JPH09143784A - Production of shielding plate for adjusting thickness of electrolytic copper foil - Google Patents

Abstract

PROBLEM TO BE SOLVED: To greatly reduce the labor, cost and time required for producing a shielding plate for adjusting the thickness of electrolytic copper foil. SOLUTION: The candidate data of a shielding plate length distribution is inputted to a neural network 10 built in accordance with the length distribution data of the existing shielding plate and the thickness distribution data of the copper foil produced by using the shielding plate. The predicted copper foil thickness distribution data is then determined in the neural network. The optimum data of the shielding plate length distribution to minimize the evaluation function value indicating the approximation of the predicted copper foil thickness distribution data and the target copper foil thickness distribution data is determined in an optimization system 20. The shielding plate is actually produced in accordance with the optimum data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for manufacturing a shielding plate for adjusting the thickness of electrolytic copper foil.

[0002]

2. Related Art It is known to manufacture a copper foil used for a printed wiring board or the like using an electrolytic copper foil manufacturing apparatus. The electrolytic copper foil manufacturing apparatus includes a rotating drum (cathode) whose lower half is housed in an electrolytic cell, and an electrode plate (anode) which surrounds the lower half of the rotating drum and is arranged in the electrolytic cell. A current is flown between the two and a copper solution (electrolyte) is supplied, and further, the copper electrodeposited on the rotary drum from the copper solution is peeled off from the rotary drum and wound on a roll.

In such an electrolytic copper foil manufacturing apparatus, the distance between the rotating drum and the electrode plate, the strength of the supplied current, the amount of the electrolytic solution supplied, etc. vary among the various parts of the apparatus. For this reason, the thickness of each part of the copper foil varies. On the other hand, there is a demand for a desired copper foil thickness distribution, particularly for a uniform copper foil thickness. Conventionally, in order to obtain the required copper foil thickness distribution, a shielding plate for adjusting the foil thickness is inserted between the rotating drum and the electrode plate, and the degree of electrodeposition of copper from the copper solution onto the rotating drum and thus the foil. It is known to adjust the thickness. That is, as the length of the shielding plate is longer, the electrodeposition of copper is suppressed and the foil thickness becomes thinner, so by using a shielding plate having an optimized length in each portion in the width direction of the shielding plate, a desired thickness distribution can be obtained. Of copper foil can be obtained.

[0004]

However, when the length of the shield plate at a certain position in the width direction of the shield plate is changed, not only the amount of electrodeposition at that position is changed but also the electrodeposition in the vicinity thereof is changed. The amount also changes. In addition to the shield plate length, there are other factors that affect the foil thickness, such as copper solution concentration, supply current intensity, and drum rotation speed. Therefore, it is difficult to express the proper length of each portion of the shield plate in the width direction by a calculation formula, and therefore it is difficult to properly set the length of each portion of the shield plate (shield length distribution). . After all, in order to obtain a desired shield plate, it is necessary for a skilled person to make many shield plates by trial and error, and the shield plate needs a lot of labor and cost.

Therefore, an object of the present invention is to provide a method for manufacturing a foil thickness adjusting shield plate relatively easily and in a short time.

[0006]

According to the present invention, a potential difference is applied between a rotating drum immersed in an electrolytic solution in an electrolytic cell and an electrode plate to remove copper electrodeposited on the rotating drum. There is provided a method for producing a shield plate for adjusting the thickness of an electrolytic copper foil, which is arranged between a rotating drum and an electrode plate in order to adjust the thickness distribution of the electrolytic copper foil obtained by.

According to the manufacturing method of claim 1, a neural network for obtaining a predicted copper foil thickness distribution from the shielding plate length distribution is manufactured by using the existing shielding plate length distribution data and the existing shielding plate. Depending on the step of building based on the thickness distribution data of the electrolytic copper foil, the step of inputting the candidate data of the shield plate length distribution to the neural network, the target copper foil thickness distribution, and the input of the candidate data. The process of inputting the predicted copper foil thickness distribution output from the neural network to the optimization system, and the shielding that minimizes the evaluation function value representing the closeness between the predicted copper foil thickness distribution and the target copper foil thickness distribution. It is characterized by including a step of obtaining optimum data of plate length distribution by an optimization system and a step of manufacturing a shielding plate for manufacturing an electrolytic copper foil having a target copper foil thickness distribution based on the optimum data. When That.

