Wire diameter prediction method, wire diameter prediction device, die determination method, and die determination system

Abstract

A technique is provided for improving the accuracy of predicting the wire diameter after wire drawing of a metal wire using a die. [Solution] At least the die hole diameter, bearing length and reduction angle are input into a wire diameter prediction model, which is a trained machine learning model that has accumulated a large number of sets of training data each of which pairs at least the die hole diameter, bearing length and reduction angle with the wire diameter after wire drawing, and the wire diameter after wire drawing is output from the wire diameter prediction model.

Description

[Technical field]

[0001] The present invention relates to a technique for predicting the wire diameter after wiredrawing of a metal wire using a die, and a technique for determining a die according to a target wire diameter by utilizing the prediction technique. [Background technology]

[0002] The wire drawing process of metal wire using a die is a type of drawing process in which the wire diameter is reduced by drawing the metal wire through a cylindrical tool called a die, the inner diameter of which narrows from the entrance side to the exit side. The history of drawing is long, but it is only in recent years that numerical analysis has started to be used to predict the shape after drawing.

[0003] Patent Document 1 discloses a micro-diameter measuring device that improves measurement accuracy by determining and storing a correlation equation between the die hole diameter and the diameter of the wire drawn by the die, and by having a calculation unit that predicts one of the hole diameter and wire diameter based on the measured value of the other from the correlation equation. Patent Document 2 discloses a method of predicting the shape after dry drawing from the relationship between the rate of shape change before and after dry drawing and the stress ratio using the finite element method. However, such conventional prediction techniques are not sufficiently accurate in predicting the wire diameter after the metal wire is drawn using a die. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 6-294619 [Patent Document 2] Japanese Patent Application Publication No. 7-144214 Summary of the Invention [Problem to be solved by the invention]

[0005] An object of the present invention is to provide a technique for improving the accuracy of predicting the wire diameter of a metal wire after wire drawing using a die. [Means for solving the problem]

[0006] The inventors focused on the fact that the shape of the die changes due to repeated wire drawing, and conducted a detailed analysis of the shape factors of the die that affect the wire diameter after wire drawing, and found that in addition to the die hole diameter, the bearing length and reduction angle also have a large effect. Therefore, the inventors created a machine learning model using training data that includes at least the die hole diameter, bearing length, and reduction angle, and the wire diameter after wire drawing, and used this as a wire diameter prediction model.

[0007] That is, according to one aspect of the present invention, the following wire diameter prediction method is provided. A wire diameter prediction method for predicting a wire diameter of a metal wire after wire drawing using a die by a wire diameter prediction model, comprising: The wire diameter prediction model is a trained machine learning model in which a large number of sets of teacher data, each set including at least a die hole diameter, a bearing length, and a reduction angle, and a wire diameter after wire drawing, are accumulated; The wire diameter prediction model is created in a narrow diameter range for a die hole diameter or a wire diameter after wire drawing in a size range in which the wire diameter of the metal wire to be processed is small, When the wire diameter of the metal wire to be processed is in a size range larger than the size range, the hole diameter of the die or the wire diameter after wire drawing is made in a diameter range wider than the diameter range, a wire diameter prediction model that outputs a wire diameter after wire drawing, the wire diameter prediction model inputting at least a die hole diameter, a bearing length, and a reduction angle. Effect of the Invention

[0008] According to the present invention, it is possible to accurately predict the wire diameter of a metal wire after it has been drawn using a die. [Brief description of the drawings]

[0009] [Figure 1] FIG. 4 is a schematic cross-sectional view showing the configuration of a die. [Diagram 2]Flow diagram for creating a wire diameter prediction model. [Diagram 3] FIG. 2 is a block diagram conceptually illustrating an example of a wire diameter prediction device. [Figure 4] FIG. 11 is a flow diagram conceptually illustrating an example of a dice determination method. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] FIG. 1 is a schematic cross-sectional view showing the configuration of a die. The metal wire is processed in the drawing direction from the maximum diameter side to the minimum diameter side of the die 1 through a bearing 3 having a hole diameter A, which is the minimum diameter inside the die 1. The inclined portion up to the bearing 3 is called the reduction 2. As described above, the inventors of the present invention analyzed in detail the shape factors of the die that affect the wire diameter after wire drawing, and found that the shape factors of the part with which the metal wire comes into contact during wire drawing, that is, the hole diameter A, the length of the bearing 2 (bearing length B), and the angle formed by the reduction 2 (reduction angle C), have a large effect. Therefore, the inventors of the present invention created a machine learning model using teacher data that includes at least the hole diameter A, the bearing length B, and the reduction angle C of the die, and the wire diameter after wire drawing, and used this as a wire diameter prediction model. In the present invention, the hole diameter of the die refers to the minimum diameter of the hole of the bearing of the die, in other words, the minimum diameter of the hole part of the die described later. Hereinafter, the present invention will be specifically described, including the configuration that is the premise of the embodiment of the present invention.

