JPH106333A - Preparation of prepreg - Google Patents

Abstract

PROBLEM TO BE SOLVED: To control a condition by an appropriate amt. of operation until characteristics of a prepreg is obtd. by determining the second manufacturing condition when production is started from the present first manufacturing condition, environmental condition and aimed characteristic values of the prepreg which are inputted by switching a grade and a relation between the first and the second manufacturing conditions. SOLUTION: When production is started, a control apparatus 9 refers to each data on the first manufacturing condition which can not be at least suddenly changed and the second manufacturing condition which can be changed from the past operational data based on inputted number of a product and aimed characteristic values of the prepreg and reads process conditions of an impregnation-drying process plant 21 into a system control means from a process data acquiring means 16. Then, operation is performed by an operational means 20 and using this result, the operational result and actual data are displayed on a displaying device 17 under control of the system control means. In addition, the operated amt. of operation is set on each controller through a process control means 18 and based on this, operation of the impregnation-drying process plant 21 is started.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for producing a prepreg for a laminated board.

[0002]

2. Description of the Related Art The control of prepreg production of a laminated board is controlled by a method of feeding back and controlling measured prepreg characteristic values as in the method described in JP-A-4-82728 or by the know-how of an operator. Is being done.

[0003]

In any of the above control methods, the method of determining the initial value of the operation amount such as the feed speed of the base material and the clearance of the squeeze roll is not clear. If the operator has determined, if the determined initial value is not an appropriate value, after a certain period of time has elapsed since startup, the prepreg characteristic value is actually measured,
Control is performed by comparing the measured characteristic value with the target characteristic value, and there is a problem that a defective product is produced until the effect appears. Further, when the operator cannot determine the initial value, it is necessary to confirm the set value by a trial operation, and there is a problem that efficiency is deteriorated.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has as its object to provide a prepreg in a manufacturing plant in which a prepreg characteristic value cannot be obtained and feedback control cannot be started unless a certain time has elapsed after the start of production. An object of the present invention is to provide a method of manufacturing a prepreg, which enables an appropriate operation amount to be set and controlled until a characteristic value is obtained.

[0005]

In order to achieve the above object, according to the first aspect of the present invention, a varnish is impregnated by moving a sheet-like substrate, and after the impregnation, the varnish-impregnated substrate is passed through a squeeze roller. In a prepreg manufacturing process in which the prepreg is sent to a dryer and dried to obtain a prepreg, the production conditions are feedback-controlled so that the characteristic value of the prepreg to be produced becomes a target characteristic value, and operation data is accumulated. A process of obtaining a relationship between first manufacturing conditions, environmental conditions, and prepreg characteristic values that cannot be rapidly changed from data, and second manufacturing conditions that can be changed without delay; Present first
At least the second production condition is determined at the start of production from the production conditions, the environmental conditions, the target prepreg characteristic value and the above relationship, and an appropriate operation amount is set until the prepreg characteristic value is obtained. Control.

According to the second aspect of the present invention, the manufacturing period under the first and second manufacturing conditions determined at the start of the production is a period until the characteristic value of the prepreg to be manufactured can be measured. First and second feedback control based on a comparison between an actually measured prepreg characteristic value and a target prepreg characteristic value.
It is characterized by setting the manufacturing conditions of the prepreg and controlling the operation by setting an appropriate operation amount until the prepreg characteristic value is obtained, and performing the feedback control by the actual measurement of the prepreg characteristic value. Good products can be prevented.

According to a third aspect of the present invention, in the first or second aspect, the temperature of the dryer is used as the first manufacturing condition, the ambient humidity is used as the environmental condition, and the line speed, Using the operation amount such as the clearance amount of the squeeze roller, at the start of production due to product change, etc., from the past dryer temperature, ambient humidity, prepreg characteristic value, line speed, clearance amount of the squeeze roller retrieved from the operation results database Read the past operation data, calculate the line speed and the clearance amount of the squeeze roller based on the past operation data and the current dryer temperature, ambient humidity and target prepreg characteristics, and calculate the line speed and squeeze roller It is characterized by setting the clearance amount as the operation amount at the start of production, Can reduce failure of the time, it is non-skilled of the work.

According to a fourth aspect of the present invention, in the third aspect of the present invention, the operation is performed based on the past operation data retrieved from the operation result database at predetermined time intervals after the start of the production, the current operation data and the target prepreg characteristic value. It is characterized by resetting the quantity, and can respond to changes in operation data after the start of production. According to a fifth aspect of the present invention, in the invention of the third or fourth aspect, when there is no operation data of the same type as the current type to be switched and changed, data of a type whose production conditions are close are used as past operation data. It is possible to deal with varieties without operation data.

According to a sixth aspect of the present invention, in the third, fourth, or fifth aspect of the present invention, for data that changes before and after the start of production, data after a certain time after the start of production is predicted from data before the start of production. However, it is characterized in that a predicted value is used instead of the current operation data until a certain time elapses, and the line speed can be stabilized rapidly.

According to a seventh aspect of the present invention, in the third or fourth or fifth or sixth aspect of the present invention, a multiple regression equation is obtained by using the operation amount data accumulated in the operation result database as an objective variable and using the operation data as an independent variable. At the start of production due to product change, etc., the current operation data and the target prepreg characteristic value are substituted into the multiple regression equation to calculate the manipulated variable. The initial value is obtained.

According to an eighth aspect of the present invention, in the third or fourth or fifth or sixth aspect, learning of the neural network is performed using the operation data accumulated in the operation result database as input teacher data and the operation amount data as output teacher data. At the start of production due to product type switching, etc., the operation amount is calculated by substituting the current operation data and the target prepreg characteristic value into the learned neural network, and the operation state and the characteristic value are calculated by the neural network. Even if the relationship is nonlinear, it can be expressed, and control can be performed in consideration of the nonlinearity.

According to a ninth aspect of the present invention, in the third or fourth or fifth or sixth aspect of the present invention, a difference between the operation data accumulated in the operation result database, the current operation data and the target prepreg characteristic value is determined by a fuzzy inference rule. It is characterized by determining the amount of correction by fuzzy inference and determining the amount of operation by using the amount of correction from the past operation amount as the input variable of the antecedent part and the amount of correction from the past operation amount as the output variable of the consequent part. Can be modified in the same manner as the operation of the operator.

