JPH01295821A - Apparatus for controlling thickness of film - Google Patents

Abstract

PURPOSE:To always receive the thickness of a film within a set value, by applying timewise integration to the differences between the actual thicknesses of the film detected by a large number of thickness meters arranged in the lateral direction of the film and a set thickness to feed back a correction order and controlling the generation of heat from a heater. CONSTITUTION:The thicknesses of a film are measured at the positions of five sets of heaters 1-5 and at the positions of the thickness meters 10 corresponding to those of the heaters to be introduced into an integrator 102 to apply timewise integration to the differences between the measured values and a set thickness value and the output of the integrator 102 is fed back to the output value of a heater generating heat order device 108 to compensate disturbance heat varying the thickness of the film by the heat generated from the heaters and the thickness of the film is always allowed to coincide with the set value. A state estimate value at time (t-L) before by a dead time L from the present time (t) is obtained by an observation device 103 and timewise integrated by the quantity of the dead time L by a state shifter 105 and a state estimate device 106 to obtain the state estimate value at the present time (t) and the deterioration of control capacity due to the dead time L is removed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Field of Application] The present invention relates to a film thickness control device in extrusion or casting production, such as film production equipment.

(Prior Art) Conventional thickness control of a resin film will be explained with reference to Fig. 2. Fig. 2 shows the process of manufacturing a single-ship film.
First, the extruder 1a melts resin particles using a screw and turns them into a liquid. The molten resin is extruded from the narrow gap 3a of the die 2a, which is kept warm.

Since this gap 3a is maintained at the same width perpendicular to the paper surface, a thin plate-shaped resin liquid 4a flows down from the die 2a. This plate-shaped resin liquid 4a comes into contact with the cooled rotating roller 5a and solidifies to form a thick film 6a, which is wound up by a winding machine 9a. Moreover, the thickness meter IO measures the thickness of the film 6a. The thickness meter IO normally uses radiation from the natural decay of radioactive substances, and measures the thickness of the film 6a based on the degree to which the intensity of the radiation weakens as it passes through the film 6a. There is one detection end, and the thickness of the film 6a in the width direction is measured by reciprocating the detection end along the width direction 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, the thin gap 3 of the die 2a
Since it is difficult to pass the resin liquid through the film 6a at the same speed in the width direction, the thickness of the film 6a is not necessarily the same in the width direction.

For this reason, conventionally, heaters 1 are placed on both sides of the gap 3a of the die 2a.
As shown in FIGS. 2 and 3, a large number of particles 2a are embedded in the die 2a and distributed in the width direction to make the flow in the width direction uniform in the gap 3a.

In other words, the heater 12 at a portion where the film 6a is too thick
When the heat generated by a is lowered, the temperature of the resin in contact with the die 2a is lowered and the viscous resistance is increased, so the speed of the resin in that part is lowered. For this reason, the thickness of the film 6a is reduced at the locations where the heat generated by the heater 12a is reduced, and the locations where the film 6a is too thick are corrected. Conversely, if the thickness of the film 6a is too thin, by increasing the heat generated by the heater 12a at that location, the resin speed at that location will increase, and the thickness of the film 6a at that location will be increased to correct the thickness. .

Conventionally, it has been considered to automatically control the thickness of the film 6a using the heater 12a using the above principle. FIG. 4 is a block diagram showing the principle of thickness control, and the difference between the film thickness measured by the thickness gauge 10 and its set value is input to the controller 13. The controller 13 includes, for example, the heater 12
The generated heat command of heater 12a is outputted to change the generated heat of heater 12a. When the heat generated by the heater changes, the speed of the resin in the die 2a changes, and the thickness of the film changes at the location where the heat generated by the heater is changed, making it possible to control the thickness.

[Problem to be solved by the invention]

The conventional film thickness control device described above had the following problems to be solved.

(1) There is a dead time L1 due to the movement of the film from the die exit to the thickness gauge 10 after a change in film thickness occurs at the die exit until the change is detected by the thickness gauge 10.

(2) There is an interference phenomenon in which when the operating end of a certain part of the die lip adjustment mechanism is operated, the thickness of the film corresponding to an adjacent part of the lip adjustment mechanism changes.

(3) In order to minimize the interference effect on the film thickness due to the mutual interference of the operating ends of the lip adjustment mechanism described in (2) above, a control method may be considered in which the operating amount commands of a large number of operating ends are updated simultaneously.

For this purpose, it is necessary to perform control calculations each time the thickness gauge reciprocates in the film width direction and reaches the end of the film where reading of the entire thickness data along the film width direction is completed. In this case, it takes time for the thickness gauge to measure the thickness of a certain part of the film and for the thickness gauge to reach the end of the film. This time is considered to be a dead time from when the thickness data is obtained to when the control calculation actually starts. Therefore, the dead time between changing the operating amount of the operating end and detecting its influence as thickness data and performing control calculations using that thickness data is the dead time described in (1) above. This is the sum of the dead time L2. Figure 5 (a
) shows a block diagram of the thickness control system including dead time L1 and dead time L8 in FIG. In the feedback control system, there is no such dead time, but in the thickness control system, as shown in FIG. 5(b), a large dead time (L++L weight) is included in the control system.

As a result, the following problem occurs; namely, the phase delay due to dead time is large, so even if phase compensation is performed to ensure stability of the control system, the gain of the controller cannot be increased.

Therefore, there are problems such as poor coordination of the control system and steady-state accuracy of the control system, and there are also problems such as the film thickness is constantly subject to disturbances due to changes in the adjacent die lip adjustment mechanism.

[Means for Solving the Problems] As a means for solving the above problems, the present invention has a die having a mechanism for adjusting the discharge amount of molten resin along the width direction of the film, and has a die position and a thickness gauge position. In film extrusion molding and casting molding equipment, the thickness gauge corresponds to a predetermined position in the film width direction. a subtracter that outputs the difference between the detected thickness value and the thickness setting value at that position, an integrator that performs time integration of the thickness difference output from the subtracter and outputs it, and the dead time L1 described above.
and the time Lt from when the thickness meter reaches the film side edge after detecting the thickness at a predetermined position. an observation device that outputs an estimated value of a state variable a dead time before the time when the stored time series of the manipulated variable of the operating end and the set value of the film thickness detection value were input; a state transition device that inputs the output and the output of the observation device and outputs an estimated state value at a predetermined time by multiplying the output by a coefficient that changes the state by a dead time, and a past discharge amount stored in the memory. a state predictor that inputs a time series of manipulated variables of an operating end and outputs a state change amount according to input settings from a certain time to a time after a dead time period; an adder that adds the outputs and outputs an estimated state value at a predetermined time; and an adder that outputs the estimated state value at a certain time, which is the output of the adder, by a state feedback gain to obtain a manipulated variable command value of the operating end. An object of the present invention is to provide a film thickness control device characterized in that it is equipped with a manipulated variable command device at an operating end that outputs the following.