In the manufacturing method of the present invention, the lengths of the existing foil thickness adjusting shield plates at respective portions in the width direction are measured, and from the measurement results,
The length distribution data of this shield is obtained. Further, using this shielding plate, an electrolytic copper foil is manufactured by a copper foil manufacturing apparatus. That is, a shield plate is interposed between the rotating drum of the copper foil manufacturing apparatus and the electrode plate, and a potential difference is applied between the rotating drum and the electrode plate, and an electrolytic solution is supplied to cause electrodeposition on the rotating drum. The removed copper is peeled off from the rotating drum to obtain an electrolytic copper foil. Next, the thickness of each part of the electrolytic copper foil is measured, and the thickness distribution data of the electrolytic copper foil is obtained from the measurement result.

The connection strength between the cells forming the neural network is adjusted so that the output of the neural network when the shielding plate length distribution data is input to the neural network is as close as possible to the copper foil thickness distribution data. change. In other words, learning of the neural network is performed using the above-mentioned electrolytic copper foil thickness distribution data as teacher data. This gives the neural network the knowledge for predicting the thickness distribution of the electrolytic copper foil manufactured using the shielding plate having a certain length distribution.

Further, the same work is performed on another existing shield having a length distribution different from that of the shield, thereby further improving the coupling strength between the cells of the neural network. In this way, a neural network for obtaining the predicted copper foil thickness distribution from the shield plate length distribution is constructed. Next, the candidate data of the shield plate length distribution that enables the production of the electrolytic copper foil having the target thickness distribution is input to the neural network. The predicted copper foil thickness distribution corresponding to the candidate data is output from the neural network. Then, this predicted copper foil thickness distribution is input to the optimization system together with the target copper foil thickness distribution. The optimization system obtains an evaluation function value representing the closeness between the predicted copper foil thickness distribution and the target copper foil thickness distribution, and outputs new candidate data that reduces the evaluation function value to the neural network. Then, the optimization system obtains an evaluation function value based on the new predicted copper foil thickness distribution output from the neural network according to the new candidate data, and then outputs further candidate data. While repeating such data processing, the optimization system obtains candidate data that minimizes the evaluation function value as the optimum data of the shield plate length distribution.

Then, the foil thickness adjusting shield plate is actually manufactured based on the optimum data. The shielding plate thus obtained has an optimum length distribution for producing an electrolytic copper foil having a target copper foil thickness distribution.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION A method of manufacturing a foil thickness adjusting shield plate according to an embodiment of the present invention will be described below. Referring to FIG. 1, a copper foil manufacturing apparatus includes an electrolytic bath 1, a cylindrical rotating drum 3 having a lower half portion immersed in an electrolytic solution 2 in the electrolytic bath 1,
A semi-cylindrical electrode plate 4 that surrounds the lower half of the rotating drum 3 in the electrolytic cell 1 is provided. The rotary drum 3 is rotatably supported by, for example, the side wall of the electrolytic cell 1, and is rotated clockwise by a drive source (not shown) in FIG.

Further, the copper foil manufacturing apparatus is provided below the rotary drum 3 with a distributor 5 for supplying the electrolytic solution 2 between the rotary drum 3 and the electrode plate 4. The distributor 5 is connected to an electrolytic solution supply device (not shown). The electrolytic solution 2 supplied from the electrolytic solution supply device through the distributor 5 passes through the electrolytic solution passage between the rotary drum 3 and the electrode plate 4, and overflows from the upper edge of the electrode plate 4 into the electrolytic cell 1. It is like this. And electrolyzer 1
The electrolytic solution 2 therein is returned to the electrolytic solution supply device through an electrolytic solution discharge port (not shown) provided in the electrolytic tank 1, and is recycled after being purified in the device.

On the electrolyte overflow side of the electrolyte passage, a foil thickness adjusting shield plate 6 is arranged between the rotary drum 3 and the electrode plate 4. The shield plate 6 has a length distribution as illustrated in FIG. That is, the length of the shielding plate 6 has a different value depending on the position in the width direction of the shielding plate. Reference numeral 6
Reference characters a and 6b represent a base end and a tip of the shield plate 6, respectively. The copper foil manufacturing apparatus includes a potential difference applying device (not shown) for applying a potential difference between the rotating drum 3 and the electrode plate 4.
This potential difference applying device includes a rectifier, the negative electrode of the rectifier is connected to the rotating drum 3, and the positive electrode is connected to the electrode plate 4. Then, while rotating the rotary drum 3, a potential difference is applied between the rotary drum 3 and the electrode plate 4 and the electrolytic solution 2 is supplied, so that copper is electrodeposited on the peripheral surface of the rotary drum 3. . The copper electrodeposited on the rotating drum 3 is peeled off from the rotating drum 3 by a peeling means (not shown) installed in the vicinity of the rotating drum 3 on the electrolyte overflow side of the electrolytic passage, thereby obtaining an electrolytic copper foil 9. . The copper foil 9 is wound by a winding roll 7 which is rotationally driven counterclockwise by a drive source (not shown). In FIG. 1, reference numeral 8 indicates
A pass roll for guiding the copper foil 9 is shown.