According to a tenth aspect of the present invention, in the ninth aspect of the present invention, the operation result of the operation amount and the data of the operation amount at the time of stable operation are accumulated, and the operation result of the operation amount is corrected according to the tendency of the difference between the two. Inference errors due to fluctuations that occur in the process but cannot be observed can be considered. The invention of claim 11 is the invention of claim 3 or 4 or 5 or 6 or 7 or 8 or 9 wherein past operation data selected based on a predetermined standard is used. By taking into account, for example, the variation of the prepreg characteristic value within the standard range, an operation result of an appropriate operation amount can be obtained.

According to a twelfth aspect of the present invention, in the ninth aspect of the present invention, the judgment is made based on the evaluation value of the guidance inference error.
Fuzzy rules are modified from past operation data, fuzzy inference is performed using the modified rules, and the amount of operation is determined.The fuzzy system is constructed again by horizontal deployment to another line or changes over time in the process. When you have to, you can save a lot of time.

According to a thirteenth aspect of the present invention, in the ninth aspect of the present invention, the amount of change of the inference error of the guidance is small and the weight of the rule conformance is corrected when the inference error exceeds the inference error evaluation value. It is possible to cope with a time-dependent change of a simple process. According to the fourteenth aspect, in the ninth aspect,
The invention according to the invention, characterized in that when the amount of change of the inference error of the guidance is large and exceeds the inference error evaluation value, the membership function of the manipulated variable is modified to cope with a sudden change in the process such as cleaning in the dryer. can do.

According to a fifteenth aspect of the present invention, in the ninth aspect of the present invention, when the characteristics of the antecedent part input variable such as a change of a varnish or a base material to be used greatly change, the antecedent part membership function corresponding to the variable is used. Is corrected, and it is possible to cope with a rapid change in process conditions due to a change in a varnish or a base material. According to a sixteenth aspect of the present invention, in the ninth aspect of the invention, when modifying the membership function of the antecedent / consequent part,
The present fuzzy rule is characterized by adding a fuzzy rule when the inference error does not sufficiently converge. This is effective when the inference error does not sufficiently converge.

[0017] The invention of claim 17 is characterized in that when a substrate, a varnish or the like to be changed has been used before, the membership function used at that time is used, so that it is possible to quickly respond to the change. it can. According to an eighteenth aspect of the present invention, in the third to eleventh aspects, the initial value of the manipulated variable obtained by the calculation and the past operation data are displayed on the display means so that the calculation result can be corrected. Characterized in that
Correction based on the know-how of a skilled operator not taken into account in the calculation can be performed.

[0018]

FIG. 17 shows a process diagram of an impregnation / drying process plant using the method of the present invention.
The drying process plant impregnates a varnish 3 in a varnish impregnation tank 2 while feeding a hoop-shaped base material (for example, a glass base material) 1 at a predetermined line speed by a group of feed rollers 4, The substrate 1 dried in the dryer 6 is dried by hot air, and the substrate 1 dried in the dryer 6 is cut into a predetermined length by a cutter 7 to form a prepreg 8.
Is to be obtained.

The controller 9 measures the thickness of the varnish-impregnated substrate 1 coming out of the dryer 6 by adding the resin based on the weight measurement data obtained by the β-ray thickness meter 10 and the moisture content data obtained by the infrared water meter 11. A prepreg characteristic value is determined based on the volatile content based on the squeeze roller 5, and the setting data of the clearance amount of the squeeze roller 5 is sent to the squeeze roller controller 12 so that the determination result matches a preset target prepreg characteristic value based on the prepreg characteristic value. Line speed setting data is sent to the line speed controller 13 of the feed roller 4 group,
Further, while sending temperature setting data to the temperature controller 14 of the dryer 6, feedback control of the dryer temperature, which is an operation condition, the line speed and the clearance amount, which are operation amounts, is performed.

The squeeze roller controller 12 controls a clearance adjustment mechanism (not shown) of the squeeze roller controller 12 so that the clearance amount detected by the clearance detector 12a and the set value match, and the clearance amount matches the set value. Let it do. The line speed controller 13 sets a speed detector (not shown) to a set value.
The rotational speed of the drive motor (not shown) of the feed roller group 4 is controlled so that the detected line speeds match.

The temperature controller 14 has a temperature sensor 14a.
The amount of hot air sent to the dryer 6 is controlled by the adjuster 14b so that the temperature detected by the controller 14 becomes a set value. FIG. 18 shows a conceptual configuration diagram of a system for realizing the method of the present invention, and the control device 9 is constituted by a computer. The computer constituting the control device 9 includes a calculating means 20 for processing input data, creating control data, and the like, and functionally obtaining input data from an external input device (keyboard) 15 and obtaining from process data obtaining means 16. Import data, CR
T, a display control of the display device 17 such as a liquid crystal display, or a process of controlling the process control means 18, and a read / write control of the operation result database 19 by the arithmetic means 2.
System control means 21 based on arithmetic processing of 0
And a storage means constituting the operation result database 19.

The process data obtaining means 16 is for obtaining process data from the impregnation / drying process plant 21 shown in FIG. 17, and includes the above-mentioned temperature sensor 14a, speed detector,
It comprises a clearance detector 12a, a β-ray thickness meter 10, an infrared water meter 11, a temperature sensor (not shown) for measuring ambient temperature and humidity as environmental conditions, a humidity sensor (not shown), and the like.

On the other hand, the process control means 17 comprises means for setting control data of each part of the plant, such as the above-mentioned controllers 12, 13, and 14. Next, features of the present invention will be described with reference to embodiments. (Embodiment 1) In this embodiment, a brand (product type) that has been manufactured up to that time is switched to another brand (product type) and control is performed when starting manufacturing. It will be described based on.

First, in the past manufacturing process, brand (product type), production date and time, operation (drying) condition, base material, varnish, operation amount set value, obtained by the impregnation / drying process plant 21 or input from the input device 15 Data such as prepreg characteristic values are stored and accumulated in the operation result database 19 as operation results, and these data are referred to when a new production is started.

According to the method of the present embodiment, when the brand (product type) manufactured up to that time is switched to another brand (product type), the product number (==) of the prepreg to be manufactured this time is used.
Input device 15: brand = type) and target prepreg characteristic value
Input in advance. And the impregnation / drying process plant 21
Starts, the control device 9 performs arithmetic processing for control in a period during which feedback control is not possible, and creates control data at the start of production.