[For production]

Since the present invention is configured as described above, it has the following effects. In other words, in contrast to the heaters that are arranged in large numbers in the width direction of the film to control the melting temperature of the molten resin that is the raw material for the film, the thickness gauge detects the film at the position corresponding to the heater in the width direction on the downstream side of the film flow. The subtractor outputs the difference between the actual thickness and the set thickness, and the integrator integrates the difference over time to feed back a correction command. This controls the heat generated by the heater, controls the temperature of the molten resin, adjusts its fluidity, and keeps the film thickness within the set value. The phase delay due to the dead time is corrected by estimating the previous state corresponding to the dead time using the observation device, and performing time integration by that amount of time using the state transition device and the state predictor.

〔Example〕

A first embodiment of the present invention will be described with reference to the drawings. First, in explaining the embodiment, we will use five sets of heaters 1 to 5 as shown in FIG. , thickness 3'
Let us consider the problem of controlling to a predetermined value. The reason why we considered the two adjacent heaters 1.2 and 4.5 on both sides in addition to heater 3 to control the thickness 3' is to control the thickness 3' in consideration of the interference of heaters 1, 2 and heater 4.5. This is to set up the system. Note that here, the influence on the thickness 3' of the heaters located outside the heaters 1 and 5 is ignored as it is small. The heat generated by heaters 1 to 5 is expressed as u+(t) and ua(
t), us(t).

ua(t) and us(t), and the thickness of the original fabric from 1' to 5' is V+ (t)lyi(t)Iyi(t)l, respectively.
Let ya(t) and ys(t).

Also Ul(t), y! (t) Laplace transform of (i = 1 to 5) as Ul(3), Y! (!l) (i =
1 to 5), then Ul (ri *Y+(s) is related by the following transfer function matrix G(3).
Hereinafter, the margin G('s) gt(s) is a transfer function that gives a time change in the thickness 3' when only the heater 3 is changed, for example. Also gt(s
) is the thickness 3' when only heater 2 or heater 4 is changed.
It is a transfer function that gives the time change of .

gs(s) is a transfer function that gives a change in thickness 3' over time when only heater 1 or heater 5 is changed. Equation (1) does not include dead time due to film flow delay from the die exit to the thickness gauge, so gt(s)+gx(
s)+gs(s) can be expressed as a rational function of the Laplace operator S. Further, the off-diagonal terms of the transfer function matrix G(s) in equation (1) represent interference with the thickness by adjacent heaters.

Input 1t(s) and output Y of equation (1)! (S) (i −
1 to 5), the following state equation, which is convenient for control system design, is used.

1(t)−＾x(t) +Bu(t) (
2) y (t) = Cx (t) (
3) X is the state vector, U is the input vector, and u(t)
=(u+(t), ut(t), us(t), u<(t)
, us(t) ) ” (T represents transpose), y is the output vector, and y(t)=(y+(t)).

yt(tLys(t), y4(t)+ys(t))
”, the state equations (2) and (3) are controllable and observable.

Also, if the time required for the thickness gauge to move from the thickness measurement point 3' to the edge of the film is L2, the dead time due to film flow delay from the die exit to the thickness gauge is LI%.
As shown in Figure (b), the total dead time on the output vector y side is given by the following equation.

(4) Therefore, the output equation (3) can be expressed as follows.

y (t) = Cx (t - L) (
5) From equations (2) and (5), the relationship between input u(t) (heater generated heat) and output y(t) (thickness gauge detection value) is shown as shown in FIG.

By the way, the thickness gauge measures the film thickness while reciprocating along the film width direction. Since the film is flowing at a certain speed, the thickness meter measures the thickness of the film along a trajectory as shown in FIG. In Fig. 8, the dead time Lm due to the movement of the thickness gauge is shown as the 0 point at the thickness 3'.
can be expressed by the travel time L, l between 00 in FIG.

On the other hand, when the control calculation is performed at the 0 point at the film end, the dead time L2 due to the movement of the thickness gauge is the time taken to move between the points ■'■ in Fig. 8! It can be expressed as #. As can be seen from Figure 8, L
, l and Lt' are generally different in size, so in this control system that controls the thickness 3' to a predetermined value, there are , the dead time shown in FIG. 7 is different.

Therefore, we are ready to design a control system in which the thickness gauge outputs an arrival point identification signal d indicating whether the thickness gauge has reached either side of both ends of the film.

First, in order to avoid the second problem of the conventional method, that is, to avoid being affected by disturbance due to heat conduction from the adjacent heater when controlling the thickness 3' to the set value, the thickness 3' is used as a disturbance compensation.
The deviation ε(
t) −rs(t) Introduce an integrator for yi(t). In the following, the setting is ri(t)-0.

It is assumed that the integrator can integrate the control deviation ε(L) up to the current time t. In reality, due to dead time, time<t-L)
It is possible to integrate only the control deviation up to
+(t) is expressed by the following formula.

Note that C5 represents the third row of the C matrix in equation (3), and the first term on the right side of equation (6) is the time integral of the amount that can actually be obtained from the thickness gauge by time t. Calculable. However, the quantity to be integrated in the second term on the right side cannot be obtained, and the time integral cannot be calculated as ↓, so X + (
Consider the following expanded system that includes X + (t) as a state variable in order to obtain a prediction at time t of t).

From formula (6), X+(t) = -CsX(t-1,)-CiX(t
)+C+X(t-L)=-C+X(t)
(7) (2) (From formula 71, the state of the expanded system [1 hectare X(t) −(L(t), X(
Expressing equation (8) using ``t))'' results in the following.

X(t) =A X(t) +Bu(t)
(9) The state feedback gain matrix for equation (9) is −f+X+(t)−hX(t)
(11) fl represents the first column of the 7 matrix, and X+
(t), if X(t) is obtained, the matrix (A-BF)
If the feedback gain matrix F is determined so that all eigenvalues of are in the stable region, the thickness ys
(t) can be stably controlled to a predetermined value. Moreover, since matrices A and B do not include the effects of dead time, this design method allows the feedback gain matrix F to be determined as if it were a system with no dead time, and the control system has sufficient coordination and steady-state accuracy. performance can be ensured. The question is whether X t (t), X Q) can be calculated. If x+(t), X(t) at current time t
If it cannot be obtained, stable control cannot be obtained with the feedback gain matrix F mentioned above, and both the coordination and steady-state accuracy of the control system will deteriorate, making it impossible to solve the conventional problem (2). . Therefore, in order to solve problem (2), we need a method that can overcome the negative effect caused by the fact that the thickness detection signal y(t) can only obtain the value up to the time y(t-L) due to the dead time. It is necessary to find out. Therefore, we will explain the means to actually obtain the (IQ formula X + (t), X (t).
By integrating the equation (9) from time (t-L) to time t with t-L) as the initial state, it can be obtained as shown in equation 02). The idea here is that since the input u(t) is known, the state quantities X+(t) and X(t) are estimated by integrating backwards by the amount of dead time. It is.