The shield plate manufacturing method according to this embodiment is carried out using a neural network and an optimization system. Referring to FIG. 3, the neural network 10
Is an input layer 11 for inputting shield plate length distribution data consisting of R × Q (Q is an integer of 1 or more) data values, and S2 ×
The output layer 14 outputs the predicted foil thickness distribution data consisting of Q data values, and the first and second hidden layers 12 and 13 interposed between the input layer 11 and the output layer 14.

The shield plate length distribution data consists of Q sets of data values, and each set consists of R data values. A set of R data values represents the length of one shield plate 6 at the point R in the width direction, that is, the length distribution of the one shield plate 6. Therefore, the total of the Q set shield plate length distribution data values is
The length distribution of Q different shield plates is shown. The predicted foil thickness distribution data is composed of Q sets of data values, and each set is composed of S2 data values. A set of S2 data values represents the predicted thickness of one copper foil 9 at the widthwise S2 point, that is, the predicted thickness distribution of the one copper foil 9. Therefore, the entire Q set of predicted foil thickness distribution data values represents the predicted thickness distribution of Q different copper foils.

The input layer 11 is composed of R × Q input cells. Each of the input cells receives one of the R × Q data values forming the Q set of shield plate length distribution data, which is assigned to each cell. The first hidden layer 12 has a plurality of first intermediate cell groups and one first bias cell group. The first bias cell group has S1 × 1 first bias cells to which the first bias inputs are respectively input. In addition, each of the first intermediate cell groups includes S1 × R first intermediate cells. A plurality of input cells assigned to each of the first intermediate cells and one first bias cell are coupled to each of the first intermediate cells. A value obtained by multiplying the output of each of these input cells by the coupling strength (weight) of each of the first intermediate cells and each of the input cells is added to each of the first intermediate cells. , And the sum of the output from the first bias cell is input. The weights of the first intermediate cells belonging to the same first intermediate cell group are preferably the same. Then, each of the first intermediate cells outputs a value obtained by converting the input of each of the first intermediate cells by, for example, a tangent sigmoid transfer function.

The second hidden layer 13 has a plurality of second intermediate cell groups and one second bias cell group. The second bias cell group is S2 for inputting the second bias input, respectively.
× 1 has a second bias cell. In addition, each of the second intermediate cell groups includes S1 × S2 second intermediate cells.
A plurality of first intermediate cells and one second bias cell assigned to the second intermediate cells are coupled to each of the second intermediate cells. In each of the second intermediate cells,
A value obtained by multiplying the output of each of the first intermediate cells by the coupling strength of each of the second intermediate cells and each of the first intermediate cells and adding up the values for these first intermediate cells, and the value of the second bias cell The sum with the output is input. It is preferable that the weights of the second intermediate cells belonging to the same second intermediate cell group are the same. Then, each of the second intermediate cells outputs a value obtained by converting the input of each of the second intermediate cells by, for example, a linear transfer function.

The output layer 14 has S2 × Q output cells. All second intermediate cells are coupled to each of the output cells. The sum of the values obtained by multiplying the output of each of the second intermediate cells by the coupling strength of each of the output cells and each of the second intermediate cells is input to each of the output cells. Each output cell outputs a value obtained by converting the input of each output cell by, for example, a linear transfer function. From the S2 × Q output cells, S2 × Q data values forming the Q sets of predicted foil thickness distribution data are output.

In this embodiment, the neural network 1
0 is configured by a computer. A program for executing software simulation of a neural network model is installed in this computer. As this program, for example, a commercially available pattern recognition tool is used. Referring to FIG. 4, the optimization system 20 outputs Q sets (Q is an integer of 1 or more) output from the neural network 10 when the candidate data of the shield plate length distribution is given to the neural network 10. Of the estimated foil thickness distribution data and the target foil thickness distribution data that is supplied separately from the same data (e.g., S2 × Q estimated foil thickness distribution data values and the same number of targets). It has an evaluation unit 21 for obtaining the sum of squares of deviations from the foil thickness distribution data value.