That is, when production (manufacturing) is started, the control device 9 changes at least abruptly from past operation data based on the input product number (= brand = product type) and the target prepreg characteristic value. In other words, the first manufacturing condition that causes a delay (time lag) before the change, ie, the dryer temperature, the ambient humidity, the prepreg characteristic value, and the operation amount (the second manufacturing condition that can be changed without the delay (no time lag)) For example, while retrieving each data of the line speed and the clearance amount of the squeeze roller 5, the process status of the impregnation / drying process plant 21, for example, the current dryer temperature and the ambient humidity, are obtained from the process data acquisition unit 16 to the system control unit 21. By the arithmetic means 20 based on the search data, the read data and the input data. The speed and the clearance amount of the squeeze roller 5 are calculated, and the calculation result is displayed on the display device 17 under the control of the system control means 21 to display the calculation result and the result data. The line speed and the clearance amount, which are the manipulated variables calculated in the above, are set, and the operation of the impregnation / drying process plant 21 is started based on the set data. This operation control is a control at the start of production. After the start of the control, the control device 9 checks through the process data obtaining means 16 whether or not the prepreg characteristic value can be actually measured. Feedback control of the clearance amount of the squeeze roller 5, the line speed, and the dryer temperature is performed by comparing the prepreg characteristic value with the target prepreg characteristic value.

FIG. 2 shows a time chart from the start of the above-mentioned production (manufacturing) to the start of the feedback control. (Embodiment 2) In Embodiment 1 described above, after the production is started, the control of the impregnation / drying process plant 21 is performed based on the calculated operation amounts, the line speed and the clearance amount, until the prepreg characteristic value can be measured. However, in this case, since it is not possible to cope with a change in data relating to the operation after the start of the production, the present embodiment requires a predetermined time as shown in FIG. 3 until the actual measurement of the prepreg characteristic value is possible after the start of the production. Each time, the current process status data is read, and an operation is performed based on the read data, the previously input data, and the past operation data retrieved from the operation result database 19, and the operation result and the operation data Is updated and the line speed and the clearance amount are reset based on the calculation result. This makes it possible to respond to changes in data relating to operations after the start of production.

FIG. 4 shows a time chart from the start of production (manufacturing) to the start of feedback control in this embodiment. (Embodiment 3) This embodiment relates to the case where the operation data of the same brand as the brand (type) of the prepreg to be produced this time in Embodiments 1 and 2 is not stored in the operation result database 19, In the first or second embodiment, when the operation data of the brand (type) of the prepreg to be manufactured this time is not stored in the operation result database 19, the production conditions (varnish type (characteristics) of the prepreg to be manufactured this time Value), base material type (characteristic value), and brand (product type) data close to the target prepreg characteristic value) are used as search data.

That is, if the control device 9 determines that the operation data of the same brand as the brand (product type) of the prepreg to be manufactured this time is not stored in the operation result database 19, as shown in FIG. Check if there is a brand (variety) that uses varnish, and if so, extract the data of the brand, extract the data of the brand with the closest substrate thickness from the extracted data, and further extract the data The data of the brand closest to the medium target prepreg characteristic value is selected.

Here, the closest one means that [(value on database)-(current value)] ² is the smallest. For example, in the case of the target prepreg characteristic value, α (curing degree target median value) Difference / current part target curing width) ² +
β (difference in target resin amount median / residual resin amount target width this time) ²
+ Γ (difference in melt viscosity target median / current part melt viscosity target width) or α (search target part cure degree target median-current part cure degree target median) ² + β (search target part resin quantity target median-current Select the minimum value that can be obtained from (part number resin amount target median) ² + γ (target part number melt viscosity target median-current part number melt viscosity target median). Where α,
β and γ indicate constants for assigning weights. The difference between the target median values means the difference between the median value of the target width of the search target product number and the median value of the target width of the current product number.

By employing this embodiment, it is possible to deal with brands not stored in the operation record database. (Embodiment 4) In this embodiment, as shown in FIG. 6, before the start of production such as dryer temperature (indicated by C1 in the figure) and after the start (C2
The data that changes with (1) is T1 before the start of production.
The data of T2 is predicted from the start of production from the start of production from the data of T2, and the predicted value is used in place of the current operation data until the lapse of a predetermined time from the start of production.

That is, the control unit 9 searches the past operation data accumulated in the operation result database 19, and obtains the temperature, humidity (ambient), characteristic values of the varnish, and characteristic values of the base material (for example, glass base material) 1. The following multiple regression equation is obtained using the dryer temperature C1 (T = T1) before the start of production as an explanatory variable and the dryer temperature C2 (T = T2) after the start of production as an objective variable.

Y1 = a0 + a1.x1 + a2.x2 + x
3 + a4.x4 + a5.x5, where a0 ... a5: regression coefficient (constant) x1: dryer temperature before starting production, x2: air temperature, x3: humidity (ambient), x4: characteristic value of varnish, x5: base material (glass) The characteristic value of the base material y1: the dryer temperature after the start of production Thus, in FIG. 18, the computing means 20 of the control device 9 functions as a multiple regression equation computing means. Temperature / humidity (ambient) / varnish characteristic value / substrate (glass substrate)
Substituting the characteristic value of (1) and the dryer temperature before the start of production, the dryer temperature after the start of production is estimated.

The multiple regression equation in this case is as follows. Dryer temperature after start of production = a0 + a1 · Dryer temperature before start of production + a2 · Temperature before start of production + a3 · Humidity before start of production (ambient) + a4 · Characteristic value of varnish before start of production +
a5. Characteristic value of base material (glass base material) before production start As described above, in this embodiment, until a certain time has elapsed from the start of production, the dryer temperature is calculated using the predicted value instead of the current operation data. Because of the setting, the dryer temperature can be stabilized quickly, and if necessary, the line speed can be stabilized rapidly.

