JP2021049778A - Control method of wound diameter distribution in width direction of plastic film-roll, and plastic film-roll - Google Patents

Abstract

To provide a control method of a wound diameter distribution in a width direction of a plastic film-roll wound on a cylindrical core, and a plastic film-roll manufactured by applying the control method.SOLUTION: There is provided a control method of a wound diameter distribution in a width direction of a plastic film-roll, comprising: a recording step to record a measured wound diameter distribution in a width direction, a variable calculated from the wound diameter distribution in the width direction and attribute data on "normal" or "abnormal", and a target wound diameter distribution ft (TD) in the width direction; a learning step in which, from the variable calculated from the recorded wound diameter distribution in the width direction and the attribute data and the attributes associated therewith, a coefficient βn to each variable that optimizes a probability "estimated value" to be classified into each attribute is obtained; a classification step that calculates and classifies the probability "estimated value" in which a newly measured wound diameter distribution in the width direction is categorized into either the "normal" or "abnormal" attribute; and a step to feed back a difference between a plastic film-roll wound diameter distribution and a target wound diameter distribution curve to a thickness control of a base when the data is classified as "abnormal" by the classification step.SELECTED DRAWING: Figure 1

Description

The present invention is a method for controlling the winding diameter distribution in the width direction of a plastic film roll in which a film is wound on a cylindrical core, particularly a resin film roll suitable for a magnetic recording medium, and is a technique for controlling the thickness of a base. The present invention relates to a method of managing the winding diameter in the width direction by learning the correction timing of the difference between the plastic film roll winding diameter distribution (TD, VAL) and the target winding diameter distribution curve _{ft (TD).}

Plastic films are used as base films for magnetic recording tapes, and in recent years, magnetic recording tapes have been in increasing demand for backup applications and archive applications in cloud computing and ICT (Information and Communication Technology). In addition, since the amount of data to be stored has increased dramatically due to the handling of big data, highly reliable and high-density recording technology is being pursued. In order to meet the demand for higher density for the base film of magnetic recording tape, not only the thickness unevenness is uniform, but also the roll diameter distribution in the width direction of the roll wound up is a smooth surface. Is required. For example, a roll having a loose and smooth drum shape in the width direction when the rolled roll is viewed from a direction orthogonal to the axis is required.

As one of the measures against the above requirement, there is a method of measuring the winding diameter distribution in the width direction of the roll in real time and correcting the deviation information from the winding diameter distribution curve of the target by controlling the thickness of the base (Patent Document 1). This method has a small correction time lag and is the most ideal, but it is a large-scale device and therefore has a high production cost.

On the other hand, there is a method (Patent Document 2 and Patent Document 3) in which the measurement result of the winding diameter distribution in the width direction of the film roll is fed back to the adjustment of the die lip temperature and the gap without using a large-scale device.

Japanese Unexamined Patent Publication No. 2002-283438 Japanese Unexamined Patent Publication No. 2017-128080 JP-A-2017-157261

However, in these methods, the thickness correction timing of the base is a sensual evaluation of the engineer based on many years of experience, and there is a concern that the judgment delay due to the change of the engineer may cause a deviation from the standard, which is an issue of technology succession. It was.

The present invention is a method for controlling the winding diameter distribution in the width direction of a plastic film roll in which a film is wound on a cylindrical core, particularly a resin film roll suitable for a magnetic recording medium, by solving the problems of the prior art. Learn the correction timing of the difference between the plastic film roll winding diameter distribution (TD, VAL) and the target winding diameter distribution curve _ft (TD), which was performed by engineers to control the thickness of the base, and learn the correction timing in the width direction. The task is to manage the winding diameter.

In order to solve the above object, the present invention has the following configuration.
(1) A method for controlling the winding diameter distribution in the width direction of a plastic film roll in which a film is wound on a cylindrical core.
Measured width direction winding diameter distribution (width position: TD, winding diameter: VAL), variables calculated from the width direction winding diameter distribution, data of "normal" and "abnormal" attribute classes, and the target width direction a recording step of recording the winding diameter distribution _f t (TD),
Each that optimizes the variable calculated from the recorded width direction winding diameter distribution and the data of the "normal" and "abnormal" attribute classes, and the probability "estimated value" of being classified into each attribute class from the associated attribute class. A learning step to find the coefficient β _{n over a variable,
From the coefficient β n} obtained in the learning step and Equation 3, the probability that the newly measured widthwise winding diameter distribution (TD, VAL) belongs to either the “normal” or “abnormal” attribute class is “estimated value”. And the classification step to calculate and classify
When classified as "abnormal" by the classification step, the reflection step of feeding back the difference between the plastic film roll winding diameter distribution (TD, VAL) and the target winding diameter distribution curve _ft (TD) to the thickness control of the base. When,
A plastic film roll width direction winding diameter distribution management method.

(2) The learning step is
Let X1 be the maximum value of the difference between the straight line f (TD) connecting both ends of the winding diameter distribution curve and the winding diameter distribution curve larger than the straight line connecting both ends of the winding diameter distribution curve.
Let X2 be the maximum value of the difference between the straight line f (TD) connecting both ends of the winding diameter distribution curve and the winding diameter distribution curve smaller than the straight line connecting both ends of the winding diameter distribution curve.
Let Y1 be the maximum value of the difference from the winding diameter distribution curve from the _{line f convex} (TD) connecting the adjacent convex and convex vertices of the winding diameter distribution curve.
Let Y2 be the maximum value of the difference from the winding diameter distribution curve from the _{line f concave} (TD) connecting the adjacent concave and concave bottom points of the winding diameter distribution curve.
Let Y5 be the difference between the diameters at both ends of the winding diameter distribution curve.
Let Y6 be the difference between the maximum diameter and the minimum diameter of the winding diameter distribution curve.
Winding diameter distribution curve (TD, VAL) and the target take-size distribution curve _f t (TD) winding diameter of the target plus a constant value k so that the area difference is the minimum of the distribution curve _f t (TD) + k and winding When the area difference of the diameter distribution curve (TD, VAL) is S,
A variable is at least one selected from the group consisting of X1, X2, Y1, Y2, Y5, Y6 and S.
The method for managing the winding diameter distribution in the width direction of a plastic film roll according to (1), wherein the attribute class is used as an objective variable.
(3) the target width direction winding diameter distribution _f t (TD) is 1 or more in the width direction winding diameter distribution (TD, VAL) obtained by averaging from, according to (1) or (2) Plastic Film roll width direction winding diameter distribution management method.
(4) A tentative widthwise winding diameter distribution (TD) synthesized from a plurality of VAL fluctuation periods and VAL fluctuation widths calculated from the plurality of measured widthwise winding diameter distributions (TD, VAL) in the recording step. , VAL), and the created temporary width direction winding diameter distribution (TD, VAL) is recorded. The plastic film roll width direction winding diameter distribution management method according to any one of (1) to (3).
(5) The plastic film roll width direction winding diameter distribution management method according to any one of (1) to (4), wherein the plastic film roll described above is a resin film roll for a magnetic recording medium.
(6) A plastic film roll produced by applying the method for controlling the winding diameter distribution in the width direction according to any one of (1) to (5).