The first term on the right side of Equation 021, X, (t-L), is a quantity that can be calculated from Equation (7), and is the output y3 at the current time.
Since it is the integral value of the control deviation of (t), the measurement is expressed as follows.

X+(L L) =X+(t) 0
41 Here, X+(t) is the detected value yz(t) of thickness 3'
is the integral value of the control deviation up to the specified time.

Next, X (t - L) can be estimated as follows.

(2) From formula (5), X(t-L) =Ax(t-L) +Bu(t-L)
05) y (t) = CxCt-L)
(16) Introduce the variable ω(1) defined by the following equation.

ω(t) −X(t−L) Q7
) The following equation holds true from the aω~q force equation.

Fu(t) =Aω(t) +Bu(t-L)
'08) Design an observer for equation Q8)01 of y(t)=Cω(t)q, and calculate the estimated value X(t-L)(7) from the thickness detection signal y(t). L
)=8ω(1) is obtained.

By the way, since the control calculation is executed every time the thickness gauge reaches 0 point or 0 point as shown in FIG. 8, it is assumed that the control calculation is executed at fixed time intervals T. Therefore, this time interval T
08) It is necessary to change the measurement to a discretized equation and design the observation device.

First, let us describe the relationship between the dead time and the control calculation execution period T. As shown in FIG. 8, it is assumed that the point 2 at the thickness 3' is near the 0 end of the film. When the control calculation is performed at the 0 point at the film edge, the dead time and 1 become large (4
) The total dead time of the equation also increases. On the other hand, when the control calculation is performed at the 0 point at the film end, the dead time L tll becomes small, so the total dead time in equation (4) also becomes small.

In this embodiment, the dead time can be divided into the following two cases, and which case it belongs to can be determined by the arrival point identification signal output when the thickness gauge reaches both ends of the film. .

Case 1. 2T<L<3T Case 2. T<L<2T (1) Case 1. (2T<L<37) As shown in FIG. 9, at time 0 b**, where control calculations are performed at times b-s to jkl, the output vector y(k+1) is obtained as a set of thickness data. Suppose that time t
It is assumed that the input vector u (k) is kept constant as the heater input set during k"'tl+*l.

When the following variable η of Bu(τ−L)dτ [phase] is introduced from the equation (18), the equation (to) is expressed as the equation (21).

η" is -◆1-τ Integration on the right side of equation (21) means integrating the double line portion in FIG. 9. Therefore, equation (21) becomes as follows.

Introduce the following variables Φ, r, rt.

Φ-eA? (23)m
= 37-L (2
6) When the discrete value ω(b) is expressed as ω(k), equation (22) can be expressed as the following equation.

ω(k+1)−Φa+(k)+r+u(k−3)Lrg
u(k-2) (27)0! The discretization formula of the D formula is given by the following formula.

y(k+1)・Cω(k+1)
(28) (27) (
28) By designing an observer for equation 1-1
The estimated value ω(k+1) at and can be found from the following two equations.

Here, K is the feedback gain matrix of the observer.

(29) According to equation (30), the state ω(k+1) at t=tk, 1 can be estimated from the thickness data set y(k÷1) at t=tk, I. Estimation error ω(k
)・ω(k)−ω(k) can be expressed by the following formula.

Therefore, if the gain matrix K of the observer is determined so that all the eigenvalues of the matrix (Φ-KCΦ) are in the stable region, the estimation error can be made smaller over time.

(2) Case 2. (When T<L<2T) As shown in FIG. 10, at time t+1k-1+jk+jk+1
Assume that control calculations are performed in . Integration on the right side of equation (21) means integrating the double line portion in Figure 1θ. At this time, the discretization equation is expressed by the following equation.

a+(k+i)=Φm+(k)+r, u(k-2)+r
zu(k-1) (31)(24) m in the equation (25) is given by the following equation.

m = 2T −L (
32) The estimated value ω(k+1) at L=b−+ can be found from the following two equations.

The equation for the estimation error is the same as equation (31), and in order to reduce the estimation error over time, use case 1. The same can be said.

From the above, 1=1. , state X(tm-+-L)
The estimated value of can be obtained in the following order:

(+) tmtk, + is the end time of the control calculation execution cycle T, and when it is determined that the thickness gauge has reached the 0 point at the film end in Fig. 8 based on the reaching point identification signal output by the thickness gauge. (29) Calculate ω(k+1) from equation (30), and estimate the value of x (th, +-t) X(tb++-L)
=ω(k+1) is obtained.

(2) When tmtk, + is the end time of the control calculation execution cycle T and it is determined that the thickness gauge has reached the 0 point at the film end in Fig. 8 based on the reaching point identification signal of the thickness gauge, then (
33) Calculate ω(k+1) from equation (34) and estimate T
The purpose is to obtain u(τ)dτ. This integral term (2) can be divided into the following two cases depending on the size of the dead time when the input u(t) is applied from time (t-L) to time. It can be determined which case the dead time belongs to by the reaching point identification signal output from the thickness gauge.

Case 1. 3T for 2T≦L Case 2. T<L<2T (1) Case 1. (When 2T≦L<37)
Integral I will integrate the double line portion in FIG.

m=37−L If we introduce the next variable η and rearrange the equation, the integral 1 becomes (37
) is expressed by the formula.

η″″tk, 1−τ (36)(
2) Case 2. (When L<T<2T) Integration I integrates the double line portion in FIG. 12.

(36) By introducing the second variable η and rearranging the previous equation, the integral I can be expressed as the following equation.

(12) (14) (29) (30) (33)
(34) (37) From equation (38), current time L=Lv
−+ Te0) state fit (X, (L), X(t) )
” (7) Estimated value (L (k+1), X(k+1)
) ” Lt is determined by the following formula.

The above contents can be organized as follows.

(1) To solve problem (2), the thickness detection value ys(
t) and the set value rz(t) deviation ε(1)・rs(t)
An integrator with -V3 (t) as input is introduced as a controller.

(2) Determine a state feedback gain matrix F that can ensure excellent performance in coordination and steady-state accuracy of the n-control system without being affected by large phase delays due to dead time.

(3) State quantity to be fed back (XI(t)).

Estimated value of X(t))' (XI(k+1), X(k+1)
)” is calculated using equation (39) without being affected by dead time.

(4) Aim to solve problems (1) and (3) from the above (21 (3)).

Based on the above, the calculation procedure of the control method that solves the conventional problems (1), (2), and (3) can be summarized as follows.

FIG. 1 shows the calculation procedure as a control device.

(]) At the time period control calculation execution time L=tk+1,
Detected value of film thickness y(k+1) (y+(k+1)
,yz(k+1).