(Embodiment 5) In Embodiment 4 described above, the dryer temperature C2 after the start of production is determined by using the dryer temperature C2 after the start of production as a target variable to obtain a multiple regression equation. Searches past operation data accumulated in the operation result database 19 in order to determine the initial setting values of the drying speed (line speed) and the clearance amount of the squeeze roller 5 at the start of production due to brand (product type) switching or the like. Using the dryer temperature, air temperature, humidity (ambient), varnish characteristic value, glass base material characteristic value, and prepreg characteristic value as explanatory variables, the drying speed and clearance of the squeeze roller 5 are used as objective variables, and the following variables are used. Find the regression equation.

Y1 = a0 + a1 · x1 + a2 · x2 + x3 + a4 · x4 + a5 · x5 + a6 · x6 (1) y2 = b0 + b1 · x1 + b2 · x2 + x3 + b4 · x4 + b5 · x5 + b6 · x6 (6) a0: Constant) x1: drying speed, x2: temperature, x3: humidity (ambient) x4: characteristic value of varnish, x5: characteristic value of substrate (glass substrate) x6: characteristic value of prepreg y1: drying speed y2: clearance amount In this embodiment, the arithmetic means 20 of the control device 9 in FIG. 18 also functions as a multiple regression equation arithmetic means in FIG. Substitute the dryer temperature, air temperature, humidity (ambient), varnish characteristic value, substrate (glass substrate) 1 characteristic value, and prepreg characteristic value to obtain the drying speed. The initial values of the degree and the clearance of the squeeze roller 5 are determined.

Drying speed set value = a0 + a1 · Drying speed at start of production + a2 · Temperature at start of production + a3 · Humidity at start of production (ambient) + a4 · Characteristic value of varnish at start of production + a5 · Base at start of production Characteristic value of material (glass substrate) + a
6. Characteristic target value of prepreg Clearance set value = b0 + b1 Drying speed at the start of production + b2 Temperature at the start of production + b3 Humidity at the start of production (ambient) + b4 Characteristic value of varnish at the start of production + b
5. Characteristic value of base material (glass base material) at the start of production + b6
Characteristic target value of prepreg As described above, in this embodiment, by using a multiple regression equation,
With a simple calculation, an appropriate initial value of the drying speed and the clearance amount of the squeeze roller 5 can be set. (Embodiment 6) In the above-described Embodiment 5, the drying speed and the squeeze roller
In this embodiment, a neural network operation unit is used as the operation unit 20 in FIG.
As shown in FIG. 7, the past operation data accumulated in the operation result database 19 is searched, and the dryer temperature, temperature, humidity (ambient), varnish characteristic value, and base material (for example, glass base material)
1 is used as input teacher data, and the drying speed. Learning of the neural network is performed using the clearance amount of the squeeze roller 5 as output teacher data, and when the production is started by switching brands (types), the process data unique to the learned neural network is
That is, the operation data (dryer temperature, air temperature, humidity (ambient), varnish characteristic value, glass substrate characteristic value) and prepreg characteristic target values shown in FIG. The initial value of the clearance amount is determined.

In this embodiment, even if the relationship between the operating condition and each characteristic value is non-linear, it can be expressed by the neural network, and control in consideration of the non-linearity becomes possible. (Embodiment 7) In the present embodiment, in order to make a correction similar to the operation of the operator on the past set value in which a non-defective product was manufactured, in the present embodiment, the calculating means 20 of the control device 9 in FIG. Fuzzy inference calculation means, and at the start of production due to brand (product type) switching, etc., the difference between the past operation data (for example, the previous production) and the data at the start of production is used as an antecedent input variable of the fuzzy inference rule. The correction amount is determined by the following fuzzy inference rule using the correction amount as a consequent output variable from the past drying speed setting value and the setting value of the clearance amount of the squeeze roller 5, and the initial values of the drying speed and the clearance amount of the squeeze roller 5 are obtained. The setting value is determined.

IF past dryer temperature-current dryer temperature> 0 THEN Correction to be slower than past drying speed; IF past dryer temperature-current dryer temperature> 0 THEN
(Eighth Embodiment) In the present embodiment, when fuzzy inference is used as in the seventh embodiment, inference errors due to fluctuations that occur in the process but cannot be observed are considered. In this embodiment, for example, as shown in FIG. 9, the estimated drying speed and the estimated clearance of the squeeze roller 5, and the data of the drying speed and the clearance of the squeeze roller 5 during stable operation are accumulated. The inference value is corrected according to the tendency of the difference between the two. For example, the THEN drying speed estimation value in which the IF drying speed estimation value has an error earlier is corrected later.

The estimated value of the clearance amount of the IF squeeze roller has a wider error. The estimated value of the clearance amount of the THEN squeeze roller is corrected to be smaller. The amount is modified by the fuzzy inference rule. (Embodiment 9) In this embodiment, when fuzzy inference is used as in Embodiment 7, a fuzzy inference rule is automatically corrected from past operation data from an evaluation value of a guidance inference error, and the corrected rule is used. The function of performing fuzzy inference and determining the amount of operation is provided in the arithmetic means 20 of the control device 9 in FIG.

Here, for example, the process variables are varnish characteristic values,
The case where the substrate characteristic value and the operation amount are the drying speed will be described. First, the membership function used for fuzzy inference is as shown in FIG. 10, and the following rules are used. Rule i: (weight of rule conformance ωi) IF x1 is M1k and x2 is M2k THEN y = Pk (i = 1,2, ...,
n) i; Rule number n: Number of rules x1: Varnish characteristic value X2: Substrate characteristic value y: Drying speed M1k: k-th membership function in the varnish characteristic value domain M2k; in the substrate characteristic value domain k-th membership function Pk: k-th membership function of output variable y Thus, the inferred value yr of the drying speed can be calculated by the following equation (1).

Yr = Fuzzy (x1, x2, ω1, ..., ωn, M11, ..., M1a, M21, ..., M2b, P1, ..., Pc) ... (1) Fuzzy (): Fuzzy inference operation using delta rule The automatic tuning of the fuzzy rule can be formulated as a problem of minimizing the evaluation function E of the following equation (2). E = 1/2 / · (y−yr) ² ... (2) y: Drying speed In order to decrease the value of the evaluation function E, the weight ωi of the rule conformity, which is a fuzzy rule parameter, and an antecedent input For the variables (varnish characteristic value, base material characteristic value), membership function M1k / M2k, consequent part output variable (drying speed), and membership function Pk, the direction vector in which the evaluation function E decreases may be calculated. . These direction vectors are -∂E / ∂ωi, -∂E / ∂M1k, -∂E / ∂M2k,-
Since it is expressed by ∂E / ∂Pk, the equation for adjusting the fuzzy rule parameters is as follows.