According to the present invention, it is possible to control the winding diameter distribution in the width direction of a plastic film roll in which a film is wound on a cylindrical core, particularly a resin film roll suitable for a magnetic recording medium. In particular, the timing of correcting the difference between the plastic film roll winding diameter distribution (TD, VAL) and the target winding diameter distribution curve _{ft (TD), which was performed by engineers to control the thickness of the base, was learned and wound.} It is possible to manage the pass / fail of the figure as a sign.

It is the schematic of the plastic film roll width direction winding diameter distribution management method which concerns on one Embodiment of this invention. It is the schematic of the straight line f (TD) calculated from the width direction winding diameter distribution (TD, VAL) of the plastic film roll which concerns on one Embodiment of this invention. It is the schematic of the _{straight line f convex} (TD) calculated from the width direction winding diameter distribution (TD, VAL) of the plastic film roll which concerns on one Embodiment of this invention. It is the schematic of the _{straight line f concave} (TD) calculated from the width direction winding diameter distribution (TD, VAL) of the plastic film roll which concerns on one Embodiment of this invention. Widthwise winding diameter distribution of the plastic film roll according to one embodiment of the present invention (TD, VAL) and a schematic diagram of a linear f _{t (TD).} It is the schematic of the plastic film roll width direction winding diameter distribution measuring method which concerns on one Embodiment of this invention. It is the schematic of the plastic film roll width direction winding diameter distribution synthesis method which concerns on one Embodiment of this invention. It is the schematic of the ROC curve which concerns on one Embodiment of this invention. It is a schematic diagram of the synthesis of (VAL fluctuation period, VAL fluctuation width) according to one embodiment of the present invention.

Hereinafter, desirable embodiments for carrying out the present invention will be described with reference to the drawings, but the present invention is not limited thereto.

The present invention is a plastic film roll width direction winding diameter distribution management method in which a film is wound on a cylindrical core.
Measured width direction winding diameter distribution (width position: TD, winding diameter: VAL), variables calculated from the width direction winding diameter distribution, data of "normal" and "abnormal" attribute classes, and the target width direction a recording step of recording the winding diameter distribution _f t (TD),
Each that optimizes the variable calculated from the recorded width direction winding diameter distribution and the data of the "normal" and "abnormal" attribute classes, and the probability "estimated value" of being classified into each attribute class from the associated attribute class. A learning step to find the coefficient β _{n over a variable,
From the coefficient β n} obtained in the learning step and Equation 3, the probability that the newly measured widthwise winding diameter distribution (TD, VAL) belongs to either the “normal” or “abnormal” attribute class is “estimated value”. And the classification step to calculate and classify
Feedback by the classification step, if it is classified as "abnormal", the width direction winding diameter distribution of the plastic film roll (TD, VAL) and the difference between the winding of the target size distribution curve f _{t (TD)} to control the thickness of the base Reflection steps and
It is characterized by having.

Here, as the cylindrical core, paper, plastic, or the like can be used as the material, but it is preferable to use fiber reinforced plastic from the viewpoint of strength. Examples of the fiber-reinforced plastic core include a core in which carbon fiber or glass filament is wound to form a cylindrical shape, which is impregnated with a thermosetting resin such as an unsaturated polyester resin and cured.

The cylindrical core preferably has an elastic modulus in the width direction of 13 GPa or more. If a core that is less than the hanging range is used, the core may be deformed due to the tension and contact pressure applied when winding the plastic film.

The plastic film of the present invention refers to a sheet-like material molded from a thermoplastic polymer or a polymer dissolved in a solvent, and examples thereof include a polyester film, an aromatic polyamide film, a polypropylene film, a nylon film, and a polycarbonate film. The thickness is not limited. In particular, the present invention is suitable for producing a thin film. The thin film here means 1 μm or more and 20 μm or less. If it is less than 1 μm, handling may be difficult and it may not be suitable for processing. Further, a film exceeding 20 μm may not be suitable for a magnetic recording medium that requires thinning.

Further, an unstretched film or a uniaxially stretched film may be used, but it is particularly suitable for producing a biaxially oriented film stretch-oriented in the longitudinal direction and the width direction.

In the present invention, the plastic film roll has a form in which the above-mentioned plastic film is wound around a cylindrical core.

The manufacturing method of the plastic film roll is not particularly limited, but the thickness of the continuously formed running film is measured with high accuracy, the widthwise winding diameter distribution of the wound film roll is measured, and the target winding is performed. A method of adjusting the film thickness by feeding back to the adjustment of the die lip temperature and the gap so as to satisfy the diameter shape is preferable.

In the present invention, the winding diameter distribution in the width direction of the plastic film roll can be measured by using a stylus type or a non-contact laser type measuring instrument. As the stylus type, for example, a bulk shape measuring instrument manufactured by Hamano Seiki Co., Ltd. can be applied.

In the winding diameter distribution measurement, the positions in the circumferential direction are measured at five points at equal intervals in the film roll circumferential direction at positions excluding the film unwinding position. This is because it is not possible to determine an error due to wrinkles on the surface of the film roll at the time of measurement at one location.

Further, in the winding diameter distribution measurement, the width of the plastic film roll records the winding diameter distribution 10 mm inside from both ends. This is because the edge of the film roll becomes noise due to the influence of the high edge.

The winding diameter distribution obtained by the measurement can be obtained by the average value of the winding diameter displacement VAL for each TD position.

Here, the position range in the width direction of the winding diameter distribution measurement will be described with reference to FIG.

First, the measurement preferably starts at the end A of the film roll and ends at the end B of the film roll in the opposite direction.

