setting value r _i (k +
1) and the difference ε (k + 1) = r _i (k + 1) −
The subtractor that outputs y _i (k + 1) and the thickness difference ε (k + 1) that is the output of the subtractor are time-integrated to obtain an integrated value x _I
An integrator that outputs (k + 1), a memory that stores a time series of past manipulated variables for the dead time L, and a past manipulated variable stored in the memory. Molded product thickness detection value y (k +) at time series and time t _{k + 1}
1) is input and the estimated value of the state variable before the time t _{k + 1} by the dead time L And an output of the integrator x _I (k + 1)
And the output of the observer And a state transition device that outputs a state estimation value at time t _{k + 1} by multiplying by a coefficient for transitioning the state only for the dead time L, and a time series of past operation amount of the operation end stored in the memory. Input and enter time t from time (t _{k + 1} -L)
State change amount I (k +) due to input u (k) up to _{k + 1}
1) is a state predictor, an adder that adds the output of the state transition unit and the output of the state predictor, and outputs a state estimation value at time t _{k + 1} , and the output of the adder. It has an operation amount commander at the operation end that outputs the operation amount command value at the operation end by multiplying the state estimation value at time t _{k + 1} by the state feedback gain, and at the time t = t _{k + 2} , the next original fabric thickness detection value. When y (k + 2) is obtained, the control calculation is performed up to each of the above-mentioned devices to update the operation amount of the operation end, which is used as means and actions for solving the problem. To do.

(Examples) The present invention will be described below with reference to the examples of the drawings. First, in explaining the present invention, five sets of heaters 1 to 5 as shown in FIG. 5 and the measured values of the original thickness 1'to 5'at the position of the thickness gauge 10 corresponding to each heater position are used. Now, consider the problem of controlling the thickness 3'to a predetermined value. In order to control the thickness 3 ', the two heaters 1, 2 and 4, 5 adjacent to each other in addition to the heater 3 are considered because the control considering the interference of the heaters 1, 2 and the heaters 4, 5 with the thickness 3'. This is to design the system. It should be noted that here, the influence on the thickness 3'of the heaters outside the heater 1 and the heater 5 is small and ignored.
The heat generated by the heaters 1 to 5 is respectively u ₁ (t), u ₂ (t) and u
₃ (t), u ₄ (t), u ₅ (t), and the original thicknesses of the thicknesses 1 ′ to ₅ ′ are y ₁ (t), y ₂ (t), y ₃ (t), y _{4 respectively.} (t),
Let y ₅ (t).

If u _i (t) and y _i (t) (i = 1 to 5) are Laplace transforms, u _i (s) and y _i (s) (i = 1 to 5), then u _i (s),
y _i (s) is related by the following transfer function matrix G (s).

g ₁ (s) is a transfer function that gives a temporal change in the thickness 3 ′ when only the heater 3 is changed, for example. Also g ₂ (s)
Is a transfer function which gives the time change of the thickness 3'when only the heater 2 or the heater 4 is changed. g ₃ (s) is heater 1
Alternatively, it is a transfer function that gives the time change of the thickness 3'when only the heater 5 is changed. Since Eq. (1) does not include the dead time due to the flow delay of the material from the die exit to the thickness gauge, g ₁ (s), g ₂ (s), g ₃ (s) are rational of the Laplace operator s. Can be expressed as a function. Further, the non-diagonal terms of the transfer function matrix G (s) of the equation (1) represent the interference with the thickness due to the adjacent heater.

In order to express the relationship between the input ui (s) and the output y _i (s) (i = 1 to 5) of the equation (1), the following state equation convenient for the control system design is used.

(t) = Ax (t) + Bu (t) (2) y (t) = Cx (t) (3) x is a state vector, u is an input vector, and u (t) = [u
₁ (t), u ₂ (t), u ₃ (t), u ₄ (t), u ₅ (t)] ^T (T
Represents transposition), y is an output vector, and y (t) = [y
₁ (t), y ₂ (t), y ₃ (t), y ₄ (t), y ₅ (t)] ^T. The above equations of state (2) and (3) are controllable and observable.