A vector consisting of yz(k+1), ya(k+1), and ys(k+1)) is obtained through the thickness meter lO and the sampler 100. The subsumer 100 performs control calculation execution time 1-
1. , , and is closed every time the thickness gauge 10 reaches the film end point 0 or 0 point in FIG. Also, when the thickness gauge 10 reaches the 0 point or 0 point of the film end in FIG. 8, it outputs a reaching point identification signal d indicating which side of the film end point has been reached.

(2) Of the film thickness detection value y(k+1), V
s(k+l) is input to the subtracter 101, and the subtracter 101
is the thickness deviation t (k+l) from the thickness setting value r, (k+1)
)□rs(k÷1)-ys(k+1) is output.

(3) The integrator 102 has a thickness deviation ε from the subtracter 101.
(k+1) is manually calculated and the time integral value of the thickness deviation is output using the following equation.

X+(k+1)□L(k)+0.5(tb,+-tk)
(t (k) + t (k+1) l (40) where ε(k) is the thickness deviation at the previous thickness detection (time 1.1.), and X r (h) is the thickness deviation at time 1=11. Integrator 102
This is the output of

The integrator 102 functions as a disturbance compensator by compensating the disturbance heat that changes the thickness y with the heat generated by the heater so that the thickness y always falls within the set value.

(4) When the thickness gauge reaches either side of the film edge, the thickness gauge outputs the reaching point identification signal d, and depending on the identification signal d, the equation (29), (30) or (33) ( 34)
Calculate ω(k÷1) using the formula. That is, the past heater generated heat time series ((29) (
In equation 30), u(k-, 3) and u(k-2), (33)
In equation (34), u(k-2), u(k-1)) and film thickness detection value y(k+1) are input to the observation device 103, and the reaching point identification signal d(th, + -L)
The estimated state variable value X (th, +-L)-ω(k+
1) Output.

(5) By calculating the first term on the right side of equation (39), the time (tb
, +−L) at the state estimate (X r (k+1) , ω
(k+1))' is overlaid with a coefficient eXL for changing the state by the dead time, and the estimated state value e'' at time 1.1 is calculated.
(X + (k+ 1) + 8ω(k+x))”, that is, the output X+(k+1) of the integrator 102 and the output ω(k+1) of the observer 103.
is input to the state transition device 105, and the state transition device 105 multiplies the state by a coefficient that causes the state to change for the dead time determined by the thickness gauge reaching point identification signal d, and then changes the state to time 1. Obtain the state estimate at l. The size of the dead time varies depending on which side of the film edge the thickness gauge 10 reaches, so the coefficient e″1
also differs depending on the thickness needle position at the time of execution of the control calculation, that is, the arrival point identification signal d of the thickness gauge.

The state transition due to the input u (k) added in the time domain corresponding to the dead time is I (k + 1) in the second term on the right side of equation (39).
This correction is performed by the next state predictor 106.

(6) The second term on the right side of equation (39), I(k÷1), is
The input time series u(k-2), u(
k-1), which represents the state transition amount by u(k), I
(k+1) is calculated from equation (37) or equation (38) depending on which side of the film edge the thickness gage has reached, that is, depending on the reaching point identification signal d output from the thickness gage.

That is, the time series of past heater generated heat determined by the length of the dead time stored in the memory 104 (here, 3 of u(k-2), u(k-1), u(k)) ) is input to the state predictor 106, and the state change amount 1 due to the input u (k) from time (th, +-L) to time t3.1
(k+1) is output.

(7) The adder 107 receives the output e″ of the state transitioner 105.
'(X + (k+1), ω(k+1)')
' and the output 1 (k+1) of the state predictor 106 are added to the state estimated value (X+(k+1), X(k+1),
+1) )". Thus, due to the dead time, only the state estimate at time (tk, 1-L) is the observed value 103.
Although the time be cannot be obtained by
It is possible to obtain a state estimate at +. By this operation, the influence of phase delay due to dead time can be removed.

(8) Time t1. ．． The heat generated by the heater u(k+1) from to the next control calculation time bet is determined by the following equation using the state feedback gain (f+, Pg).

That is, the adder 107 outputs the estimated state value at time t, .

(9) The above control calculation is performed at the next film thickness detection value y(k+2 ) is obtained from the sampler 100 and executed.

Next, as a first concrete example, the number g+ (
s).

A design example will be shown where gz (s) 1g+ (s) is given by the following equation.

u+(t) (i = l~5) indicates the amount of change in heat generated by each heater (unit: Kcal/s), and yi(L) (i = 1~5)
5) shows the amount of thickness change (unit: cm) at the thickness needle position corresponding to each heater position. The dead time due to film flow delay and the time required for the thickness gauge to move from the thickness control point 3' to the end of the film L2' and Lx'' (see FIG. 8) are assumed to have the following values.

Ll=30 seconds Lx'-15 seconds Lt'=1.5 seconds It is assumed that the thickness control point 3' is on the 0 point side of the film end in FIG. The control calculation execution period T assumes the following value.

T=16-. 5 seconds (45) In order to design a control system, we need to express the relationship between the input u(t) and the output y(t) in equation (1), and use the controllable and observable state equation (2) (31 It is necessary to obtain the formula (42) to (4
4) G (s) consisting of g+(s), gz(s), and g+(s) in the equation can be expressed by a 77th-order state equation, but the controllable and observable state equation is 39th-order. I understand. Therefore, from G (s), the 39th order state equation (2)
(Formula 31 was obtained.

(1) Determination of state feedback gain matrix 100 (11)
The state feed bank gain matrix F in the equation was obtained as a solution to the optimal regulator problem for the state equation (8), which is expanded to the 40th order based on the equation (2). Since equation (8) is a state equation for a continuous time system, it can be expressed as the sampling period T
= 16.5 seconds, the regulator solution was applied instead of the discretized state equation. As a result of finding the state feedback gain matrix F using an appropriate evaluation function, the following values were obtained as the eigenvalues of the matrix (A-BF).

0.876±0.02i, 0.79.0.50±0.
07i, 0.60±0.09i, 0.60 +0.
06+, 0.51 and the other 30 eigenvalues are:
The absolute value is small, less than 0.1, and the attenuation is fast, so it is not described. It can be seen that stable control is possible because all the eigenvalues are within a circle with a radius of 1. Since the slowest decaying eigenvalue is 0.88±0.02i, the settling time T is limited.
When defined with an error of 1%, from (0,876)3Sζ0.01, the settling time T1 can be predicted to be approximately 10 minutes as follows.

T-T x35=16.5x35 seconds=577.5 seconds=
9.6 minutes (2) Determination of the feedback gain of the observer The feedback gain matrix of the observer in equation (30) or equation (34) is determined by the 39th order state equation (27)2 or (3).
1) and the fifth-order output equation (28).