When it was difficult to start from the end, the measurement was started from the measurement start position A inside the film roll from the end A at the same position in the circumferential direction of the film roll, and measured to the end B of the film roll in the opposite direction in the measurement direction A. Later, the measurement is performed from the measurement start position B inside the inside of the film roll, which has just started measurement, to the end portion A of the film roll, and to the end portion A of the film roll in the measurement direction B opposite to the previous one.

Next, a method of synthesizing the winding diameter distribution in the width direction obtained in the measurement direction A and the measurement direction B will be described with reference to FIG. 7.

When the width of the measured plastic film roll is, for example, 500, the start position of the measurement direction A is 50 from the end, and the start position of the measurement direction B is 100, the respective width direction winding diameter distributions obtained are It becomes like "7-1".

Next, the TD of the widthwise winding diameter distribution (TD _A , VAL _A ) in the measurement direction A is uniformly set to +50 to obtain the _{measurement direction A'(TD A'} , VAL _A'). Further, the TD of the widthwise winding diameter distribution (TD _B , VAL _B ) in the measurement direction B is multiplied by a minus, and _{the maximum value of the TD B} is uniformly added to obtain the measurement direction B'(TD _B' , VAL _B'). .. The winding diameter distribution in the width direction is the measurement direction A'and the measurement direction B'as in "7-2". Next, the measuring direction _{B '(TD B', VAL} B ') adding _{VAL TDAmin} of TD _A minimum of uniform the measuring direction A to VAL of' corresponds to the TD _A minimum value of the measurement direction B 'VAL Is subtracted to obtain the measurement direction B ″ (TD _{B ″} , VAL _{B ″} ). The winding diameter distribution in the width direction becomes the measurement direction A'and the measurement direction B'as shown in "7-3" and is combined.

The recording step of the present invention refers to the widthwise winding diameter distribution of each plastic roll measured by the winding diameter distribution measuring method, and the unique number of the plastic film roll. Variables X1, X2, Y1, Y2, Y5, Y6, S, attributes calculated from the array data (TD, VAL) with the displacement of the winding diameter for each width position as VAL and the winding diameter distribution in the width direction. "Normal" and "abnormal" data are stored as a class.

Table 3 shows an example of the recorded sequence data (TD, VAL), and an example of the variables X1, X2, Y1, Y2, Y5, Y6, S, and the attribute class recorded in Table 1.

In the recording step of the present invention, a tentative widthwise winding diameter distribution (TD) synthesized from a large number of VAL fluctuation periods and VAL fluctuation widths calculated from a plurality of measured widthwise winding diameter distributions (TD, VAL). , VAL) is preferably created and recorded.

A large amount of data is required for the learning step described later. If the amount of data to be trained is small, not only the accuracy of the plastic film roll width direction winding diameter distribution management method cannot be obtained, but also the data used for training is adapted with high accuracy, resulting in overfitting that does not conform to unknown data. Sometimes.

The data preferably exists at least 20 times or more, more preferably 50 times or more, the number of variable items calculated from the width direction winding diameter distribution.

When the measured data is small, it is difficult to obtain the judgment formula as it is, but by Fourier transforming the measured multiple width direction winding diameter distributions (TD, VAL), a large number (VAL fluctuation period, VAL fluctuation width) can be obtained. ) Can be generated (obtained).

From the large number of obtained (VAL fluctuation cycle, VAL fluctuation width), 20 are randomly selected in descending order of fluctuation width, and at least 3 or more, and at most 10 or less are selected, and the winding diameter in the width direction to be the target described later distribution _f t (TD) in the width direction winding diameter distribution of temporary by combining (TD, VAL) can be obtained. Further, when synthesizing the temporary winding diameter distribution (TD, VAL) in the width direction, the VAL fluctuation widths of a large number (VAL fluctuation period, VAL fluctuation width) obtained are 0 at the center and standard deviation = fluctuation width ×. Random selection with a normal distribution of 1 to 3 is preferable because more various tentative width direction winding diameter distributions (TD, VAL) can be obtained.

An example is shown in FIG.

The obtained provisional width direction winding diameter distribution (TD, VAL) was obtained in the recording step in the same manner as the measured width direction winding diameter distribution (TD, VAL), and the unique No. of the plastic film roll was obtained. Variables X1, X2, Y1, Y2, Y5, Y6, S, attributes calculated from the array data (TD, VAL) with the displacement of the winding diameter for each width position as VAL and the winding diameter distribution in the width direction. "Normal" and "abnormal" data are stored as a class.

The recording step of the present invention, the width becomes a target direction winding diameter distribution f _{t (TD)} is stored.

The width direction winding diameter distribution f _t to be the target _(TD) is preferably from 1 or more in the width direction winding diameter distribution (TD, VAL) is obtained by averaging.

Further, the widthwise winding diameter distribution f _{t (TD),} the width when viewed rolled up rolls required for the recent high-density recording capacity for a base film for magnetic recording tape from a direction perpendicular to the axis A drum-shaped widthwise winding diameter distribution with a loose and smooth directional winding diameter distribution is preferable. More preferably, select was good film roll by coating unevenness condition of the magnetic tape processing, be the widthwise winding distribution selected as target width direction winding diameter distribution f _{t (TD),} the future It is preferable because it can quickly respond to the required technology.

Next, the variable X1 in the present invention will be described with reference to FIG. The variable X1 in the present invention is the maximum value obtained by calculating the difference between the recorded widthwise winding diameter distribution (TD, VAL) and the straight line f (TD) by the following formula.

+ {VAL (TD) -f (TD)}
VAL (TD): VAL value of position TD in the width direction
f (TD): Straight line f value of the width direction position TD The variable X2 in the present invention is the difference between the recorded width direction winding diameter distribution (TD, VAL) and the straight line f (TD) calculated by the following formula. , The maximum value obtained.

-{VAL (TD) -f (TD)}
VAL (TD): VAL value of position TD in the width direction
f (TD): The position in the width direction TD of the straight line f value where the straight line f (TD), the recorded width direction winding diameter distribution (TD, VAL), the point _{(TD min, VAL TDmin) (} TD max, It is a linear type f (TD) connecting two points of VAL _TDmax).

The variable Y5 in the present invention is the absolute value difference between _{VAL TDmin} and VAL _TDmax with respect to the recorded widthwise winding diameter distribution (TD, VAL).