Further, the optimization system 20 includes a discriminator 22.
And a calculation unit 23. The determination unit 22 is the evaluation unit 2
It is determined whether or not the evaluation function value obtained by 1 is zero or minimum. If the evaluation function value is not zero or minimum, a candidate data calculation command is sent to the calculation unit 23. On the other hand, if the evaluation function value becomes zero or the minimum, the determination unit 22
Outputs candidate data corresponding to this evaluation function value (predicted foil thickness distribution) as optimum data of the shield plate length distribution.

In accordance with the candidate data calculation command, the calculation unit 23 calculates new candidate data of the shield plate length distribution that reduces the evaluation function value by a conventionally known optimization method. The new candidate data is output to the neural network 10. The neural network 10 obtains the predicted foil thickness distribution corresponding to this candidate data. In this embodiment, the optimization system 20 is composed of a computer. This computer is shared by the neural network 10 and the optimization system 20, and is created by a conventionally known programming technique for achieving the functions of the respective elements 21, 22 and 23 of the optimization system 20. The implemented program has been implemented.

The manufacturing process of the shielding plate according to this embodiment will be described below. First, the worker uses the existing first shield plate 6
The length at the point R in the width direction is measured to create shield plate length distribution data consisting of R data values. Further, using the first shielding plate 6, the copper foil 9 is manufactured by the copper foil manufacturing apparatus shown in FIG. That is, while the first shield plate 6 is interposed between the rotating drum 3 and the electrode plate 4 of the copper foil manufacturing apparatus, a potential difference is applied between the rotating drum 3 and the electrode plate 4 and the electrolytic solution 2 is supplied. By doing so, copper is electrodeposited on the rotating drum 3, and the electrodeposited copper is peeled off from the rotating drum 3 to obtain a copper foil 9. Next, the operator measures the thickness of the copper foil 9 at the point S2 in the width direction and obtains the copper foil thickness distribution data consisting of S2 data values.

Next, the operator inputs the shielding plate length distribution data and the copper foil thickness distribution data obtained as described above into the computer as the neural network 10 via the keyboard of the computer. The neural network 10 performs learning according to the learning rule implemented therein, for example, the backpropagation rule.
In this learning, first, with the coupling strength between the cells of the neural network 10 and the first and second bias inputs set to their respective initial values, the shield plate length distribution data is input to the input layer 11 of the neural network 10. Entered in. In response to this data input, in the neural network 10, the outputs of the first hidden layer 12, the second hidden layer 13, and the output layer 14 are obtained in this order. Then, the output of the output layer 14 (predicted copper foil thickness distribution data) is compared with the copper foil thickness distribution data as teacher data. If the prediction data and the teacher data do not match,
A new coupling strength between the corresponding cells of the output layer 14 and the cells of the second hidden layer 13, a new second bias input,
The new coupling strength between the corresponding cells of the second hidden layer 13 and the cells of the first hidden layer 12, the new first bias input, and the cells of the first hidden layer 12 and the cells of the input layer 11 New bond strengths between the counterparts are determined in this order.

Next, in the neural network 10,
The output (prediction data) of the output layer 14 when the shield plate length distribution data is input to the input layer 11 with the coupling strength between the cells and the first and second bias inputs set to new values is obtained. Then, this prediction data is compared with the teacher data. If the two data do not match, the above procedure is repeated until the bond strength between the cells and the first
And the second bias input is optimized.

When the predicted data and the teacher data match, the same work as in the case of the first shield plate is performed for the existing second shield plate having a length distribution different from that of the first shield plate. Done. As a result, the coupling strength between the cells and the first and second bias inputs become more appropriate.
In this way, the neural network 10 for obtaining the predicted copper foil thickness distribution from the shield plate length distribution is constructed.

Next, the operator selects candidate data consisting of R data values of the shielding plate length distribution that enables the production of the electrolytic copper foil having the target thickness distribution, from the neural network 1.
The target copper foil thickness distribution data consisting of S2 data values is input to the optimization system 20. In response to the candidate data input, the neural network 10 includes a first hidden layer 12, a second hidden layer 13, and an output layer 14.
The respective outputs of are obtained in this order. The predicted copper foil thickness distribution data from the output layer 14, which corresponds to the candidate data and includes S2 data values, is output to the evaluation unit 21 of the optimization system 20.