The gain matrix K is obtained as a solution to the optimal regulator problem so that the matrix (Φ"-(CΦ)"K"l has stable eigenvalues.) The following values were obtained as the eigenvalues of Φ-KCΦ).

0.9077+0.0002i, 0.90?6.0.
9075.0.9075゜0.722±O,0OO1i
, 0.722.0.722.0.722.0.576
+ I X10-'i, 0.576f= I Xl0
-'i, 0.232.0.232°0.232.0.
232.0.232 The other 30 eigenvalues are concentrated at the origin. Since both are within a circle with a radius of 1, the estimation error can be reduced over time. The time T0 required for the estimation error to decay to 1% of the initial value is (0,9077) since the slowest eigenvalue is 0.9077.
”~0.01, the following prediction can be made.

T, = TX45 = 16.5 x 45 seconds = 742.5 seconds - 1
2. By performing control calculations using the gain matrices F and K obtained for 742 seconds, an example of the control results obtained is shown in the 13th example.
As shown in the figure. In Fig. 13(a), the setting value of thickness y is 0.0.
5 thicknesses y1 when only 2ai is changed into a step shape
Similarly, FIG. 13(b), which shows the change over time in the amount of change in y (the amount of change in the detected value of the thickness gauge), shows the amount of change in the heat generated by the heaters u1 to U at that time.

Execution cycle time of control calculation after changing thickness setting value 16
．． Since the control calculation is performed after 5 seconds, the change in the heat generated by the heater occurs 16.5 seconds after changing the thickness setting value.

Heat generated by the heater remains at the same value until 16.5 seconds after the next control calculation is performed, and after 16.5 seconds, control calculation is performed based on the new detected thickness value and the heat generated by the heater is changed. . Therefore, the heat generated by the heater changes stepwise as shown in FIG. 13(b).

On the other hand, a change in the detected thickness value is detected after a further dead time L=31.5 seconds after the heat generated by the heater changes for the first time 16.5 seconds after the set value changes. That is, the thickness change is detected 16.5 seconds + 31.5 seconds - 48 seconds after the thickness setting value changes. The thickness y has been changed to the correct setting value, and changes almost symmetrically with respect to the thickness y. On the other hand, the amount of change in heat generated by heater U is the largest, and heater 1. The change in u is the next largest, and the change in heater ul+us is the smallest.

The settling time is about 18.5 minutes, which is considerably longer than the settling time of 12.4 minutes, which is based on the eigenvalue of the observer (the settling time based on the eigenvalue of the regulator is much shorter). This is due to the following reason.

In order to prevent the heater generated heat command value from changing significantly at the control calculation time, the heater generated heat command value was determined by adding weights as shown in the following equation.

ua, ni-Wua, *-+ + (I W) um
(46) Here ud+o: Control calculation time t xi
, the heater generated heat command value u4+1I-1'previous control calculation time t-1. Heater generation heat command value uk determined by −1: Control calculation time 1=1. Heater generated heat command value W calculated by: Weighting is a coefficient W-0.8 in this simulation. Considering the control calculation cycle T-16.5 seconds, this means that the time constant is 74.6 seconds.
This means that a time delay element equivalent to a first-order delay of 5 seconds is added to the heater generation heat command device. Therefore, the settling time of the thickness control shown in Fig. 13 is considered to be longer than the settling time estimated from the characteristic value of the observation device.Even if the zero-order thickness control enters the settling state, the heater generated heat command value is changed for each control calculation. It's changing.

This is because the magnitude of the eAL dead time in the first term on the right side of equation (39), which is one of the control calculation equations, is different between the 0 point side and the 0 point side of the film end in FIG.

When actually applying this control method, each thickness y.

The same control equation as applied to thickness y for 1t+ya+Vs can be applied, and each heater generated heat command value can be output as the sum of the results of each control equation. The time required for the change in thickness to settle when the thickness is changed stepwise without changing the thickness is approximately 10 minutes, which shows that the present control method considerably reduces the influence of dead time.

Next, to explain the second specific example with reference to FIG.
Figure 4 shows the control results when a disturbance heat of 8.4 W is applied to the heater uz in a stepwise manner. Figure 14 (a) shows the time course of the amount of change in thickness y1 to (b
) shows the change over time in the amount of heat generated by the heaters u1 to U. As shown in FIG. 14(a), the thickness Vs increases once due to the disturbance heat of the heater U, but By changing the heat, the thickness y returns to the original setting value, and the settling time is 18.5 minutes, which is the same as in Figure 13.The introduction of an integrator using the zero-line control method provides good disturbance compensation. I can see what is being done.

Thickness! ! 1y4 is affected by disturbance heat due to heat conduction in the die width direction, and increases by -1. Thickness! ++ys is also affected in the same way, but the effect is naturally greater than that of Vt*Va, and in order to offset the effect of such disturbance heat, the amount of decrease in heat generated by heater u3 is the largest, and heater uM+u4
The amount of decrease in heater ul is the next largest.

When disturbance heat is applied to heater LI3, where the amount of decrease in us is the smallest, other thicknesses V++Vt+Va+c are also changing, but each thickness νl+V! Thickness y for +V4+VS
By applying the same control equation as , such interference effects can be canceled out.

Next, a second embodiment of the present invention will be described. In the second embodiment as well, the apparatus to be used and equations (1) to (19) are derived, and an observation device for equations (18) and (19) is designed, and the thickness detection signal y (t) is -L) Since it is exactly the same as the example of the estimated value, the explanation will be omitted. Also, in the second embodiment, the control calculation is performed when the thickness gauge reaches the 0 point or 0 point shown in FIG.
Since it is executed every time a point is reached, it is assumed that it is executed at fixed time intervals T. Here, time T is the time required for the thickness gauge to traverse the film width. On the other hand, thickness 3
The dead time for the 0 point is different between the 0 point side and the 0 point side, that is, for the 0 point, it is clear that * > L c,
In the following description, it is assumed that both Lm and Lc satisfy T<L<27.

Figure 15 shows the estimated value ω(tk, +-
We show how to obtain L,). Assume that control calculations are performed from time t2 to [k]. Also, between control calculations t, ~b++
Assume that the heater power u (k) is kept constant.

Time t5. ．． Assume that the thickness gauge reaches 0 point. Therefore, at time i, which is a time T past time that, the thickness gauge would have reached the 0 point. In the control calculation at the 0 point performed at time t, the known ω(tb) is used in the control calculation at the 0 point performed using the estimated value.