Ωi (t + 1) = ωi (t) −K1 (∂E / ∂ωi) (3) M1k (t + 1) = M1k (t) −K2 (∂E / ∂M1k) (4) M2k (t + 1) ) = M2k (t) −K3 (∂E / ∂M2k) (5) Pk (t + 1) = Pk (t) −K4 (∂E / ∂Pk) (6) Kn: constant t: number of learning A fuzzy rule that minimizes the evaluation function E is obtained by repeatedly performing the calculation of the above expression by the calculation means 20 shown in FIG.

It is to be noted that the same concept as described above holds when the process variables increase or when the operation amount is the clearance amount of the squeeze roller 5. FIG. 11 shows the rule conformity ωi.
Is shown in the flowchart (algorithm) of the arithmetic processing of the arithmetic means 20 in the case of automatically correcting the rule. As shown in this flowchart, first, the weight ωi (1) of the currently used rule conformance is set as an initial value, and Is set to 1 and a variable s indicating the number of tuning data (data for learning fuzzy rule parameters extracted from past operation data) is set to 1. After this setting, read the s-th tuning data and input data (varnish characteristic value, base material characteristic value) x1, x2
Fuzzy inference is performed on to obtain the inference result ys. Using the obtained inference result ys and the output data (drying speed) yrs, the weight ωi (p) of the rule conformance is updated from the equation (3).

The above processing is repeated until the value of s becomes equal to or greater than the maximum number of tuning data. When the value of s reaches the maximum number of tuning data, the following equation is calculated.

[0046]

(Equation 1)

In this operation, if D is not less than the learning end threshold, the maximum learning number is compared with the value of p. If the value of p is smaller than the maximum learning number, the above processing is repeated. If D is equal to or less than the learning end threshold value or if the value of p is equal to or greater than the maximum number of times of learning, the correction process ends. In the above formula, yn
Indicates the drying rate for the nth tuning data, and yrn indicates the inferred value for the nth tuning data.

The above flow chart shows the case where the rule conformity ωi is automatically corrected. However, when the membership function of the drying speed is corrected, Pk is set. When the membership function of the varnish characteristic value is corrected, M1k is set. When modifying the membership function of the substrate characteristic value, the above-mentioned algorithm is established by replacing M2k with ωi. In the present embodiment, judging from the evaluation value of the inference error of the guidance,
Fuzzy rules are automatically corrected from past operation data, fuzzy inference is performed using the corrected rules, and the amount of operation is determined. When you have to build, you can save a lot of time.

In addition, when the variance of the inference error of the guidance is large and the inference error exceeds the evaluation value (threshold), the gradual process is performed by automatically correcting the weight ωi of the rule conformance. It can respond to aging. If the membership function of the manipulated variable (the output variable of the consequent part) is automatically corrected like the drying speed, it is possible to cope with a rapid change in a process such as cleaning in the dryer 6.

Further, when the characteristics of the input variable of the antecedent part greatly change, such as changing the varnish or the base material to be used, the membership function of the antecedent part (varnish characteristic value or base material characteristic value) corresponding to the variable is changed. With automatic correction, it is possible to cope with a sudden change in process conditions due to a change in varnish or base material. When the membership function of the precedent or the consequent is automatically corrected as described above, the inference error may not sufficiently converge with the current fuzzy rule. In this case, a fuzzy rule may be added.

For example, the antecedent variable is a varnish characteristic value, the consequent variable is a drying speed, the membership function is shown in FIG. 12, and the rule is an IF varnish characteristic value is M11 THEN drying speed = P1 IF varnish characteristic. Value is M12 THEN drying speed = P2 IF varnish characteristic value is M13 THEN When drying speed = P3, the inference error does not sufficiently converge even if learning is performed using the algorithm shown in FIG. 11 (FIG. 13). Indicates a case in which learning is performed, and x indicates tuning data, and inference errors of the data are not sufficiently converged in a range surrounded by a circle in the figure), and a rule is added as follows, for example.

Additional rule IF Varnish characteristic value is M14 THEN Drying speed = P4 When this rule is added, re-learning is performed using the additional rule and the algorithm shown in FIG. 11 to obtain an inference error. If the inference error does not sufficiently converge in this learning, further rules are added. In this way, the above processing is repeated until the inference error sufficiently converges.

When the characteristics of an antecedent part input variable such as a change of a varnish or a base material to be used greatly change, the membership function of the antecedent part (varnish characteristic value or base material characteristic value) corresponding to the variable is changed. Although the automatic correction has been described above, if the varnish or substrate to be changed has been used before, the membership function used at that time may be used.

In this case, for example, when the type of varnish is changed from [A] to [C] as shown in FIG.
.. Of the varnish [C] are searched from the varnish membership functions MEMα registered in the fuzzy rule database 24 provided in the actual operation number database 19. Where the varnish [C] membership function MEMγ
If there is, the membership function MEMγ is read and adopted as the membership function of varnish [C] to be changed. If not, the automatic correction is performed as described above.

(Embodiment 10) In this embodiment, past operation data used in the above embodiment is selected based on a predetermined standard. That is, in the present embodiment, the input (dryer temperature, air temperature, etc.) space of the operation result database 19 in FIG. 18 is divided into n regions as shown in FIG. It is assumed that operation data can be stored in each area... Up to m pieces of common data can be stored, and thereafter, old data and new data are sequentially replaced.
For example, in FIG. 15A, when an attempt is made to add data k to the area, since the maximum number of data has already been accumulated, the data 2 at the head is deleted and the data k is sequentially moved up to the m-th data. Are stored as shown in FIG. 10 (b).

Since this embodiment is adopted as a part of each embodiment, the description other than the above-described method will be omitted. (Embodiment 11) In principle, the above embodiments are shown in FIG.
Although realized by the configuration shown in FIG. 18, when the initial value of the operation amount is calculated and obtained in each embodiment,
The obtained manipulated variable initial value is displayed on the display device 17 shown in FIG. 18 on the monitor, and at the same time, the operation data is displayed on the display device 17 as shown in FIG. 16, and the calculation result is corrected by the input device 15 by the operator's judgment. This embodiment is capable of performing the above operations. That is, the control device 9 is provided with a program for this correction operation and a program for display.