Further, if the dead time due to the flow delay of the material from the die exit to the thickness meter is L, the detection value y (t) at the thickness meter is expressed by the following equation (3).

y (t) = Cx (t−L) (4) From equations (2) and (4), input u (t) (heat generated by the heater) and output y (t)
The relationship of (thickness meter detection value) is shown in FIG.
Now you are ready to design the control system.

First, in order to avoid the influence of the disturbance due to heat conduction from the adjacent heater in controlling the thickness 3'to the preset value of the above-mentioned second problem, the thickness 3'is detected as the disturbance compensation. Deviation ε (t) = between the value y ₃ (t) and the set value r ₃ (t)
Introduce an integrator for r ₃ (t) −y ₃ (t). In the following, the setting r ₃ (t) = 0.

It is assumed that the integrator can integrate up to the current time t of the control deviation ε (t). In reality, since the dead time is L, only the control deviation up to the time (t−L) can be integrated. Also, the output _{I of the} integrator
(t) is expressed by the following equation.

C ₃ represents the third row of the C matrix in equation (3). Also (5)
The first term on the right side of the equation can be calculated because it is the time integral of the amount that can be detected by the actual thickness meter by the time t. However, the second term on the right side cannot be detected because the thickness gauge has a dead time L,
The time integral cannot be calculated as it is. for that reason
To obtain a prediction of _I (t) at time t, consider the following extended system that includes ₁ (t) as a state variable.

From equation (5) From equations (2) and (6) The state vector (t) = [ _I (t), x (t)] ^T of the expansion system is used to express the equation (7) as follows.

The state feedback gain matrix for Eq. (8) =
[F ₁ , F ₂ ], the input u (t) is f ₁ represents the first column of the matrix. Further, if (t) and x (t) are obtained, if the feedback gain matrix is determined so that all eigenvalues of the matrix (−) are in the stable region, the thickness y ₃ (t ) Can be stably controlled to a predetermined value. Moreover, since the influence of the dead time L is not included in the matrix, the feedback gain matrix can be determined as a system without the dead time L in this design method, and the quick response and steady-state accuracy of the control system are sufficient. Performance can be secured. The question is whether _I (t), x (t) can be calculated. If at the current time t
_{If I} (t) and x (t) cannot be obtained, stable control cannot be obtained with the feedback gain matrix described above, and both the quick response and steady-state accuracy of the control system deteriorate and the conventional problems
The solution of (2) cannot be achieved. Therefore, in order to solve the problem (2), the detection signal y of the thickness is determined by the dead time L.
It is necessary to find a means capable of overcoming the adverse effect that (t) can only obtain a value up to the time of y (t-L).

Therefore, the means for actually obtaining _I (t), x (t) in the equation (10) will be described. _I (t), x (t) is calculated as in equation (11) by integrating equation (8) from time (t-L) to time t with time (t-L) in the initial state. To be The idea here is that the input u (t) is known,
By integrating retroactively by the dead time L,
The state quantity _I (t), x (t) is to be estimated.

From the equation (6), _I (t−L) of the first term on the right side of the equation (11) is The right side of the equation (12) is a calculable amount, and the output y ₃ at the current time t
Since it is the integral value of the control deviation of (t), equation (12) is expressed as follows. _I (t−L) = x _I (t) (3) where x _I (t) is the integrated value of the control deviation of the detected value y ₃ (t) of the thickness 3 ′ until time t.

Then x (t-L) can be estimated as follows.

From equations (2) and (4), (t-L) = Ax (t-L) + Bu (t-L) (14) y (t) = Cx (t-L) (15) Defined by the following equation Introduce the variable ω (t)

ω (t) = x (t−L) (16) From the equations (14) to (16), the following equation holds.

(t) = Aω (t) + Bu (t−L) …… (17) y (t) ＝ Cω (t) …… (18) Designing an observer for the equations (17) (18), Estimated value of x (t-L) from detection signal y (t) To get

The thickness gauge measures the thickness of the original fabric while reciprocating along the width direction of the original fabric. Since the original fabric flows at a certain speed, the thickness meter measures the thickness of the original fabric along the locus shown in FIG. Therefore, the thickness at the point position of the original fabric is the time interval T ₁
And a detection signal is obtained every T ₂ . Therefore (1
It is necessary to design the observer by changing equations (18) and (18) to discretized equations.