The evaluation unit 21 calculates the sum of squares of deviations between the predicted copper foil thickness distribution data value and the target copper foil thickness distribution data value corresponding to each other, and calculates the predicted copper foil thickness distribution and the target copper foil thickness. The evaluation function value representing the closeness to the distribution is obtained, and this evaluation function value is output to the determination unit 22 of the optimization system 20. Discrimination unit 22
Determines whether the evaluation function value is zero or the minimum value, and if the evaluation function value is not zero or the minimum value, sends a candidate data calculation command to the calculation unit 23 of the optimization system 20.

The calculation unit 23 calculates new candidate data of the shield plate length distribution that reduces the evaluation function value in response to the candidate data calculation command, and outputs the new candidate data to the neural network 10. The neural network 10 obtains new predicted foil thickness distribution data corresponding to the new candidate data and outputs it to the optimization system 20. The optimization system 20 obtains an evaluation function value based on the new predicted copper foil thickness distribution data and the target copper foil thickness distribution data, and then outputs further candidate data for the shield plate length distribution. .

If the evaluation function value becomes zero or the minimum while such data processing is repeated, the discriminator 22 uses the R data values as a basis for calculating the evaluation function value. The candidate data consisting of is output as the optimum data of the shield plate length distribution. The optimum data is stored in the memory of the computer as the optimization system 20, is displayed on the CRT screen, and is printed out.

The operator determines the shield plate length at the point R in the width direction of the shield plate 6 based on the optimum data of the shield plate length distribution obtained as described above. Next, the shield plate 6 having the shield plate length distribution thus obtained is actually manufactured. The shielding plate 6 thus obtained has an optimum thickness distribution for producing the copper foil having the target copper foil thickness distribution by the copper foil producing apparatus as illustrated in FIG.

The present inventor evaluated the prediction accuracy of the above-mentioned neural network 10 by simulation. First, the first method, which was manufactured by trial and error using the conventional method
Through, the shield plate length distribution data for the fifth shield plate and the thickness distribution data of the copper foil manufactured using the first to fifth shield plates were obtained. In collecting the copper foil thickness distribution data, the weight of each of the copper foil pieces having a predetermined area extracted at 26 points in the copper foil width direction is measured, and the measured weight is converted into a unit area weight to extract 26 pieces. The copper foil thickness distribution data value at the site was obtained. FIG. 5 shows copper foil thickness distribution data corresponding to the first to fifth shielding plates. As shown in FIG.
The thickness of the copper foil manufactured using the shielding plate is considerably larger than the target foil thickness and the variation of the foil thickness is also considerably large. The variation of the foil thickness is getting smaller as it gets closer to.

Next, shield plate length distribution data for the first to third shield plates, and copper foil thickness distribution data (copper foil unit area weight distribution data) corresponding to these three types of length distribution data. The neural network 10 was made to learn using. Furthermore, the length distribution data of the fourth shielding plate was input to the learned neural network 10, and the predicted copper foil thickness distribution data corresponding to this length distribution data was obtained by the neural network 10. This predicted data and actual data are shown in FIG.
The error between the predicted data value and the actual data value at 6 points is shown in FIG. As shown in FIG. 7, the prediction error was within ± 2%. That is, the error with respect to the target thickness distribution is ± 2
It was found that three trial productions of the shielding plate were sufficient to produce a copper foil having a content of about 10%.

Next, shield plate length distribution data for the first to fourth shield plates, and copper foil thickness distribution data (copper foil unit area weight distribution data) corresponding to these four types of length distribution data. The neural network 10 was made to learn using. Further, the length distribution data of the fifth shielding plate was input to the learned neural network 10, and the predicted copper foil thickness distribution data corresponding to this length distribution data was obtained by the neural network 10. The predicted data and the actual data are shown in FIG.
FIG. 9 shows the error between the predicted data value and the actual data value at 6 points. As shown in FIG. 9, the prediction error was within ± 1%. That is, the error with respect to the target thickness distribution is ± 1
It was found that it was sufficient to prototype the shielding plate four times in order to produce a copper foil having a content of about 10%.

The present invention is not limited to the above embodiment, but can be variously modified. For example, the configuration of the neural network for carrying out the shield plate manufacturing method of the present invention is not limited to that used in the above embodiments.