The variable Y6 in the present invention is the difference between the maximum value of VAL and the minimum value of VAL with respect to the recorded widthwise winding diameter distribution (TD, VAL).

The variable Y1 in the present invention is the maximum value obtained by calculating the difference between the recorded widthwise winding diameter distribution (TD, VAL) and the f- _{convex (TD) by the following formula.}

-{VAL (TD) -f _convex (TD)}
VAL (TD): VAL value of position TD in the width direction
f- _convex (TD): Line f_convex value connecting the convex vertices of the width direction position TD Here, the line f- _convex (TD) is dVAL (TD) in the recorded width direction winding diameter distribution (TD, VAL). ) / DTD is 0 and d (dVAL (TD) / dTD) / dTD is a positive value, which is a convex vertex, and the convex vertices are connected by a straight line.

The variable Y2 in the present invention is the maximum value obtained by calculating the difference between the recorded width direction winding diameter distribution (TD, VAL) and the f _{concave (TD) by the following formula.}

+ {VAL (TD) -f _concave (TD)}
VAL (TD): VAL value of position TD in the width direction
f _concave (TD): Line f_concave value connecting the concave bottom points of the width direction position TD Here, the line f _concave (TD) is dVAL (TD) in the recorded width direction winding diameter distribution (TD, VAL). ) / DTD is 0 and d (dVAL (TD) / dTD) / dTD is a negative value, which is a concave bottom point, and is a line connecting the concave bottom points with a straight line.

The variable S in the present invention, it is recorded widthwise winding diameter distribution (TD, VAL) and the target roll diameter value of the area difference S obtained calculated from the equation 1 of the distribution curve f _{t (TD)} + k.

Widthwise winding diameter distribution of the target in the present invention f _{t (TD)} is, when viewed rolled up rolls required for the recent high-density recording capacity for a base film for magnetic recording tape from a direction perpendicular to the axis A drum-shaped winding diameter distribution in the width direction is preferable because the winding diameter distribution in the width direction is loose and smooth. More preferably, select was good film roll by coating unevenness condition of the magnetic tape processing, be the widthwise winding distribution selected as target width direction winding diameter distribution f _{t (TD),} the future It is preferable because it can quickly respond to the required technology.

Here, the constant value k, a mean value _{VAL ave} of VAL of recorded widthwise winding diameter distribution (TD, VAL), the average value f a data target winding diameter distribution curve _f t in (TD) (TD) _This is the difference from the table.

k = VAL _ave −f _tave
The attribute class in the present invention is the unique No. of the recorded plastic film roll. It is an objective variable that distinguishes between "normal" and "abnormal" for each. Although not particularly limited, it is preferable to input with "0" and "1" from the viewpoint of obtaining the coefficient β multiplied by each variable that optimizes the probability "estimated value" to be classified into each attribute class.

The learning step of the present invention is the data of the variable calculated from the recorded widthwise winding diameter distribution and the attribute class of "normal" and "abnormal", and more specifically, the unique No. of the plastic film roll. _{From the objective variable for each and the data in which the variables are stored, the coefficient β n of} each variable is obtained in order to optimize the probability “estimated value” of being classified into each attribute class.

Since the learning step of the present invention aims to classify as "normal" or "abnormal", it is possible to use a neighbor method, logistic regression, a support vector machine, a decision tree, a neural network, or the like as the method. it can. Decision trees and neural networks have high classification accuracy, but optimization calculation formulas are complicated. Neighboring method, logistic regression, and support vector machine are preferable because the optimization calculation formula can be simplified. More preferably, it is a logistic regression in which the structure of the calculation formula is simplified.

In this learning step, it is preferable that at least one selected from the group consisting of X1, X2, Y1, Y2, Y5, Y6 and S described above is used as a variable and the attribute class is used as the objective variable.

Hereinafter, the specific contents of the learning step will be described based on logistic regression.

The purpose of the learning is to obtain the coefficient β multiplied by each variable that optimizes the probability "estimated value" of "normal" and "abnormal" classified by the attribute class.

In general regression analysis, both the objective variable and the variable are continuous variables, whereas in the present invention, the objective variable is a labeled variable of "normal" or "abnormal". Set the label type variable to 0 for "normal" and 1 for "abnormal", and consider the probability that "abnormal" will occur. The probabilities range from 0% to 100% and the logit function is adapted to make it a continuous variable. The expression obtained by expanding this is the expression 3. The goal is to optimize β in Equation 3.

Next, the definition of optimization will be described.

Likelihood (predicted value) is
Likelihood = estimated value ^{attribute class} × (1-estimated value) ^{1-attribute class}
Will be.

If this is applied to all data

The value of the coefficient β is determined so that this equation is maximized.

However, multiplication is complicated to calculate on a computer, and when the total power of probabilities is calculated, the value approaches 0 infinitely, and it may be calculated as 0. Therefore, in order to deal with this problem, the logarithmic equation 2 is used.

The value of the coefficient β is determined so that the sum of all the data of the log-likelihood is maximized.

Next, a method for evaluating the versatility of the _{coefficient β n} multiplied by each of the obtained variables will be described.

Since the prepared data may be easily learned or predicted by chance, it is preferable to adopt the cross-validation method as the versatility evaluation method.

The cross-validation method is performed according to the following procedure.

First, the data is divided into M pieces.

Next, m-1 pieces are learned as learning data, and the m-th data is evaluated for versatility of the optimization calculation formula obtained as versatility evaluation data.

This is repeated for m = 1, 2, ... M.

The average of the obtained versatility evaluation results is used as the versatility evaluation result of the optimization calculation formula.

Generally, the number of divisions M is preferably 5 to 10.

Here, the calculation formulas for the correct answer rate, the precision rate, the recall rate, and the F value, which are the items of the versatility evaluation method, are described below.

Correct answer rate = (TP + TN) / (TP + TN + FP + FN)
Conformity rate = TP / (TP + FP)
Reproducibility = TP / (TP + FN)
F value = 2 x (compliance rate x reproducibility) / (compliance rate + reproducibility)
TP = number of data with positive true value and positive predicted value TN = number of data with negative true value and negative predicted value FP = number of data with negative true value and positive predicted value FN = true value Number of data with positive and negative predicted value The classification step of the present invention is based on the newly measured widthwise winding diameter distribution (TD, VAL) using _{the coefficient β n obtained in the learning step and Equation 3.} From the calculated variables, the probability "estimated value" belonging to the attribute class is calculated and classified.