[0011] <Wire drawing process using a die> In wire drawing, a die is set on a fixed base, and the metal wire is drawn through it to thin it. In this process, the metal wire is subjected to a large tension, so in order to prevent it from breaking, it is necessary to keep the reduction in cross-sectional area per process to a minimum. Therefore, in order to obtain the product size, the wire is repeatedly drawn several to several dozen times, gradually thinning the thick wire. As the metal wire is repeatedly passed through the die, the reduction and bearing parts of the die, i.e., the hole in the die, become deformed due to wear. The dies used in the wire drawing process are generally used dies for metal wire processing, and are preferably made of a high-hardness material such as die steel, cemented carbide, diamond, etc., which reduces the amount of wear in the hole of the die and increases the accuracy of predicting the wire diameter after wire drawing. There is no limit to the type of metal wire to be processed, and it may be light metals such as aluminum, heavy metals such as iron or copper, precious metals such as gold, silver or platinum, or alloys. In the present invention, even in the case of a hard tungsten wire, the wire diameter can be accurately predicted without being limited by the amount of wear in the die hole.

[0012] <Creating a wire diameter prediction model> FIG. 2 is a flow chart for creating a wire diameter prediction model. In the present invention, the wire diameter prediction model is a machine learning model, and as the machine learning model, a known machine learning model such as wire diameter regression, random forest, or neural network is used.

[0013] In creating a wire diameter prediction model, training data is created to be trained by a machine learning model. In the present invention, the training data is a set of at least the die hole diameter, the bearing length, and the reduction angle, and the wire diameter after wire drawing. That is, in the present invention, the training data has at least the die hole diameter, the bearing length, and the reduction angle as explanatory variables, and the wire diameter after wire drawing as a response variable.

[0014] The shape of the die is measured to obtain explanatory variables. A shape measuring instrument is used for the measurement. As long as it can measure at least the hole diameter of the die, the bearing length, and the reduction angle, either a contact type or a non-contact type may be used. Preferably, a three-dimensional measuring instrument or an image measuring instrument is used. As described above, in the present invention, at least the die hole diameter, bearing length, and reduction angle are the explanatory variables, but other explanatory variables such as the roundness of the die hole diameter, maximum bearing length, minimum bearing length, and bearing length difference can also be added. These additional explanatory variables can also be obtained by measuring the shape of the die. Furthermore, manufacturing conditions such as the environmental temperature during wire drawing, the processing speed, and the area reduction rate can also be added to the explanatory variables. When the area reduction rate is used, it is calculated as {(wire diameter before processing) - (wire diameter after processing)} / wire diameter before processing × 100% during wire drawing. On the other hand, the wire diameter after wiredrawing, which is the objective variable, may be a value obtained by actually measuring the wire diameter of the wire after wiredrawing, or if it cannot be measured, it may be calculated by conversion from the weight value after wiredrawing.

[0015] In this way, many sets of training data are created, each set consisting of at least the die hole diameter, bearing length, and reduction angle, and the wire diameter after wire drawing, and this is trained by a machine learning model to create a trained machine learning model, i.e., a wire diameter prediction model. Note that the more sets of training data required for training, the better the prediction accuracy, but if there are at least 10 sets or more, the prediction error can be reduced. Furthermore, if there are 15 sets or more, the prediction error can be kept within 2%.

[0016] In the present invention, the wire diameter prediction model can be created for the entire range regardless of the size of the die hole diameter or the wire diameter after wire drawing. However, as a result of various tests conducted by the present inventors, it was found that the prediction accuracy is further improved by creating the wire diameter prediction model for each range of the die hole diameter or the wire diameter after wire drawing. In particular, for sizes smaller than 0.1 mm, it is better to create a learning model in a narrower size range. This is because the smaller the size, the more minute the size difference affects the predicted value. It is preferable to create a wire diameter prediction model in a range of 10% or less around the die hole diameter or the target wire diameter. For sizes exceeding 0.1 mm, the prediction error is small even if the wire diameter prediction model is created in a wider range. In this case, a highly accurate prediction value can be obtained even if the wire diameter prediction model is created in a range of 30% or more around the die hole diameter or the target wire diameter.