[0036]

The thickness distribution of the electrolytic copper foil obtained by peeling the copper electrodeposited on the rotating drum by applying a potential difference between the rotating drum immersed in the electrolytic solution in the electrolytic cell and the electrode plate. In the manufacturing method of the shield plate for adjusting the electrolytic copper foil thickness arranged between the rotating drum and the electrode plate for adjusting the temperature, the manufacturing method according to claim 1, wherein the copper foil thickness predicted from the shield plate length distribution is used. A step of constructing a neural network for determining the thickness distribution based on the length distribution data of the existing shielding plate and the thickness distribution data of the electrolytic copper foil manufactured using the existing shielding plate,
The process of inputting the candidate data of the shield plate length distribution to the neural network, the target copper foil thickness distribution, and the predicted copper foil thickness distribution output from the neural network according to the input of the candidate data are used as the optimization system. Input process,
The process of obtaining the optimum data of the shielding plate length distribution that minimizes the evaluation function value that represents the closeness between the predicted copper foil thickness distribution and the target copper foil thickness distribution by the optimization system, and the target copper foil thickness distribution Since the shield plate for manufacturing the electrolytic copper foil that is provided is manufactured based on the optimum data, it is not necessary to manufacture the shield plate a number of times by trial and error. According to the present invention, the shield plate basically needs to be manufactured only twice. As a result, the labor, cost and time required for the shield plate can be greatly reduced.

[Brief description of the drawings]

FIG. 1 is a schematic view showing an electrolytic copper foil manufacturing apparatus having a foil thickness adjusting shield plate.

FIG. 2 is a schematic plan view illustrating the length distribution of the shielding plate shown in FIG.

FIG. 3 is a conceptual diagram showing a neural network used to carry out a method for manufacturing a shielding plate according to an embodiment of the present invention.

4 is a schematic diagram functionally showing an optimization system used for manufacturing a shielding plate together with the neural network shown in FIG.

FIG. 5 is a graph showing a plurality of types of copper foil thickness distribution data used in a simulation for evaluating the prediction accuracy of a neural network.

FIG. 6 is a graph showing predicted copper foil thickness distribution data obtained by a prediction accuracy evaluation simulation together with actual data.

FIG. 7 is a graph showing an error between the predicted data and the actual data shown in FIG.

FIG. 8 is a graph showing predicted copper foil thickness distribution data obtained by a prediction accuracy evaluation simulation performed under another condition together with actual data.

9 is a graph showing an error between the predicted data and the actual data shown in FIG.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Electrolyzer 2 Electrolyte 3 Rotating drum 4 Electrode plate 5 Distributor 6 Shield plate for foil thickness adjustment 7 Winding roll 8 Pass roll 9 Copper foil 10 Neural network 11 Input layer 12 First hidden layer 13 Second hidden layer 14 Output layer 20 Optimization system 21 Evaluation unit 22 Discrimination unit 23 Calculation unit

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Kenji Sugano, Kenji Sugano 2 601, Ojizawa, Imaichi City, Tochigi Prefecture Furukawa Sark Foil Co., Ltd. Imaichi Plant (72) Inventor Yoshio Hoshi, 601, Ogizawa, Imaichi City, Tochigi Prefecture Furukawa Saki Tofoil Co., Ltd. Imaichi Office

Claims (1)

[Claims] 1. The thickness of an electrolytic copper foil obtained by applying a potential difference between a rotating drum immersed in an electrolytic solution in an electrolytic cell and an electrode plate to peel off the copper electrodeposited on the rotating drum. In the method for manufacturing the shield plate for adjusting the electrolytic copper foil thickness arranged between the rotating drum and the electrode plate to adjust the distribution, a neural network for obtaining a predicted copper foil thickness distribution from the shield plate length distribution is used. A step of constructing based on the length distribution data of the existing shield plate and the thickness distribution data of the electrolytic copper foil manufactured using the existing shield plate, and the candidate data of the shield plate length distribution in the neural network. A step of inputting a target copper foil thickness distribution, and a step of inputting a predicted copper foil thickness distribution output from the neural network according to the input of the candidate data to an optimization system, the predicted copper foil Thickness A step of obtaining optimum data of the shielding plate length distribution that minimizes the evaluation function value indicating the closeness of the distribution and the target copper foil thickness distribution by the optimization system, and an electrolysis having the target copper foil thickness distribution And a step of manufacturing a shielding plate for manufacturing a copper foil based on the optimum data, and a method for manufacturing a shielding plate for adjusting an electrolytic copper foil thickness.