Since the mass-produced plastic film rolls have high process capability and the widthwise winding diameter distribution of the plastic film rolls is "normal" with high probability, it was obtained in the learning step with the attribute class "normal" set to 0. The probability “estimated value” calculated by the coefficient β _n and the equation 3 approximates the γ distribution in which the average of the data number distributions for the attribute class “normal” is close to 0.

Therefore, it is possible to classify "abnormal" by issuing a warning before deviating from "normal" by setting a target threshold value from the probability of not being "normal" obtained from the γ distribution.

On the contrary, the plastic film roll produced in a small amount has a low process capability, its width direction winding diameter distribution has a low probability of being "normal", and the provisional width direction winding diameter distribution (TD, VAL) according to the present invention. Is randomly synthesized, so the obtained probability "estimated value" and the data number distribution of the attribute class "normal" may not approximate the γ distribution whose average is close to 0.

Therefore, it is desirable to optimize the threshold value for determining "normal" or "abnormal" from the probability "estimated value".

As a method of optimizing the threshold value, there is a method of maximizing the AUC (area under the curve) using the ROC curve (receiver operating characteristic curve).

Here, the ROC curve is a graph in which the "estimated values" are arranged in descending order, and the 1-specificity is plotted on the x-axis and the recall is plotted on the y-axis. An example is shown in FIG.

It can be said that the closer the RCO curve is to the upper left of the graph, the more accurately the classification can be performed.

Here, the specificity is a value obtained by TN / (FP + TN).

Further, as a method of optimizing the threshold value, there is a method of minimizing BER (Balanced Error Rate).

Here, BER is a value obtained by 0.5 × (FP / (TN + FP) + FN / (FN + TP)).

In the reflection step of the present invention, when the classification step is classified as "abnormal", the difference between the plastic film roll winding diameter distribution (TD, VAL) and the target winding diameter distribution curve ft (TD) is controlled by the thickness of the base. It is to give feedback to.

The feedback to the thickness control of the base is the width of the plastic film roll winding diameter distribution (TD, VAL) and the target winding diameter distribution curve ft (TD) in the same manner as in Patent Document 1 (Japanese Unexamined Patent Publication No. 2002-283438). The deviation target value per plastic film in each direction is obtained and set as the target value of the parameter for controlling the gap of the base.

At this time, it is preferable to obtain the plastic film roll winding diameter distribution (TD, VAL) by adopting the VAL average value around the width direction by the adjacent averaging method for the plastic film roll winding diameter distribution (TD, VAL).

By executing each of the above steps, it is possible to manage the winding diameter distribution in the width direction of the plastic film roll, and it is possible to manage the pass / fail of the wound shape as a sign. In other words, it is possible to prevent the occurrence of rejected products in roll form by catching the signs and feeding back information to the upstream part of the process such as the base before the actual failure products occur. It is possible to obtain a plastic film roll with a good winding shape with high efficiency.

Hereinafter, the present invention will be described with reference to Examples, but the present invention is not limited to these Examples.
(Plastic film roll A in which a film is wound on a cylindrical core)
In N-methyl-2-pyrrolidone (hereinafter NMP), 2-chloroparaphenylenediamine (hereinafter CPA) corresponding to 85 mol% and diphenyl ether (hereinafter DPE) corresponding to 15 mol% are dissolved, and this solution is filtered with precision. After filtering through a filter made of 1.0 μm polypropylene, the particles were transferred to a polymerization tank, which was passed through a filter made of polypropylene having a filtration accuracy of 1.0 μm, which corresponds to 98.5 mol% of chloroterephthalic acid chloride (hereinafter referred to as chloroterephthalic acid chloride). CTPC) is added to add 0.02% by mass of silica particles having an average particle size of 80 nm to the aromatic diamine component and 6.0% by mass of organic particles having an average particle size of 100 nm to the aromatic diamine component before polymerization. It was added so as to be%, and the mixture was stirred at 30 ° C. or lower for 2 hours to obtain a polymerized polymer.

Next, 98.5 mol% of lithium carbonate was added to hydrogen chloride in the polymer polymer for 4 hours of neutralization, and 10 mol% of triethanolamine was added to hydrogen chloride in the polymer polymer. Then, the mixture was stirred for 1 hour to obtain an aromatic polyamide solution A having a polymer concentration of 10.8% by mass.

Further, CPA corresponding to 85 mol% and DPE corresponding to 15 mol% were dissolved in NMP, and this solution was filtered through a filter made of polypropylene having a filtration accuracy of 1.0 μm, and then transferred to a polymerization tank. CTPC corresponding to 98.5 mol% passed through a filter made of polypropylene having a filtration accuracy of 1.0 μm was added thereto, and the mixture was stirred at 30 ° C. or lower for 2 hours to obtain a polymerized polymer.

Next, 98.5 mol% of lithium carbonate was added to hydrogen chloride in the polymer polymer to neutralize it for 4 hours, and silica having an average primary particle size of 16 nm (“AEROSIL” manufactured by Nippon Aerosil Co., Ltd.” R972 type) was added to the polymer in an amount of 1.6% by mass and stirred for 1 hour, and then 10 mol% of triethanolamine was added to hydrogen chloride in the polymer polymer and stirred for 1 hour. This was carried out to obtain an aromatic polyamide solution B having a polymer concentration of 10.8% by mass.

Next, the aromatic polyamide solution A is passed through a metal fiber filter made of stainless steel having a filtration accuracy of 1.2 μm, and the aromatic polyamide solution B is passed through a metal fiber filter made of stainless steel having a filtration accuracy of 5.0 μm, and then inside the mouthpiece. Laminated with. At this time, the lamination ratio (thickness ratio) of the aromatic polyamide solutions A and B was set to 50% each. Next, the laminated polymer solution is cast on a stainless steel belt having a mirror-like surface, and heated with hot air at 180 ° C. for 2 minutes on the belt on the downstream side in the traveling direction from the mouthpiece to evaporate the solvent to improve self-retention. The obtained film body was continuously peeled off from the belt.

Next, the membrane was introduced into the tank of the solvent extraction treatment apparatus and stretched 1.15 times in the MD direction while water-extracting the residual solvent and the inorganic salt generated by neutralization.