Integrating equation (17) from time t _o to time t, Here, the time t is the end time of the cycle T ₁ , as shown in FIG. 8, and the time t _o is the time T ₁ before the time t. That is, the following formula is established.

t = t _o + T ₁ (20) Further, the dead time L is assumed to be in the following range here as shown in FIG.

T ₁ <L <T ₁ + T ₂ (21) That is, the following equation is established.

L = T ₁ + T ₂ −m (22) Next, a new variable η is introduced.

η = t−τ (23) (19) (20) (23) The integration on the right side of the equation (24) means that the double line portion in FIG. 9 is integrated. In addition, the input u is provided between the time intervals T ₁ and T _2.
Let (t) be a constant value and take the following discretized quantity.

From time _{t o} during the period from to time t u = u (k) time during the period from _(t o -T ₂₎ to time _{t o u = u (k-} 1) time _(t 0 _-T 2 _{-T 1)} Until time (t _o −T ₂ ), u = u (k−2) Therefore, the equation (24) is expressed as follows.

ω (t) is also a discretized quantity as follows.

Furthermore, the following quantities are defined.

Φ = e ^AT1 (27) From equations (25) to (29), the discretization equation of equation (17) is expressed by the following equation.

ω (k + 1) = Φω (k) + Γ ₁ u (k-2) + Γ ₂ u (k-1) (30) In order to express Eq. (30) as a normal state equation, the following new state Introduce variables.

The new discretized equation of state from Eqs. (30) to (32) is expressed as follows.

Note that I is a 5 × 5 identity matrix. 0 is a zero matrix of appropriate size.

The discretization equation of Eq. (18) is expressed by the following equation.

Equations (33) and (34) are represented by the following equations.

Estimate at t = t _k And Predicted value at t = t _{k + 1} Is as follows from Eq. (34).

Estimated value at t = t _{k + 1} Is calculated from the following formula.

L _o is the feedback gain matrix of the observer. (37)
According to the equation (38), when t = t _{k + 1} , the thickness detection value y
At the same time that _{(k + 1)} is input, the state (k + 1) at t = t _{k + 1} can be estimated. Estimation error at this time Can be expressed by the following equation.

Therefore the matrix If the gain matrix L _o of the observer is determined such that all the eigenvalues of ω are in the stable region, the estimation error can be reduced with the passage of time.

As shown in FIG. 10, when the time t is the end time of the cycle T ₂ , and t _o is a time T ₂ before the time t, ω
(t) is as follows.

The integral on the right side of the equation (40) is to integrate the double line portion in FIG. Therefore The following quantities are defined.

Φ '= e ^AT2 (42) From equations (26), (41) to (44), the discretization equation of equation (17) is as follows.

ω (k + 1) = Φ′ω (k) + Γ ′ ₁ u (k−2) + Γ ′ ₂ u (k−1) (45) The expanded discrete equation of state corresponding to (33) is become that way.

Equation (46) is expressed as follows.

The estimation formula at this time is based on the following two formulas.

Observer gain L _o'is a matrix All eigenvalues of should be determined so that they are in the stable region.

From the above, the estimated value of x (t _{k + 1} −L) at t = t _{k + 1} can be obtained in the following order.

(1) When t = t _{k + 1} is the end time of the cycle T ₁ , (3
From 7) and (38) When t = _{k + 1} is the end time of the period T ₂ ,
From equations (48) and (49) To calculate.

(2) And the estimated value of ω (k + 1), that is, x (t _{k + 1)}
-L) estimate To get

Finally, the integral term of Eq. (11) Is to get. When the current time t = t _{k + 1} is the end time of the period T ₁ , the integral term I means to integrate the double line portion of FIG. Therefore Introduce the following variable η.

From equations (50) to (52) If σ = η−T ₁ , then I (k + 1) is Introducing the following definition formula, From equations (54) to (56), the integral term is obtained as follows.