Therefore, by adopting the display and correction method of the present embodiment, the correction can be performed by the know-how of the skilled operator which is not considered in the calculation.

[0058]

According to the first aspect of the present invention, the varnish is impregnated by moving the sheet-like substrate, and after the impregnation, the varnish-impregnated substrate is sent through a squeeze roller into a dryer and dried to obtain a prepreg. At the same time, in the prepreg manufacturing process in which the manufacturing conditions are feedback-controlled so that the characteristic value of the prepreg to be manufactured becomes the target characteristic value and the operation data is accumulated, the first manufacturing method that cannot be rapidly changed from the past operation data is used. A step of obtaining a relationship between the condition, the environmental condition, the prepreg characteristic value, and the second manufacturing condition that can be changed without delay, and the current first manufacturing condition, environmental condition, At least the second manufacturing condition is determined at the start of production from the target prepreg characteristic value and the above relationship, and until the prepreg characteristic value is obtained. There is an effect that it becomes possible to perform control by setting the appropriate operation amount.

According to a second aspect of the present invention, the manufacturing period under the first and second manufacturing conditions determined at the time of starting the production is defined as a period until the characteristic value of the prepreg to be manufactured can be measured. The first and second manufacturing conditions are set by feedback control based on a comparison between an actually measured prepreg characteristic value and a target prepreg characteristic value, and an appropriate operation amount is set until the prepreg characteristic value is obtained. This makes it possible to prevent the occurrence of defective products until the feedback control based on the actual measurement of the prepreg characteristic value becomes possible.

According to a third aspect of the present invention, in the first or second aspect, the temperature of the dryer is used as the first production condition.
Using the ambient humidity as the environmental condition, the operation amount such as the line speed and the clearance amount of the squeeze roller as the second manufacturing condition, and the past drying machine searched from the operation result database at the start of the production due to the type change or the like. Temperature, ambient humidity, prepreg characteristic value, line speed,
Reads past operation data consisting of the squeeze roller clearance and calculates the line speed and squeeze roller clearance based on the past operation data and the current dryer temperature, ambient humidity and target prepreg characteristics. It is characterized by setting the line speed and the clearance amount of the squeeze roller as the operation amount at the start of production, and has an effect that defects at the start-up can be reduced and work can be made unskilled.

According to a fourth aspect of the present invention, in accordance with the third aspect of the present invention, an operation is performed based on past operation data retrieved from an operation result database at predetermined time intervals after the start of production, current operation data and target prepreg characteristic values. The method is characterized in that the amount is reset, and there is an effect that it is possible to respond to a change in operation data after the start of production. The invention of claim 5 is the invention of claim 3 or 4
Is characterized in that, when there is no operation data of the same type as the current type to be switched and changed, data of a type with similar production conditions is used as past operation data, and it is possible to cope with a type without operation data. effective.

The invention of claim 6 is the invention of claim 3 or 4 or 5
In the invention of the above, with respect to data that changes before and after the start of production, data after a certain period of time from the start of production is predicted from data before the start of production, and the current operation data is predicted until a certain period of time elapses It is characterized by using a value, and has an effect that the line speed can be stabilized rapidly.

The invention of claim 7 is the invention of claim 3 or 4 or 5
In the invention of the sixth aspect, a multiple regression equation is obtained by using the data of the operation amount accumulated in the operation result database as an object variable and the operation data as an independent variable. And calculating the manipulated variable by substituting the target prepreg characteristic value,
Use of the multiple regression equation has an effect that an appropriate initial value can be obtained by a simple calculation.

The invention of claim 8 is the invention of claim 3 or 4 or 5
Or, in the invention of 6, the learning of the neural network is performed by using the operation data accumulated in the operation result database as the input teacher data and the operation amount data as the output teacher data. The operation amount is calculated by substituting the current operation data and the target prepreg characteristic value, and it can be expressed by a neural network even if the relationship between the operation situation and the characteristic value is non-linear. There is an effect that can be.

The ninth aspect of the present invention is the third aspect of the present invention.
Or, in the invention of 6, the difference between the operation data accumulated in the operation result database, the current operation data and the target prepreg characteristic value is used as an antecedent input variable of the fuzzy inference rule, and the correction amount from the past operation amount is The amount of correction is determined by fuzzy inference as the output variable of the part, and the amount of operation is determined. The effect is that the same setting as the operator's operation can be performed on the past set values for which good products could be manufactured. There is.

According to a tenth aspect of the present invention, in the ninth aspect of the invention, the operation result of the operation amount and the data of the operation amount at the time of stable operation are accumulated, and the operation result of the operation amount is corrected based on the tendency of the difference between the two. This is advantageous in that an inference error caused by a variation that occurs in the process but cannot be observed can be considered. The invention of claim 11 is the invention according to claim 3 or 4 or 5 or 6 or 7 or 8 or 9
It is characterized by using past operation data selected according to predetermined criteria, and considering the aging of the process and the variation of the prepreg characteristic value within the standard range, the calculation result of the appropriate operation amount can be obtained. There is an effect that it can be obtained.

According to a twelfth aspect of the present invention, in the ninth aspect of the present invention, a fuzzy rule is corrected from past operation data by judging from an evaluation value of an inference error of guidance, and fuzzy inference is performed using the corrected rule. It is characterized by determining the amount of operation to be performed, and when the fuzzy system has to be constructed again due to horizontal development to another line or changes in the process over time, the time spent on it can be greatly reduced. .

According to a thirteenth aspect of the present invention, in the ninth aspect of the present invention, when the amount of change in the inference error of the guidance is small and exceeds the inference error evaluation value, the weight of the rule conformance is corrected. There is an effect that it is possible to cope with a time-dependent change of a simple process. According to a fourteenth aspect, in the ninth aspect, the membership function of the manipulated variable is corrected when the amount of change in the guidance inference error is large and exceeds the inference error evaluation value. There is an effect that rapid changes in the process such as cleaning can be coped with.