Then, the obtained film body was gripped by a clip of a transverse stretching machine to dry and heat-treat the moisture, and then dried and heat-treated at 245 ° C. for 1.5 minutes and stretched 1.43 times in the TD direction. , The film was slowly cooled at a rate of 20 ° C./sec, the film edge was removed, and the film was wound on a core with an oscillation of 60 mm to obtain a jumbo roll of an aromatic polyamide film having an average thickness of 3.6 μm.

Next, a slitter is used to slit into a small width (1,014 mm) at a winding speed of 30 m / min, a winding surface pressure of 200 N / m, and a winding tension of 100 N / m, and the cylindricity is 0.05 mm and the elastic modulus in the width direction is 14. It was wound into a roll on a cylindrical core of a fiber reinforced plastic (FWP) core A (manufactured by Tenryu Industries Co., Ltd.) FWP10) having a flexural strength of .5 GPa, a bending strength of 210 MPa, a surface hardness of 85, an outer diameter of 167 mm, and an inner diameter of 152.5 mm. ..

Similarly, a total of 1025 plastic film rolls A in which a film was wound on a cylindrical core were collected.

(Plastic film roll B in which a film is wound on a cylindrical core)
The MD direction stretching ratio in the solvent extraction treatment device was 1.05 times, the TD direction stretching ratio in the transverse stretching machine was 1.08 times, and the heat treatment in the transverse stretching machine was performed at 280 ° C. for 4.3 minutes. A jumbo roll of an aromatic polyamide film having an average thickness of 12 μm was obtained and slit into a narrow width (660 mm) with a slitter at a winding speed of 20 m / min, a winding surface pressure of 300 N / m, and a winding tension of 100 N / m. Was performed in the same manner as (Plastic film roll A in which a film was wound on a cylindrical core).

Similarly, a total of 30 plastic film rolls B in which a film was wound on a cylindrical core were collected.

(Measurement of winding diameter distribution in the width direction of a plastic film roll)
All of the 1025 plastic film rolls A and 30 plastic film rolls obtained above were placed inside the film roll end A using a raw fabric shape measuring instrument (bulk shape measuring instrument) manufactured by Hamano Seiki Co., Ltd. After starting from the measurement start position A of 50 mm and measuring to the end B of the film roll in the opposite direction in the measurement direction A, from the measurement start position B 100 mm inside the inside of the film roll that started the measurement earlier to the end A of the film roll. Measured up to the end A of the film roll in the measurement direction B, which is the opposite direction.

Next, the TD of the width direction winding diameter distribution (TD _A , VAL _A ) in the measurement direction A is uniformly +50, and the TD of the width direction winding diameter distribution (TD _B , VAL _B ) in the measurement direction B is negatively added. Further, _{the maximum value of TD B} was uniformly added to obtain the winding diameter distribution in the width direction (TD _A' , VAL _A' ) and (TD _B' , VAL _B' ).
Next, the VAL TDAmin of the TD _{A minimum value in the measurement direction A'is uniformly} added to the VAL of the width direction winding diameter distribution (TD _B' , VAL _B' ), and corresponds to _{the TD A} minimum value in the measurement direction B'. VAL is subtracted to _{obtain (TD B "} , VAL _B" ), and the widthwise winding diameter distributions (TD _A' , VAL _A' ) and (TD _{B "} , VAL _B" ) are combined to obtain the widthwise winding diameter distribution (TD B ", VAL B"). TD, VAL) was obtained.

Similarly, the winding diameter distribution was measured in the width direction at five locations excluding the film unwinding position at intervals of 60 degrees in the circumferential direction of the film roll, and the average value was obtained. Specifically, the linear gauge, which is the detector of the measuring device, was brought into contact with the roll surface and traveled on a dedicated rail at a speed of 12.5 mm / sec for measurement. Data from the linear gauge is set to zero at intervals of 0.01 seconds using a digital recorder, the displacement of the winding diameter for each width position is set to VAL, and the TD position is determined by the following formula, and the data (TD). , VAL) was recorded in the shared server in csv format.

TD [mm] = measurement time [[sec] x speed 12.5 [mm / sec]]
(Straight line f (TD))
The data recorded in the shared server (TD, VAL), was calculated point _{(TD min, VAL TDmin) (} TD max, VAL TDmax) connecting two points of the linear equation f (TD). The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.

(Tentative winding diameter distribution in the width direction)
Each of the data (TD, VAL) obtained by the measurement of the plastic film roll B was subjected to Fourier transform to obtain the top 20 fluctuation widths (VAL fluctuation period, VAL fluctuation width). An example is shown in Table 4.

From the top 20 (VAL fluctuation cycle, VAL fluctuation width) of each of the obtained 30 fluctuation widths, 6 were randomly selected and selected from those whose VAL fluctuation cycle was larger than the cycle range (VAL fluctuation cycle, VAL fluctuation width,). the VAL variation width) by combining the target width direction winding diameter distribution f _{t (TD),} to obtain a provisional widthwise winding distribution of the plastic film roll B 1000 duty.

Here, the periodic range is a value obtained by doubling the sampling period of TD of data (TD, VAL).

Further, when synthesizing the temporary winding diameter distribution (TD, VAL) in the width direction, the VAL fluctuation width obtained (VAL fluctuation cycle, VAL fluctuation width) is set to 0 at the center and standard deviation = fluctuation width × 3. A tentative width direction winding diameter distribution (TD, VAL) was obtained by randomly selecting from the normal distribution.

The obtained data (TD, VAL) for 1000 temporary winding diameter distributions in the width direction were recorded in the shared server in the csv format.
(Variable X1)
For the recorded data (TD, VAL), the difference from f (TD) was calculated by the following formula, and the maximum value of the obtained numerical value was defined as X1. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.

+ {VAL (TD) -f (TD)}
VAL (TD): VAL value of position TD in the width direction
f (TD): Straight line f value of position TD in the width direction (variable X2)
For the recorded data (TD, VAL), the difference from f (TD) was calculated by the following formula, and the maximum value of the obtained numerical value was defined as X2. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.