I (k + 1) = u (k−1) (L−T ₁ ) Φ (T ₁ ) Ψ (L−T ₁ ) + u (k)
T ₁ Ψ (T ₁ ) ... (57) When the current time t = t _{k + 1} is the end time of the period T ₂ , the integral term I (k + 1) means to integrate the double part of FIG. Therefore By exchanging variables as before, Eq. (58) becomes as follows.

I (k + 1) = u (k + 1) (L−T ₂ ) Φ (T ₂ ) Ψ (L−T ₂ ) + u (k)
T ₂ Ψ (T ₂ ) ... (59) (11) (13) (37) (38) (48) (49) (54) (59) From the equation, the current time t =
Estimated value of state quantity [ _I (t), x (t)] ^T at t _{k + 1} Is calculated from the following equation.

The above contents are summarized as follows.

(1) To solve the problem (2), input the deviation ε (t) = r ₃ (t) −y ₃ (t) between the thickness detection value y (t) and the set value r ₃ (t). The integrator that does this is introduced as a controller.

(2) Determine the state feedback gain that can secure the performance with excellent quick response and steady-state accuracy of the control system without being affected by a large position delay due to the dead time L.

(3) State quantity to be fed back [ ₂ (t), x (t)]
Estimated value of ^T Is calculated without being affected by the dead time L from the equation (60).

(4) The problem (1) is solved by the above (2) and (3).

From the above, the calculation procedure of the control method for solving the conventional problems (1) and (2) is summarized as follows. FIG. 1 shows the calculation procedure by imitating the control device.

(1) At time t = t _{k + 1} , a new detection value y of the original thickness is obtained.
(K + 1) (y ₁ (k + 1), y ₂ (k + 1), y
_A vector consisting of ₃ (k + 1), y ₄ (k + 1), y ₅ (k + 1)) is obtained through the thickness meter 10 and the sampler 100.

(2) Of the original fabric thickness detection value y (k + 1), y ₃ (k +
1) is input to the subtractor 101, and the subtractor 101 has a thickness deviation ε (k + 1) = r ₃ (k from the thickness setting value r ₃ (k + 1).
+1) -y ₃ (k + 1) is output.

(3) The integrator 102 calculates the thickness deviation ε from the subtractor 101.
(K + 1) is input and the time integral value of the thickness deviation is output from the following equation.

x ₁ (k + 1) = x ₂ (k) +0.5 (t _{k + 1} −t _k ) {ε (k) + ε (k + 1)}
... (61) where epsilon (k) is thickness deviation at the last thickness detection (time _{t = t k), x I} (k) is the output of the integrator 102 at time t = t _k.

The integrator 102 serves as a disturbance compensator that compensates the disturbance heat that changes the thickness y ₃ with the heat generated by the heater so that the thickness y ₃ is always located at the set value.

From (4) (37) (38) or (48) (49) To calculate. From equation (36) From Ask for. That is, the past heater generated heat time sequence (here, one of u (k)) and the original fabric thickness detection value y (k + 1) stored in the memory 104 are input to the observer 103, and from time t _{k + 1} . Estimated value of state variable before dead time L Is output.

(5) In the calculation of the first term on the right side of the equation (60), the time (t _{k + 1} −
State estimation value in L) Coefficient for changing state by dead time L And the state estimate at time t _{k + 1} To get That is, the output x _I (k + 1) of the integrator 102 and the output of the observer 103 Is input to the state transition unit 105, and the state transition unit 105 multiplies the state transition by the dead time L and multiplies it by time t.
Obtain the state estimate at _{k + 1} .

The state transition due to the input u (k) added in the time domain of the dead time L is represented by I (k + 1) of the second term on the right side of the equation (60), and this correction is performed by the next state predictor 106.

(6) I (k + 1) of the second term on the right side of the equation (60) is the time (t
_{(k + 1-} L) to time tk _{+ 1} represents the state transition amount due to the time series u (k), u (k-1) of the input added to the time domain of the dead time L. I (k + 1) is equation (57) or (59)
Calculate from the formula. That is, the time series of the past heater generated heat (here, two of u (k) and u (k−1)) stored in the memory 104, which is determined by the length of the dead time L, is the state predictor 106. And the state change amount I (k +) from the time (t _{k + 1} −L) to the time t _{k + 1} by the input u (k).
1) is output.