[0017] <Wire diameter prediction device> 3 is a block diagram conceptually showing an example of a wire diameter prediction device. This wire diameter prediction device has a shape measuring device 4 and a management computer 5. The measurement data measured by the shape measuring device 4 is input by an operator to the management computer 5 using an information recording medium such as an updatable flash memory, a built-in hard disk or a hard disk connected via a data communication terminal, or a memory card, and a read / write device for the same. The data may also be set to be stored directly in the management computer. The management computer 5 includes an inventory database into which the identification data and shape data of the die and the measured wire diameter after wire drawing are input, a data calculation unit having a program for processing the shape data and the measured wire diameter into numerical values ​​for teacher data, and a wire diameter prediction model which is a machine learning model. The wire diameter prediction model is a trained machine learning model in which many sets of teacher data, each set consisting of die shape data (at least the die hole diameter, bearing length, and reduction angle) and the wire diameter after wire drawing, are accumulated, and when at least the die hole diameter, bearing length, and reduction angle are input to this wire diameter prediction model, the wire diameter after wire drawing is predicted and output by this wire diameter prediction model. The data added to the inventory database may be processed by the calculation unit and added as training data, and a highly accurate prediction value may be obtained by reconstructing the linear prediction model.

[0018] <How the dice are determined> FIG. 4 is a flow diagram conceptually showing an example of a dice determination method. system The wire diameter prediction device 6 includes a die diameter recording means for outputting the wire diameter after wire drawing from the wire diameter prediction model using the wire diameter prediction device shown in Fig. 3 and recording the output wire diameter as the die diameter, and a die determination means for selecting the die diameter closest to the target wire diameter from the recorded multiple die diameters and determining the die corresponding to the selected die diameter as the die corresponding to the target wire diameter. system When the desired wire diameter (target wire diameter) is input in 6, a die corresponding to the desired wire diameter (target wire diameter) is output. EXAMPLES

[0019] Tables 1 to 3 show the predicted wire diameters predicted by the wire diameter prediction device shown in FIG. 3 and the actual measured values.

[0020] [Table 1]

[0021] [Table 2]

[0022] [Table 3]

[0023] Table 1 shows a wire diameter prediction model created using 18 sets of training data for die hole diameters in the range of 0.04 mm or more and less than 0.05 mm, Table 2 shows a wire diameter prediction model created using 36 sets of training data for die hole diameters in the range of 0.05 mm or more and less than 0.08 mm, and Table 3 shows a wire diameter prediction model created using 17 sets of training data for die hole diameters in the range of 0.1 mm or more and less than 0.2 mm. In all wire diameter prediction models, the absolute error was kept within 2%. [Explanation of symbols]

[0024] 1 Dice 2. Reduction 3. Bearings 4 Shape measuring device 5 Management Computer 6 Dice Decision system A Hole diameter B Bearing length C Reduction angle

Claims

1. A wire diameter prediction method for predicting a wire diameter of a metal wire after wire drawing using a die by a wire diameter prediction model, comprising: The wire diameter prediction model is a trained machine learning model in which a large number of sets of teacher data, each set including at least a die hole diameter, a bearing length, and a reduction angle, and a wire diameter after wire drawing, are accumulated; The wire diameter prediction model is created in a narrow diameter range for a die hole diameter or a wire diameter after wire drawing in a size range in which the wire diameter of the metal wire to be processed is small, In a size range in which the wire diameter of the metal wire to be processed is larger than the size range, the hole diameter of the die or the wire diameter after wire drawing is made to be in a diameter range wider than the diameter range, a wire diameter prediction model that outputs a wire diameter after wire drawing, the wire diameter prediction model inputting at least a die hole diameter, a bearing length, and a reduction angle.

2. A method for determining a die according to a target wire diameter from a plurality of dies, comprising the steps of: outputting wire diameters after wire drawing from the wire diameter prediction model for the plurality of dies using the wire diameter prediction method according to claim 1; The output diameter is recorded as the die diameter. Select a die wire diameter that is closest to the target wire diameter from the recorded plurality of die wire diameters; A die determination method for determining a die corresponding to a selected die wire diameter as a die according to a target wire diameter.

3. A wire diameter prediction device for predicting a wire diameter after wire drawing of a metal wire using a die, comprising: The wire diameter prediction model is a trained machine learning model in which a large number of sets of training data, each set including at least the hole diameter of the die, the bearing length, and the reduction angle, and the wire diameter after wire drawing, are accumulated; The wire diameter prediction model is created in a narrow diameter range for a die hole diameter or a wire diameter after wire drawing in a size range in which the wire diameter of the metal wire to be processed is small, A wire diameter prediction device, wherein when the wire diameter of the metal wire to be processed is in a size range larger than the size range, the range of the die hole diameter or the wire diameter after wire drawing is created in a diameter range wider than the diameter range.

4. A die selection system for selecting a die according to a desired wire diameter from a plurality of dies, The wire diameter prediction device according to claim 3 ; a die wire diameter recording means for recording a wire diameter after wire drawing from the wire diameter prediction model using the wire diameter prediction device as a die wire diameter; Select a die wire diameter that is closest to the target wire diameter from the recorded plurality of die wire diameters; A die determination system comprising a die determination means for determining a die corresponding to a selected die wire diameter as a die according to a target wire diameter.

Images