-{VAL (TD) -f (TD)}
VAL (TD): VAL value of position TD in the width direction
f (TD): Straight line f value of position TD in the width direction (line f _convex (TD))
In the recorded data (TD, VAL), the point where dVAL (TD) / dTD is 0 and d (dVAL (TD) / dTD) / dTD is a positive value is defined as the convex vertex, and the convex vertex. _{The line f convex} (TD) connecting the above with a straight line was calculated. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll.
(Line f _concave (TD))
In the recorded data (TD, VAL), the point where dVAL (TD) / dTD is 0 and d (dVAL (TD) / dTD) / dTD is a negative value is defined as the bottom point of the concave, and the concave _{The line f concave} (TD) connecting the base points with a straight line was calculated. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll.
(Variable Y1)
For the recorded data (TD, VAL), _{the difference from the f-convex} (TD) was calculated by the following formula, and the maximum value of the obtained numerical value was defined as Y1. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.

-{VAL (TD) -f _convex (TD)}
VAL (TD): VAL value of position TD in the width direction
f- _convex (TD): Line f_convex value connecting the convex vertices of the width direction position TD (variable Y2)
For the recorded data (TD, VAL), _{the difference from the f concave} (TD) was calculated by the following formula, and the maximum value of the obtained numerical value was defined as Y2. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.

+ {VAL (TD) -f _concave (TD)}
VAL (TD): VAL value of position TD in the width direction
f _concave (TD): Line f_concave value connecting the concave bottom points of the width direction position TD (variable Y5)
For the recorded data (TD, VAL), the absolute value difference between _{VAL TDmin} and VAL _{TDmax was defined as Y5.} The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.
(Variable Y6)
For the recorded data (TD, VAL), the difference between the maximum value of VAL and the minimum value of VAL was defined as Y6. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.
(Of the target width direction winding diameter distribution _f t (TD))
Regarding the recorded data (TD, VAL), 10 data are arbitrarily selected from the data (TD, VAL) judged to be "normal" as the target winding diameter distribution of each of the plastic film roll A and the plastic film roll B. The selected data was used as each target data. For each target data, the VAL values at each TD position were averaged and the target data _ft (TD) was recorded.
(Constant value k)
Recorded data (TD, VAL) and the average value _{VAL ave} of VAL of the difference between the average value _{f tave} data target winding diameter distribution curve _f t in (TD) (TD) was k. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll.

k = VAL _ave −f _tave
(Variable S)
The area difference S of recorded data (TD, VAL) a target winding diameter distribution curve _f t (TD) was calculated from Equation 1. The data 10 mm from both ends was deleted in order to exclude the influence of the high edge at the end of the film roll. An example is shown in Table 1.

(Attribute class judgment)
Regarding the recorded width direction winding diameter distribution (TD, VAL) 1025 data of 1025 plastic film rolls, the width direction winding diameter distribution data exceeding the standard range and the width direction winding diameter distribution produced immediately before exceeding the standard range The attribute class of 218 data of the width direction winding diameter distribution data for which feedback to the data and the automatic base was performed was set as "abnormal". An example is shown in Table 1.

Similarly, the width direction winding diameter distribution (TD, VAL) 30 data of 30 plastic film rolls B and the provisional width direction winding diameter distribution (TD, VAL) 1000 data of the plastic film roll B exceeded the standard range. Width direction winding diameter distribution data, width direction winding diameter distribution data produced immediately before exceeding the standard range, and temporary width direction winding similar to the width direction winding diameter distribution produced immediately before exceeding the standard range. The attribute class of the diameter distribution was set to "abnormal".

The "similarity" here will be described.

When multiple people view the width direction winding diameter distribution data produced immediately before and the temporary width direction winding diameter distribution side by side, if more than half of the people judge that they are "similar", they are said to be "similar". did.
(Learning step)
For the widthwise winding diameter distribution (TD, VAL) 1025 data of the recorded plastic film rolls A1025, X1, X2, X1 + X2, Y1, Y2, Y5, Y6, S obtained by the method described above are prepared as variables. did.

For the prepared 1025 data of X1, X2, X1 + X2, Y1, Y2, Y5, Y6, S, the width direction winding diameter distribution with the attribute class set to "abnormal" is set as the objective variable, and the other "normal" is set. The coefficient β that maximizes the sum of the log-likelihoods obtained by the following equations was obtained from Equations 2 and 3.

Similarly, the widthwise winding diameter distribution (TD, VAL) 1030 data of the recorded plastic film roll B30 and the provisional widthwise winding diameter distribution (TD, VAL) 1000 data of the plastic film roll B is described as a variable. X1, X2, X1 + X2, Y1, Y2, Y5, Y6, and S obtained by the above method were prepared.

For the 1030 data of the prepared X1, X2, X1 + X2, Y1, Y2, Y5, Y6, S, the width direction winding diameter distribution with the attribute class set to "abnormal" is set as the objective variable, and the other "normal" is set. The coefficient β that minimizes the value of log-likelihood × -2 obtained by the following equation was obtained from Equations 2 and 3.

Since the cross-validation method is used as a versatility evaluation, the number of data divisions M is set to 5.
Table 2 shows a part of the results for 1025 plastic film rolls A.

Here, the part where the coefficient column in the coefficient β that minimizes the log-likelihood of each variable in Table 2 is blank is the part that is blank when finding the coefficient that minimizes the value of log-likelihood × -2. It means that the variable of was not used.

As a result of 1025 plastic film rolls A, the correct answer rate, recall rate, and F value were the highest in learning No. 12 with X1 + X2, Y1, and S as variables.
(Classification step)
As a classification step in which the attribute class is “normal”, the “estimated value” was calculated by Equation 3 from _{the coefficient β n obtained in the learning step.}

In the results for 1025 plastic film rolls A, the estimated value of the attribute class that is "normal" is close to the γ distribution of 5.5% and σ0.3% on average, so the probability of being "normal" is 10. It can be calculated that the "estimated value" that is less than% is less than 13%, and if the "estimated value" is 13% or more, it can be predicted that the winding shape will be rejected, and it can be the timing to correct the thickness of the base. We were able to reduce the time required for the evaluation of engineers based on our experience.

In addition, as a result of learning the timing of correcting the thickness of the base with 1025 plastic film rolls and correcting the thickness of the base before the winding is rejected, the winding appearance has been rejected at a rate of 10%. The yield that was out has disappeared.

For 1030 plastic film rolls, the AUC value exceeds 50% when the "estimated value" threshold, which has a low log-likelihood x-2 and the minimum BER value, is taken in 255 combinations of all variables. Moreover, the combination of variables with high recall was obtained.

Table 5 shows a part of the results for 1030 plastic film rolls.