(7) Output of the state transition unit 105 to the adder 107 And the output I (k + 1) of the state change predictor 106 are added, and the state estimated value at time t _{k + 1} Is output. Thus, the time (t
Only the state estimation value at ( _{k + 1−} L) can be obtained by the observer 103, but the state transition unit 105 and the state change predictor 106 perform the integration operation for the dead time L, so that the time t
A state estimate at _{k + 1} can be obtained. This operation can eliminate the influence of the phase delay due to the dead time L.

(8) Heat generated by heater u (k + 1) after time t _{k + 1}
Is determined by the following equation using the state feedback gain (f ₁ , F ₂ ).

That is, the adder 107 calculates the state estimation value at the time t _{k + 1.} Heater generation heat command estimator The heater generated heat command value is determined by multiplying by with the state feedback gain.

(9) The above control calculation is performed each time the next original fabric thickness detection value y (k + 2) is obtained from the sampler 100 at time t = t _{k + 2} .

Next, as the first specific example, the transfer function g ₁ (s), g of the equation (1) is
A design example when ₂ (s) and g ₃ (s) are given by the following equations is shown.

u _i (t) (i = 1 to 5) indicates the amount of change in heat generated by each heater [unit kcal / sec], and y _i (t) (i = 1 to 5) indicates the thickness gauge corresponding to each heater position. Amount of change in thickness at position (unit: cm)
Indicates. The dead time L and the detection periods T ₁ and T ₂ of the thickness gauge assume the following values.

L = 31.5 seconds T ₁ = 16.5 seconds T ₂ = 1
6.5 seconds In order to design the control system, input u (t) and output y in equation (1) are used.
represents the relationship between (t) and is a controllable and observable equation of state
It is necessary to obtain equations (2) and (3). G ₁ (s) in the equations (63) to (65),
G (s) consisting of g _i (s) and g ₃ (s) can be expressed by the 77th-order state equation, but the controllable and observable state equation is 39
I knew it would be next. Therefore, the 39th-order equations of state (2) and (3) were obtained from G (s).

(1). Determining the state feedback gain matrix The state feedback gain matrix in Eq. (10) is based on Eq. (2).
Then, the extended equation of state (7) is solved as a solution of the optimal regulator problem. Since Equation (7) is a discrete-time system state equation, the sampling period T = T ₁ = T
_The regulator solution method was applied in place of the discretized equation of state of ₂ = 16.5 seconds. As a result of obtaining the state feedback gain matrix using an appropriate evaluation function, the matrix (−
The following values were obtained as eigenvalues of).

0.876 ± 0.02i, 0.79, 0.50 ±
0.07i, 0.60 ± 0.09i, 0.60 ± 0.06i,
0.51 Further, the 30 eigenvalues other than the above are not described because their absolute values are as small as 0.1 or less and their attenuation is fast. Since all eigenvalues are within the circle with radius 1, it can be seen that stable control is possible. The slowest decaying eigenvalue is 0.88 ± 0.
Therefore, if the settling time T _s is defined as a control error of 1%, then (0.876) ³⁵ ≈0.01
_s can be predicted to be about 10 minutes as follows.

T _s = T ₁ × 35 = 16.5 × 35 seconds = 577.5 seconds =
9.6 minutes (2). Determination of the observer feedback gain L _o The observer feedback gain matrix L _o in Eq. (38) is 4
It was calculated for the 9th order equation (32) and the 5th order output equation (33). The gain matrix L _o is the matrix Is obtained as a solution of the optimal regulator problem so that has a stable eigenvalue. Gain matrix L _o using an appropriate evaluation function
As a result, the matrix The following values were obtained as the eigenvalues of.