Figure 16 shows the estimated value X(b-+-L
Suppose that the thickness gauge reaches 0 point at time tk,l, which shows how to obtain c), and the thickness gauge reaches 0 point at time t□, which is past time T.
The estimated value ω(b) = x (b-Ls
) Δ = X (tm, +-Lc) needs to be found. 1st
As can be seen from Fig. 5 and Fig. 16, the time interval T of control calculation
The time difference is different from the time interval T. Therefore, from equation (18) and (19), the estimated value ω(b-+) is the predicted value X of the state M variable at time t1 from the subsequent inputs as follows.
(t+) is estimated from equation (18) as follows.

t, = t, -LC 1, 1 battle 1 - LI + Lji Lo
``T-LB+L (When new variables η=t, -τ are introduced, the integration on the right side of equation (49) is to integrate the double line part in Figure 15, so equation (49) becomes the following equation. It can be expressed as

ff1l = 2 T L@u(k-1) is the heater input between time tk-1 and tk, o
(k-2) is time tk-! ~tk- is the heater input between. X (t+) = ω (b, +), X (to) = ω
(th), equation (50) can be expressed as the following equation.

Here, the discrete value ω(tk) is expressed by ω(k). Next, consider the control calculation at the 0 point shown in FIG. 16. In this case, Lo□ti+−Li+t+=th , +-Lc,
By integrating the double line part in Figure 16 at t+-to-T-Lc+L1, the predicted value ω(th, +) is obtained from the following formula % formula % Formula (52) The discretization formula of formula (19) is as follows. It is given by Eq.

y(k+1) = Cω(k+1) Therefore, by designing an observer for equations (51), (53), and (54), 1=1. The estimated value ω(tb-+) at , , is determined as follows.

Calculation formula for estimated value ω(k+1) at 0 points (% formula %) K@= Observer gain matrix Calculation formula for estimated value ω (k+1) at 80 points %) K, - Observer gain According to matrix (55) and equation (58), 1=1. From the set of thickness data y(k+1) at , 1.1. The state variable ω(k+1) at , , can be estimated. At this time, the estimation error at the point where the estimation error is 0 is ω(k÷1) ω(k+1) = (ΦwKscΦ1]ω(k)
(59) Estimation error at point 0 ω(k+1) ω(k+1) −(ΦcKccΦ,]ω(k) (
60) Therefore, the matrix [ΦmKscΦ,] [ΦcKcc
Φ. ] If the gain matrices Km and Kc of the observers are determined so that all the eigenvalues are in the stable region, the estimation error can be made smaller over time.

From the above, the estimated value of the state IQX(b-+-L) at L=jl++ can be obtained in the following order.

(1) When tmtm, + is the end time of the control calculation execution cycle T, and it is determined that the thickness gauge has reached the 0 point at the film end in Fig. 8 based on the reaching point identification signal output by the thickness gauge. Calculate ω(k+1) from equations (55) and (56), and calculate the estimated value X(tm-
t-Lm)8=ω(k+1) is obtained.

(2) When tmt, l is the end time of the control calculation execution cycle T, it is determined from the arrival point identification signal of the thickness gauge that the thickness gauge has reached the 0 point at the film end in FIG.

Finally, the integral term on the right side of equation 02) This integral term 1 is from time (t-L) to time t. Integration (2) integrates the double line portion in FIG.

m=27−L If we introduce new variables η=ih, +−τ and organize the previous two, we get
Integral 1(k+1) can be expressed by the following equation.

Since the dead time differs between control calculations at 0 point and 0 point, (
61) The formula is as follows.

Integral value at 0 point 1 m(k+1) Integration at 0 point 1 c(k+1) of.

(12), (14), (55), (56)
, (57), (5B), (62), (
63) It can be found by Eq.

Estimated value at 0 point Estimated value at 0 point However, when the calculation formula is switched between the 0 point side and the 0 point side, such as the estimated value of state ai at time tk+1, the size of the dead time LI + LC is different, so the estimated value The values also become non-continuous and change stepwise each time the calculation formula is switched. When the dead time Lv increases, the estimated value of equation (64) becomes (65
) becomes larger than the estimated value of equation (11), and therefore the manipulated variable according to equation (11) also repeats uneven fluctuations. FIG. 18 shows simulation results when estimated values are calculated using equations (64) and (65). Control calculation interval t
= 22.5 seconds, dead time Ll = 39 seconds, and Lc = 37 seconds. Further, FIG. 18 shows the control results when 8 W of disturbance heat is applied to the heater U in a stepwise manner. Figure 18 (
a) shows the time course of the amount of change in thickness yl to c, and the 18th
Figure (b) shows the heat U generated by the heater.

As shown in FIG. 18(a), which shows the time course of the change in ~U, the thickness I/ increases by -1 times due to disturbance heat, but by changing the heat generated by the heaters u1~U, the thickness I/ 8 has returned to its original setting. However, even in a steady state, the heat generated by each heater repeatedly fluctuates, and the thickness also repeats small changes. If the position of 5/3 of the thickness is closer to the edge of the film, there will be dead time.

Since the difference between the two becomes larger, if the estimation calculation formulas (64) and (65) are used, the fluctuation range of the heat generated by the heater becomes larger, and the fluctuation of the thickness also becomes more noticeable. In order to improve such a problem, the dead time L used for (64), (65) 2, 7Ls and Lc (7) average value L = 'A (t
, a+t, c) was used. Therefore, even on the 0 point side, it is 0
The following equation is commonly used to calculate the estimated value on the point side.

L = 'A (t, m+t, c)
(68) In Figure 19, the calculation formula for the estimated value is (66),
(67) This is a simulation result when using equations (6B), and the conditions are the same as in FIG. 18. Fluctuations in heater heat generation in steady state are resolved. The above contents are summarized as follows.

(1) To solve problem (2), thickness detection value y: +
Deviation between (t) and set value rs(t) ε(1)・r+(t)
An integrator with -ys(t) as input is introduced as a controller.

(2) Determine a state feedback gain matrix F that can ensure excellent performance in control system coordination and steady-state accuracy without being affected by large phase delays due to dead time.

(3) State 1 to be fed back (X + (
t).

x(t)] Estimated value of T (X+(k+1), X(k+1)
) ' C Calculate without being affected by dead time using equation (66).

(4) Aim to solve problems (1) and (3) from (2) and (3) above.

Based on the above, the calculation procedure of the control method that solves the conventional problems (1), (2), and (3) can be summarized as follows.

FIG. 1 shows the calculation procedure as a control device.

(At the control calculation execution time 1.1 of the 11-hour period T,...
Detected value of film thickness y(k+1) (y+(k+1)
,yg(k+1).

ys (k+1), ya (k+1), ys (k+
1) is the thickness meter 10 and the sampler 10
Obtained through 0. The sampler 100 performs control calculation execution time 1=1. , , and is closed every time the thickness gauge 10 reaches the film end point 0 or 0 point in FIG. When the thickness meter IO reaches the film end point 0 or 0 point in FIG. 8, it outputs a reaching point identification signal d which indicates which side of the film edge it has reached.