According to a fifteenth aspect of the present invention, in the ninth aspect of the invention, when the characteristics of the antecedent part input variables such as a change of the varnish or the base material to be used are greatly changed, the antecedent part membership function corresponding to the variable is changed. Is corrected, and there is an effect that it is possible to cope with a rapid change in process conditions due to a change of a varnish or a base material. According to a sixteenth aspect of the present invention, in the invention of the ninth aspect, a fuzzy rule is added when the inference error does not sufficiently converge with the current fuzzy rule when modifying the membership function of the antecedent / consequent part. This is effective when the inference error does not sufficiently converge.

A seventeenth aspect of the present invention is characterized in that, when a substrate, varnish, or the like to be changed has been used before, the membership function used at that time is used, so that it is possible to quickly respond to the change. There is an effect that can be. The invention of claim 18 is the invention of claims 3 to 11, wherein the initial value of the operation amount obtained by calculation and the past operation data are displayed on the display means, and the calculation result can be corrected. This is advantageous in that the correction can be performed by the know-how of a skilled operator that is not taken into account in the calculation.

[Brief description of the drawings]

FIG. 1 is a flowchart of a first embodiment of the present invention.

FIG. 2 is a time chart for explaining the above operation.

FIG. 3 is a flowchart according to a second embodiment of the present invention.

FIG. 4 is a time chart for explaining the above operation.

FIG. 5 is a flowchart according to a third embodiment of the present invention.

FIG. 6 is a time chart for explaining the operation of the fourth embodiment of the present invention.

FIG. 7 is an explanatory diagram of Embodiment 6 of the present invention.

FIG. 8 is an explanatory diagram of the above.

FIG. 9 is a time chart for explaining the operation of the eighth embodiment of the present invention.

FIG. 10 is an explanatory diagram of a membership function according to a ninth embodiment of the present invention.

FIG. 11 is a flowchart (algorithm) for explaining the above operation.

FIG. 12 is an explanatory diagram of another membership function of Embodiment 1;

FIG. 13 is an explanatory diagram of the above.

FIG. 14 is an explanatory diagram of the above.

FIG. 15 is an explanatory diagram of Embodiment 10 of the present invention.

FIG. 16 is an explanatory diagram of a screen of the display device according to the eleventh embodiment of the present invention.

FIG. 17 is a configuration diagram of an impregnation / drying process plant using the present invention.

FIG. 18 is a conceptual configuration diagram of the above system.

[Explanation of symbols]

 19 Operational Performance Database

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] December 25, 1996

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0030

[Correction method] Change

[Correction contents]

Here, the closest one means that [(value on database)-(current value)] ² is the smallest. For example, in the case of the target prepreg characteristic value, α (curing degree target median value) Difference / current part target curing width) ² +
β (difference in target resin amount median / residual resin amount target width this time) ²
+ Γ (difference in melt viscosity target median / current part melt viscosity target width) ² or α (search target part cure degree target median-current part cure degree target median) ² + β (search target part resin quantity target median- Current part number target median value) ² + γ (Search target part melt viscosity target median-Current part number melt viscosity target median) Select the minimum value obtained from ² . Where α,
β and γ indicate constants for assigning weights. The difference between the target median values means the difference between the median value of the target width of the search target product number and the median value of the target width of the current product number.

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0033

[Correction method] Change

[Correction contents]

Y1 = a0 + a1.x1 + a2.x2 + a
3 · x3 + a4 · x4 + a5 · x5 However, a0 ... a5: regression coefficient (constant) x1: dryer temperature pre-production, x2: Temperature, x3: Humidity (ambient), x4: characteristic value of the varnish, x5: substrate (Glass base material) characteristic value y1: Dryer temperature after start of production Thus, in FIG. 18, operation means 20 of control device 9 functions as multiple regression equation operation means, and starts production by switching brands (types). Sometimes, the temperature, humidity (ambient), characteristic value of varnish, substrate (glass substrate)
Substituting the characteristic value of (1) and the dryer temperature before the start of production, the dryer temperature after the start of production is estimated.

[Procedure amendment 3]

[Document name to be amended] Statement

[Correction target item name] 0034

[Correction method] Change

[Correction contents]

The multiple regression equation in this case is as follows. Dryer temperature after start of production = a0 + a1 · Dryer temperature before start of production + a2 · Temperature before start of production + a3 · Humidity before start of production (ambient) + a4 · Characteristic value of varnish before start of production +
a5 · base for setting the drier temperature using a prediction value instead of the current operational data until the lapse of a predetermined time from the start of production in the characteristic value or more as the embodiment of (glass substrate), The dryer temperature can be stabilized quickly, and if it is, the line speed can be stabilized rapidly.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0036

[Correction method] Change

[Correction contents]

Y1 = a0 + a1 · x1 + a2 · x2 + a3 · x3 + a4 · x4 + a5 · x5 + a6 · x6 (1) y2 = b0 + b1 · x1 + b2 · x2 + b3 · x3 + b4 · x4 + b5 · x5 + b6. ... b6: regression coefficient (constant) x1: drying speed, x2: temperature, x3: humidity (ambient) x4: characteristic value of varnish, x5: characteristic value of base material (glass base material) x6: characteristic value of prepreg y1: Drying speed y2: Clearance amount In this embodiment as well, the calculating means 20 of the control device 9 in FIG. 18 functions as a multiple regression calculating means in FIG. The dryer temperature, air temperature, humidity (ambient), varnish characteristic value, substrate (glass substrate) 1 characteristic value, and prepreg characteristic value are substituted into the multiple regression equation. To is to determine the initial set value of the drying speed and the amount of clearance of the squeeze roller 5.

[Procedure amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0037

[Correction method] Change

[Correction contents]

Drying speed set value = a0 + a1 · Drying speed at start of production + a2 · Temperature at start of production + a3 · Humidity at start of production (ambient) + a4 · Characteristic value of varnish at start of production + a5 · Base material (glass base) Material value + a6 · Prepreg characteristic target value Clearance set value = b0 + b1 · Drying speed at start of production + b2 · Temperature at start of production + b3 · Humidity at start of production (ambient) + b4 · Varnish at start of production Characteristic value of + b
5 In the present embodiment as described characteristic above the target value of the characteristic value + b6 · prepreg Substrate (glass substrate) by using a multiple regression equation,
With a simple calculation, an appropriate initial value of the drying speed and the clearance amount of the squeeze roller 5 can be set.