Here, the place where the coefficient column in the coefficient β that minimizes the log-likelihood of each variable in Table 5 is blank is a variable that is blank when finding the coefficient that minimizes the value of log-likelihood × -2. Means that was not used.

As a result of 1030 plastic film rolls B10, learning No. 79 had a low log-likelihood × -2, a high AUC value, and a high reproducibility, so the variables were set to Y2, Y5, and S.

It was possible to set the timing to correct the thickness of the base, and it was possible to reduce the time required for the evaluation of engineers based on many years of experience.
(Reflection step)
Classification Step "estimate" is a plastic film roll winding diameter distribution became "estimate" threshold or (TD, VAL) of _{VAL TD} was subtracted by the average VAL VAL _{TD 'targeting} of winding diameter distribution curve ft (TD the VAL difference in the width direction Dear between) VAL _{TD 'calculated} as -ft (TD), averaged VAL difference range of the width direction ± 30 mm by the adjacent averaging method, VAL _TD' and 'measured plastic film roll The number of sheets was calculated from the winding length of _{the plastic film, and VAL TD} '' was divided by the number of sheets to obtain the deviation target value in the width direction per plastic film. The obtained deviation target value in the width direction was input as a gap control parameter for each corresponding position in the width direction of the base.

With 30 plastic film rolls B and 1000 provisional data, the timing of correcting the thickness of the base was learned, and the thickness of the base was corrected before the winding was rejected. The yield that had been rejected has disappeared.

1 Measured winding diameter distribution in the width direction (width position: TD, winding diameter: VAL)
2 Straight line f (TD) connecting both ends of the winding diameter distribution curve
_{3 Line f-convex} (TD) connecting adjacent convex and convex vertices of the winding diameter distribution curve
_{4 Line f concave} (TD) connecting the adjacent concave and concave bottom points of the winding diameter distribution curve
5 Target winding diameter distribution curve _ft (TD) + k
6 Maximum value of the difference between the straight line f (TD) connecting both ends of the winding diameter distribution curve and the winding diameter distribution curve larger than the straight line connecting both ends of the winding diameter distribution curve X1
7 Maximum value of the difference between the straight line f (TD) connecting both ends of the winding diameter distribution curve and the winding diameter distribution curve smaller than the straight line connecting both ends of the winding diameter distribution curve X2
8 Difference in diameter at both ends of winding diameter distribution curve Y5
9 Difference between maximum diameter and minimum diameter of winding diameter distribution curve Y6
10 Maximum value of the difference from the winding diameter distribution curve from the _{line f convex} (TD) connecting the adjacent convex and convex vertices of the winding diameter distribution curve Y1
11 Maximum value of the difference from the winding diameter distribution curve from the _{line f concave} (TD) connecting the adjacent concave and concave bottom points of the winding diameter distribution curve Y2
12 cores 13 rolls

Claims (6)

This is a plastic film roll width direction winding diameter distribution control method in which a film is wound on a cylindrical core.
Measured width direction winding diameter distribution (width position: TD, winding diameter: VAL), variables calculated from the width direction winding diameter distribution, data of "normal" and "abnormal" attribute classes, and the target width direction a recording step of recording the winding diameter distribution _f t (TD),
Each that optimizes the variable calculated from the recorded width direction winding diameter distribution and the data of the "normal" and "abnormal" attribute classes, and the probability "estimated value" of being classified into each attribute class from the associated attribute class. A learning step to find the coefficient β _{n over a variable,
From the coefficient β n} obtained in the learning step and Equation 3, the probability that the newly measured widthwise winding diameter distribution (TD, VAL) belongs to either the “normal” or “abnormal” attribute class is “estimated value”. And the classification step to calculate and classify
Feedback by the classification step, if it is classified as "abnormal", the width direction winding diameter distribution of the plastic film roll (TD, VAL) and the difference between the winding of the target size distribution curve f _{t (TD)} to control the thickness of the base Reflection steps and
A plastic film roll width direction winding diameter distribution management method.

The learning step
Let X1 be the maximum value of the difference between the straight line f (TD) connecting both ends of the winding diameter distribution curve and the winding diameter distribution curve larger than the straight line connecting both ends of the winding diameter distribution curve.
Let X2 be the maximum value of the difference between the straight line f (TD) connecting both ends of the winding diameter distribution curve and the winding diameter distribution curve smaller than the straight line connecting both ends of the winding diameter distribution curve.
Let Y1 be the maximum value of the difference from the winding diameter distribution curve from the _{line f convex} (TD) connecting the adjacent convex and convex vertices of the winding diameter distribution curve.
Let Y2 be the maximum value of the difference from the winding diameter distribution curve from the _{line f concave} (TD) connecting the adjacent concave and concave bottom points of the winding diameter distribution curve.
Let Y5 be the difference between the diameters at both ends of the winding diameter distribution curve.
Let Y6 be the difference between the maximum diameter and the minimum diameter of the winding diameter distribution curve.
Winding diameter distribution curve (TD, VAL) and the target take-size distribution curve _f t (TD) winding diameter of the target plus a constant value k so that the area difference is the minimum of the distribution curve _f t (TD) + k and winding When the area difference of the diameter distribution curve (TD, VAL) is S,
A variable is at least one selected from the group consisting of X1, X2, Y1, Y2, Y5, Y6 and S.
The plastic film roll width direction winding diameter distribution management method according to claim 1, wherein the attribute class is used as an objective variable. The target in the width direction winding diameter distribution f _{t (TD)} is 1 or more in the width direction winding diameter distribution (TD, VAL) is obtained by averaging from a plastic film roll width direction winding according to claim 1 or 2 Diameter distribution management method. The recording step is a temporary width direction winding diameter distribution (TD, VAL) synthesized from a plurality of VAL fluctuation periods and VAL fluctuation widths calculated from a plurality of widthwise winding diameter distributions (TD, VAL) measured. The plastic film roll widthwise winding diameter distribution management method according to any one of claims 1 to 3, wherein the plastic film roll widthwise winding diameter distribution management method is created and the created temporary widthwise winding diameter distribution (TD, VAL) is recorded. The method for managing the winding diameter distribution in the width direction of a plastic film roll according to any one of claims 1 to 4, wherein the plastic film roll is a resin film roll for a magnetic recording medium. A plastic film roll produced by applying the method for controlling the winding diameter distribution in the width direction according to any one of claims 1 to 5.

Images