(2) Of the film thickness detection value y(k+1), y
s(k+1) is input to the subtracter 101, and the subtracter 101
is the thickness deviation e (k+1) from the thickness setting value rs(k+1)
)=rs(k+1)−ys(k+1) is output.

(3) The integrator 102 has a thickness deviation ε from the subtracter 101.
(k+1) is input and the time integral □ value of the thickness deviation is output from the following equation.

X, (k+1)・XI(k)+0.5(tk, 1-tk
) (ε (k)+ ε (k+1)l (69)
Here, ε(k) is the thickness deviation at the previous thickness detection (time t=b), and XI(k) is the thickness deviation at time 1=1. Integrator 10 at
This is the output of 2.

The integrator 102 functions as a disturbance compensator by compensating the disturbance heat that changes the thickness y with the heat generated by the heater so that the thickness y always matches the set value.

(4) When the thickness gauge reaches either side of the film edge, the thickness gauge outputs the reaching point identification signal d, and depending on the identification signal d, the formula (55) (56) or (57) ( 58)
Calculate ω(k+1) from the formula ＾. That is, the past heater generated heat time series (u(k-2)
, u(k-1)) and film thickness detection value y(k+1) are input to the observation device 103, and the time (( The estimated value of the state variable X (tk
, +-t) =^ ω(k+1).

(5) By calculating the first term on the right side of equation (66), time (tk
, 1-L), the state estimate (X + (k)1), ω
(k+1))' is multiplied by a coefficient that causes the state to change by the average dead time determined by equation (68) to obtain the estimated value of the state at time tk. . The size of the dead time τ is (68)
As shown in the formula, the average value of the dead time at both ends of the film is adopted.

Average Input u (
The state transition due to 〒(k) in the second term on the right side of equation (66)
+1), and this correction is performed by the next state predictor 106.

(6) The second term on the right side of equation (66), I (k+1), is the time series u(k-
I(k+1), which represents the amount of state transition due to ILu(k)
) is calculated from equation (67) using the average dead time.

That is, the time series of past heater generated heat (here, two, u(k-1) and u(k)) determined by the length of the dead time n stored in the memory 104 is sent to the state predictor 106. and outputs the state change amount 1 (k+1) due to the input u(k) from time (b-+-L) to time jk+1.

e”LX+(k+lLa1Ck+1)]” and the output 1 (k+1) of the state predictor 106 are added to the time E.
State estimate at h*+ (XI(k+1), X(k+1)
)".Thus, due to the dead time, only the state estimate at time (th, +-L) can be obtained from the observed value 103, but the state transitioner 105 and the state predictor 106 output the average dead time. By performing an integral operation for the time t,
1 can be obtained. By this operation, the influence of phase delay due to dead time can be removed.

(8) The heat generated by the heater u(k+1) from time tk-1 to the next control calculation time tk*2 is determined by the following equation using the state feedback gain (f+, h).

The heater generated heat command value is determined by multiplying by the gain.

(9) The above control calculation is performed so that the thickness gauge moves in the film width direction after the time period T and reaches the opposite film end at the control calculation execution time L = tv-z. ) is obtained from the sampler 100 and executed.

Next, as a first concrete example, the transfer function g+(s
).

A design example will be shown where gz(s) and g3(s) are given by the following equations.

u+(t) (i = 1 to 5) indicates the amount of change in heat generated by each heater (in watts), and yt(t) (+ = 1 to 5)
5) shows the amount of thickness change (in microns) at the thickness needle position corresponding to each heater position. Dead time due to film flow delay and time L2' required for the thickness gauge to move from the thickness control point 3' to the edge of the film.

L2'' (see FIG. 8) assumes the following value.

Then, = 26 seconds Lz' = 17 seconds Lx' = 7.5 seconds Therefore, L exposure = 43 seconds Lc = 33.5 seconds The thickness control point 3' is assumed to be on the 0 point side of the film edge in Fig. 8. . The control calculation execution period T assumes the following value.

T = 22.5 seconds (72) In order to design the control system, input u(t) of equation (1) and output V
It is necessary to obtain controllable and observable state equations (2) (31) that express the relationship between (t). (69) to (71
) of g+(s)2gz(s)9gs(s)
(s) can be expressed by a 77th order state equation, but it was found that the controllable and observable state equation is 29th order. Therefore, from G (s), the 29th order state equation (21(3
One set was obtained.

(1) Determination of state feedback gain matrix F 01) The state feedback gain matrix F in equation (2) was found as a solution to the optimal regulator problem for equation (8), which is a 30th expansion of equation (2). . Since equation (8) is a one-state equation for a continuous time system, it can be expressed as the sampling period T
= 22.5 seconds, the regulator solution method was applied instead of the discretized state equation. As a result of finding the state feedback gain matrix F using an appropriate evaluation function, the following values were obtained as the eigenvalues of the matrix (A-BF), which are the main values that determine the response of the control.

0.856.0.8119.0.7755.0.761
8. Eigenvalues other than those listed above are not described because their absolute values are small and their attenuation is fast. It can be seen that stable control is possible because all the eigenvalues are within a circle with a radius of 1. Since the eigenvalue with the slowest decay is 0.8560, the settling time T, defined with a control error of 1%, is (0,876)"#0.0
1, the settling time T can be predicted to be approximately 11 minutes as follows.

T, -T X35=22.5X30 seconds=675 seconds=1
1.3 minutes (2) Determining the feedback gain of the observer The feedback gain matrix of the observer in equation (56) or equation (58) is determined by the 29th order state equation (55) or (5
8) and the fifth-order output equation (54).

In the gain matrix, K is the matrix (Φ, 〒−(CΦ,)DingKl'') and the matrix (Φc''(CΦc)'Kcf)
The solution to the optimal regulator problem is obtained so that the eigenvalues are stable. For example, in the gain matrix, the equation of state that determines Equation 4 (%) is 13 seconds. As a result of finding the gain matrix K using an appropriate evaluation function, the following values were obtained as the eigenvalues of the matrix (Φ1-KsCΦ,) that mainly determine the convergence of the observer.

0.9183.0.9183.0.9183.0.91
83.0.9183゜0.7654.0.7654.0
．． 7654.0.7654.0.7654° Eigenvalues other than the above are not described because their absolute values are small and convergence is fast.

Since both are within a circle with a radius of 1, the estimation error can be reduced over time. Since the slowest eigenvalue is 0.9183, the time T0 required for the estimation error to decay to 1% of the initial value is (0,9183
)5S! =i0.01, it can be predicted as follows.

T0 = (T-1, m+Lc) x55 = 13x55 seconds - 715 seconds - 12 minutes Also for the matrix Φ, the settling time T0
A gain matrix of approximately 12 minutes was obtained.