[Procedure amendment 6]

[Document name to be amended] Statement

[Correction target item name] 0039

[Correction method] Change

[Correction contents]

IF past dryer temperature-current dryer temperature> 0 THEN Correction to be slower than past drying speed; IF past dryer temperature-current dryer temperature> 0 THEN
Modified wider than the clearance of the past squeeze roller 12 (Embodiment 8) In this embodiment, in the case of using the file <br/> di I reasoning as in the embodiment 7, but has occurred in step unobservable change In the present embodiment, for example, as shown in FIG. 9, the estimated value of the drying speed and the estimated value of the clearance amount of the squeeze roller 5, the drying speed at the time of stable operation and the clearance amount of the squeeze roller 5 are considered. the accumulated data, intended to modify the inference value by the tendency of the difference between, for example, IF drying speed inference value you modify the relatively slow tHEN drying speed inference value error out early.

[Procedure amendment 7]

[Document name to be amended] Statement

[Correction target item name] 0040

[Correction method] Change

[Correction contents]

The estimated value of the clearance amount of the IF squeeze roller has a wider error. The estimated value of the clearance amount of the THEN squeeze roller is corrected to be smaller. Fuzz
The amount is modified by the inference rule. (Embodiment 9) In this embodiment, when fuzzy inference is used as in Embodiment 7, a fuzzy inference rule is automatically corrected from past operation data from an evaluation value of a guidance inference error, and the corrected rule is used. The function of performing fuzzy inference and determining the amount of operation is provided in the arithmetic means 20 of the control device 9 in FIG.

Claims (18)

[Claims] A varnish is impregnated by moving a sheet-like substrate, and after impregnation, the varnish-impregnated substrate is sent to a dryer through a squeeze roller and dried to obtain a prepreg. In the prepreg manufacturing process of performing feedback control of manufacturing conditions and accumulating operation data so that the characteristic value becomes a target characteristic value,
A step of obtaining a relationship between first manufacturing conditions, environmental conditions, and prepreg characteristic values that cannot be rapidly changed from past operation data, and second manufacturing conditions that can be changed without delay; A method for manufacturing a prepreg, wherein at least a second manufacturing condition is determined at the start of production from the input current first manufacturing condition, environmental condition, target prepreg characteristic value, and the above relationship. 2. The manufacturing period under the first and second manufacturing conditions determined at the start of production is defined as a period until the characteristic value of the prepreg to be manufactured can be measured, and after the measurement can be performed, the measured prepreg characteristic value 2. The method according to claim 1, wherein the first and second manufacturing conditions are set by feedback control based on a comparison between the first and second target prepreg characteristic values. 3. A method using a temperature of a dryer as a first manufacturing condition, an ambient humidity as an environmental condition, and an operation amount such as a line speed and a clearance of a squeeze roller as a second manufacturing condition. At the start of production due to switching or the like, the past operation data including the past dryer temperature, ambient humidity, prepreg characteristic value, line speed, and the amount of clearance of the squeeze roller retrieved from the operation result database are read, and the past operation data and Calculate the line speed and the clearance amount of the squeeze roller based on the current dryer temperature, ambient humidity, and target prepreg characteristics, and set these line speed and the clearance amount of the squeeze roller as the operation amount at the start of production. The method for producing a prepreg according to claim 1 or 2, wherein 4. An operation amount is reset based on past operation data retrieved from an operation result database every predetermined time after the start of production, current operation data, and target prepreg characteristic values. 3. The method for producing a prepreg according to 3. 5. The method according to claim 3, wherein when there is no operation data of the same kind as the current kind to be changed, data of a kind whose production conditions are close are used as past operation data.
Or the manufacturing method of the prepreg of 4. 6. For data that changes before and after the start of production, data after a certain period of time from the start of production is predicted from data before the start of production. The method for producing a prepreg according to claim 3, wherein a predicted value is used. 7. A multiple regression equation is obtained by using the operation amount data accumulated in the operation result database as an objective variable and the operation data as an independent variable. At the start of production due to product change or the like, the current operation data and the current operation data are added to the multiple regression equation. 7. The method for producing a prepreg according to claim 3, wherein a manipulated variable is calculated by substituting a target prepreg characteristic value. 8. Learning of a neural network is performed by using operation data accumulated in an operation result database as input teacher data and operation amount data as output teacher data. 4. An operation amount is calculated by substituting the operation data and the target prepreg characteristic value.
Or the method for producing a prepreg according to 4 or 5 or 6. 9. A difference between operation data accumulated in an operation result database, current operation data and a target prepreg characteristic value is used as an antecedent input variable of a fuzzy inference rule, and a correction amount from a past operation amount is a consequent. 7. The prepreg manufacturing method according to claim 3, wherein a correction amount is obtained by fuzzy inference as a unit output variable, and an operation amount is determined. 10. The prepreg according to claim 9, wherein the operation result of the operation amount and the data of the operation amount at the time of stable operation are accumulated, and the operation amount operation result is corrected according to the tendency of the difference between the two. Manufacturing method. 11. The method for producing a prepreg according to claim 3, wherein past operation data selected based on a predetermined standard is used. 12. A fuzzy rule is corrected from past operation data by judging from an evaluation value of a guidance inference error,
10. The prepreg manufacturing method according to claim 9, wherein an operation amount is determined by performing fuzzy inference using the modified rule. 13. The prepreg manufacturing method according to claim 12, wherein the weight of the rule conformance is corrected when the amount of change in the inference error of the guidance is small and exceeds the inference error evaluation value. 14. The prepreg manufacturing method according to claim 12, wherein the membership function of the manipulated variable is corrected when the amount of change in the inference error of the guidance is large and exceeds the inference error evaluation value. 15. The antecedent membership function corresponding to a variable when the characteristic of an antecedent input variable such as a change of a varnish or a substrate to be used greatly changes. Prepreg manufacturing method. 16. The prepreg according to claim 9, wherein a fuzzy rule is added when the inference error does not sufficiently converge with the current fuzzy rule when modifying the membership function of the antecedent / consequent part. Production method. 17. The method for producing a prepreg according to claim 9, wherein when a substrate to be changed or varnish is used before, the membership function used at that time is used. 18. An apparatus according to claim 3, wherein an initial value of the manipulated variable obtained by the calculation and past operation data are displayed on a display means to enable correction of the calculation result. Prepreg manufacturing method.