0.9077 ± 0.0002i, 0.9076,
0.9075, 0.9075, 0.722 ± 0.0001i,
0.722, 0.722, 0.722, 0.576 ± 1i × 1
0 ⁻⁵ i, 0.576 ± 1 × 10 ⁻⁵ i, 0.232, 0.
232, 0.232, 0.232, 0.232 Since the 30 eigenvalues other than the above are in the circle of radius 1 concentrated at the origin, the estimation error is reduced with the passage of time. I can go. The estimation error is 1
% Time _{T o} required to decay to, because the slowest eigenvalue is 0.9077, ^{(0.9077) 45} ≒
From 0.01, it can be predicted as follows.

_{T 0 = T 1 × 45 =} 742.5 seconds = 12.4 minutes or more than the obtained gain matrix, by performing a control operation using the _{L o,} showing an example of control results obtained 13 FIG. 13 view (a) shows the time course of change in the amount of five thickness y ₁ ~y ₅ when changing the set value of thickness y ₃ to 0.02mm by stepwise (the amount of change in the detection value of the thickness gauge) . Similarly, FIG. 13 (b) shows the amount of change in the heater-generated heat u _{1 to} u ₅ at that time.

Sampling time 1 of the thickness meter after changing the thickness setting value
Since the thickness detection value is obtained after 6.5 seconds, the change in the heat generated by the heater occurs 16.5 seconds after the thickness setting value is changed. The heater generated heat gives the following thickness detection value: 16.
The same value is connected until 5 seconds later, and when a new thickness detection value is obtained after 16.5 seconds, the heat generated by the heater is also changed. Therefore, the heat generated by the heater changes stepwise as shown in FIG. 13 (b).

On the other hand, as for the thickness detection value, the dead time L = 31.5 after the heater generated heat changes for the first time 16.5 seconds after the set value changes.
Changes are detected after seconds. That is, the change in thickness is detected 16.5 seconds + 31.5 seconds = 48 seconds after the thickness setting value has changed. The thickness y ₃ is correctly changed to the set value, and the settling time is about 10 minutes, which is almost the same as the settling time supported by the eigenvalue. The thickness changes almost symmetrically with respect to the thickness y ₃ . On the other hand, in the heat generated by the heater, the change amount of the heater u ₃ is the largest, the change of the heaters u ₂ and u ₄ is the next largest, and the change of the heaters u ₁ and u ₅ is the smallest.

When actually applying this control method, each thickness y ₁ ,
The same control method as that applied to the thickness y ₃ may be applied to y ₂ , y ₄ , and y ₅ to cancel such an interference effect. It should be noted that, since the time required for the change in thickness to settle when the heater generated heat is changed in steps without being controlled is about 10 minutes, this control method is hardly affected by dead time. I know there isn't.

Next, the second specific example will be described with reference to FIG.
FIG. 4 shows the control result when the turbulent heat of 8.4 w (watt) was added to the heater u ₃ stepwise. Fig. 14
(a) shows the time course of the variation of the thicknesses y _{1 to} y ₅ , and
FIG. 4 (b) shows the change over time in the amounts of heat generated by the heaters u _{1 to} u ₅ . Although Figure 14 the thickness y ₃ by disturbances heat of the heater u _3, as seen in (a) is increased once, by changing the heat generated the heater u ₁ ~u _5, the thickness y ₃ back to its original setting The settling time is about 10 minutes as expected. It can be seen that the introduction of the integrator in the present control method provides good disturbance compensation.

The thicknesses y ₂ and y ₄ are temporarily increased due to the influence of external heat due to heat conduction in the die width direction. The thicknesses y ₁ and y ₅ are also similarly affected, but the influence is naturally smaller than that of y ₂ and y ₄ . In order to cancel the influence of such disturbance heat, the amount of decrease in the heat generated by the heater u ₃ is the largest, and the heaters u ₂ , u
₄ is the next largest decrease, and heaters u ₁ and u ₅ are the smallest. When the disturbance heat is applied to the heater u ₃ , the other thicknesses y ₁ , y ₂ , y ₄ , y ₅ also change, but the thickness y is different for each thickness y ₁ , y ₂ , y ₄ , y ₅ . By applying the same control method as in No. ₃ , such interference effect can be canceled out.