By performing control calculations using the gain matrices F, Km, and Kc obtained above, an example of the control results obtained is shown in the second example.
Figure 20 (a) shown in Figure 0 shows five thicknesses y1~ when the set value of thickness y is changed stepwise by 5 microns.
FIG. 20(b), which also shows the time course of the amount of change in C (the amount of change in the detected value of the thickness gauge), shows the heat generated by the heater 11. It shows the amount of change in %ll.

Execution cycle time of control calculation after changing thickness setting value 22
．． Since the control calculation is performed after 5 seconds, the change in the heat generated by the heater occurs 22.5 seconds after the thickness setting value is changed.

Heat generated by the heater remains at the same value until 22.5 seconds after the next control calculation is performed, and after 22.5 seconds, control calculation is performed based on the new thickness detection value and the heat generated by the heater is changed. . Therefore, the heat generated by the heater changes stepwise as shown in FIG. 20(b).

On the other hand, a change in the detected thickness value is detected after a further dead time L=33.5 seconds after the heat generated by the heater changes for the first time at 22.5 seconds after the set value changes. That is, the thickness change is detected 22.5 seconds + 33.5 seconds = 56 seconds after the thickness setting value changes. The thickness y has been changed to the correct setting value, and changes almost symmetrically with respect to the thickness y. On the other hand, in the heat generated by the heater, the amount of change in heater U is the largest (the change in heater ul, u% is the next largest, and the change in heater ut+u4 is the smallest.
This is to reduce the interference of 4. Gain matrix F, Km
The fact that the settling time estimated from the eigenvalue determined by , Kc is 12 minutes is supported by FIG. Also, the reason why there is no fluctuation in heater heat generation during steady state is that to compensate for dead time, instead of formula (64) and (65), formula (66) is used.
This is due to the use of Eq.

When actually applying this control method, each thickness Vr+V*
+Va+The same control formula as applied to the thickness ys is applied to Vs, and each heater generated heat command value can be output as the sum of the results of each control formula. Note that the heat generated by the heater is not controlled. Since the time required for the change in thickness to stabilize when the thickness is changed stepwise is approximately 13 minutes, it can be seen that this control method considerably reduces the influence of dead time.

Next, to explain the second specific example with reference to FIG.
Figure 1 shows the control results when 8 watts of disturbance heat is applied to the heater u3 in a stepwise manner. b)
shows the change over time in the heater generated heat ul""uz. As shown in Fig. 21(a), the thickness y increases by -1 times due to the disturbance heat of the heater, but by changing the heat generated by the heater U, ~U, the thickness y returns to the original setting value. It can be seen that the settling time is about 12 minutes as in FIG. 20, and that disturbance compensation is achieved well by introducing an integrator using the zero line control method.

The thickness yt+Va is affected by disturbance heat due to heat conduction in the die width direction, and increases by -1. The thickness V++Vs is similarly affected, but the effect is naturally smaller than that of Vt+Va.To offset the effect of such disturbance heat, the amount of decrease in heat generated by heater U is the largest, and heater L11. u
The next largest percentage decrease is Heater LI:.

The amount of decrease in u4 is the smallest. This is because the amount of decrease in heat tam and uJ was made so as not to affect the thickness 18 so much. When disturbance heat is applied to heater U, other thickness Vl*
Although Vz*Va+Vs also changes, each thickness V++Vt
By applying the same control equation for the thickness y to rVa+Vs, such interference effects can be canceled out.

〔Effect of the invention〕

Since the present invention is configured as described above, it has the following effects. That is, by introducing an integrator that time-integrates the difference between the thickness detection value and the thickness setting value at a predetermined position of the film thickness, and feeding back the output of this integrator to the heater generated heat command value, the disturbance heat that changes the thickness is It is possible to compensate for this with the heat generated by the heater so that the thickness always falls within the set value. In addition, in order to avoid a large phase delay due to dead time, the observation device uses the time (t-
By obtaining the state estimate at the current time t, the state transition unit and the state predictor time-integrate the state by the amount of dead time, thereby obtaining the state estimate at the current time t and reducing the deterioration of control performance due to the dead time. Can be removed.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a control device showing a first embodiment of the present invention, Fig. 2 is an explanatory diagram showing the configuration of a conventional film manufacturing plant, and Fig. 3 shows an example of arrangement of heaters embedded in a die. 4 is a block diagram of a conventional film thickness control device, and FIG. 5 is a block diagram of a conventional film thickness control device including dead time. , (B) is a diagram showing the same block, FIG. 6 is an explanatory diagram showing the correspondence between the heater position of 5&ll and the thickness detection position of 5&iI, and FIG. 7 is a dynamic mathematical model of film thickness. Block diagram, Figure 8 is a line diagram showing the trajectory of the thickness meter that detects film thickness, Figure 9 is a line diagram explaining the time interval and time integration section of control calculation, Figure 1
0, 11, and 12 are diagrams explaining various time integration intervals, and FIGS. 13 and 14 are simulation results using the apparatus of the first embodiment of the present invention (for changing the thickness setting value). FIGS. 15 to 21 are diagrams related to the second embodiment of the present invention, and FIGS. 15, 16, and 17 are diagrams showing control calculations. Figures 18 and 19 are diagrams illustrating the time interval and time integral interval of , and Figures 18 and 19 are diagrams illustrating simulation results illustrating the effect of using the average dead time as the integral interval of the state transition device and the state predictor. 20 and 21 are explanatory diagrams showing simulation results (when changing the thickness setting value and when disturbance heat is applied to the heater).

Claims (1)

[Claims] It has a die with a mechanism to adjust the discharge amount of molten resin along the width direction of the film, and has a dead time L_ which is the time required for the film to flow between the die position and the thickness gauge position.
In a film extrusion molding and casting molding device having a thickness gauge that detects thickness changes using 1, the thickness gauge detects a thickness value corresponding to a predetermined position in the film width direction, and a thickness setting value at that position. a subtracter that outputs the difference in thickness, and an integrator that performs time integration of the thickness difference output from the subtracter and outputs the result.
A memory that stores a time series of past operation amounts of the discharge amount operation end by the length of the sum of the dead time L_1 and the time L_2 from when the thickness gauge detects the thickness at a predetermined position until it reaches the film side edge. , an observation device that outputs an estimated value of a state variable a dead time L before the time when the time series of the operation amount of the operation end in the past and the set value of the film thickness detection value stored in the memory were input; a state transition device that inputs the output of the integrator and the output of the observation device, multiplies it by a coefficient that causes the state to transition by a dead time L, and outputs an estimated state value at a predetermined time; and a state transition device that is stored in the memory. a state predictor that inputs the time series of the operation amount of the past discharge amount operation end and outputs the state change amount according to the input setting from a certain time to the time after dead time L; an adder that adds the outputs of the state predictor and outputs a state estimate at a predetermined time;
and a manipulated variable command device for the operating end that outputs a manipulated variable command value for the operating end by multiplying the estimated state value at a certain time, which is the output of the adder, by a state feedback gain. Film thickness control device.