(Effects of the Invention) The present invention is configured as described above in detail, and an integrator that integrates the difference between the thickness detection value and the thickness setting value at a predetermined position of the original thickness is introduced. By feeding back the output to the heater generated heat command value, it is possible to compensate the disturbance heat that changes the thickness with the heater generated heat so that the thickness always matches the set value. Also, in order to avoid a large phase delay due to the dead time L, the observer obtains the state estimation value at the time (t−L) before the current time t by the dead time L, and the state transition device and the state change predictor Dead time L
By performing the time integration by the amount of, it is possible to obtain the state estimation value at the current time t and remove the deterioration of the control performance due to the dead time L.

[Brief description of drawings]

FIG. 1 is a block diagram of a control device showing an embodiment of the present invention,
FIG. 2 is an explanatory view showing the structure of a film manufacturing plant, FIG. 3 is a front view showing an arrangement example of heaters embedded in a die,
FIG. 4 is a block diagram of a conventional original fabric thickness control device, FIG. 5 is an explanatory view showing correspondence between 5 sets of heater positions and 5 sets of thickness detection positions, and FIG. 6 is a dynamic formula model of the original fabric thickness. 7 is a block diagram showing the locus of the thickness gauge for detecting the original thickness, FIG. 8 is a diagram showing the time interval at which the thickness gauge outputs the thickness detection value, FIG. 9, FIG. Figures, 11 and 1
2 is a diagram for explaining the time integration section, FIG. 13 and FIG.
FIG. 4 is an explanatory diagram showing simulation results (when changing the thickness setting value) and (when the heater is subjected to disturbance heat) of the respective embodiments of the present invention. Description of main parts of the drawing 101 ... subtractor, 102 ... integrator 103 ... observer, 104 ... memory 105 ... state transition device, 106 ... state predictor 107 ... adder, 108 ... operation Quantity commander

Claims (1)

[Claims] 1. A die having a mechanism for adjusting a discharge amount of a molten resin along a width direction, and a dead time L which is a time required for a molded product to flow between the die position and a thickness gauge position is set. In the extrusion molding and casting molding apparatus having a thickness gauge for detecting the thickness change, the die has a plurality of operation ends for changing the discharge amount at a predetermined position along the width direction of the molded product. 1 to N are attached, and discrete time t =
At t _{k + 1} , a plurality of original fabric thicknesses y ₁ (k + 1), y ₂ (k +) at positions corresponding to the operation end positions along the width direction of the molded product.
1) ... A thickness gauge for detecting y _N (k + 1) and a thickness detection value y _i (k + 1) corresponding to a predetermined position in the width direction of the molded product.
And the thickness setting value r _i (k + 1) at that position,
The difference ε (k + 1) = r _i (k + 1) −y _i (k + 1)
And the thickness difference ε, which is the output of the subtractor.
(K + 1) performs time integration and outputs an integrated value x _I (k + 1), a memory that stores a time series of past operation amount of the operation end for the length of the dead time L, and the same memory The time series of the operation amount of the past operation end and the time t stored in
_{k + 1} moldings thickness detection value y in (k + 1) is input, the time t _{k + 1} from the dead time L only estimates of the previous state variable And an output of the integrator x _I (k + 1)
And the output of the observer And a state transition device that outputs a state estimation value at time t _{k + 1} by multiplying by a coefficient for transitioning the state only for the dead time L, and a time series of past operation amount of the operation end stored in the memory. Input and enter time t from time (t _{k + 1} -L)
State change amount I (k +) due to input u (k) up to _{k + 1}
1) is a state predictor, an adder that adds the output of the state transition unit and the output of the state predictor, and outputs a state estimation value at time t _{k + 1} , and the output of the adder. It has an operation amount commander at the operation end that outputs the operation amount command value at the operation end by multiplying the state estimation value at time t _{k + 1} by the state feedback gain, and at the time t = t _{k + 2} , the next original fabric thickness detection value. When y (k + 2) is obtained, a control calculation to each of the above-mentioned devices is performed to update the operation amount of the operation end, and the molded product thickness control device is characterized.