Characteristic estimation system, quality determination system, characteristic estimation method, and quality determination method

Abstract

To estimate characteristics of a long intermediate product manufactured from a casting.SOLUTION: This characteristic estimation system for estimating characteristics of a long material manufactured from a casting machine-learns a relationship among appearance inspection data indicating the appearance inspection result of the long material, first step data regarding a step of manufacturing the long material, and characteristic inspection data indicating the characteristic inspection result of the long material to generate a regression model indicating a correlation among the appearance inspection data, the first step data, and the characteristic inspection data (ST104), and estimates characteristics of the long material on the basis of the appearance inspection data, the first step data, and the regression model of the long material to be estimated (ST105).SELECTED DRAWING: Figure 19

Description

[Technical Field]

[0001] The present invention relates to a characteristic estimation system and a characteristic estimation method for estimating the characteristics of a long intermediate product manufactured from a casting material, and a quality determination system and a quality determination method for determining the quality of the characteristics of a long product manufactured from the intermediate product. [Background technology]

[0002] International Publication No. 2020-090848 (Patent Document 1) describes a technology for deriving an optimal solution for design conditions that satisfy desired material properties. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2020-090848 Summary of the Invention [Problem to be solved by the invention]

[0004] The properties of long intermediate products manufactured from cast materials and long end products manufactured using the intermediate products can be evaluated, for example, by cutting out the end of the intermediate product or the end product and performing a twist test on the end. However, because the twist test can only evaluate the properties of the end, the evaluation does not guarantee the quality of the intermediate product or the end product over the entire length. [Means for solving the problem]

[0005] In one embodiment, a property estimation system for estimating the properties of long material manufactured from cast material includes a regression model creation unit that machine-learns the relationship between appearance inspection data indicating the inspection results of the appearance of the long material, first process data related to the process of manufacturing the long material, and property inspection data indicating the inspection results of the properties of the long material, and creates a regression model that represents the correlation between the appearance inspection data, the first process data, and the property inspection data, and a property estimation unit that estimates the properties of the long material to be estimated based on the appearance inspection data, first process data, and the regression model of the long material to be estimated.

[0006] In addition, in one embodiment, a quality determination system for determining the quality of the characteristics of a linear material manufactured using a long material produced from a cast material includes: a regression model creation unit that performs machine learning on the relationship between long material data related to the long material, second process data related to the process of manufacturing the linear material, and characteristic investigation data indicating the investigation results of the characteristics of the linear material, and creates a regression model that represents the correlation between the long material data, the second process data, and the characteristic investigation data; a characteristic estimation unit that estimates the characteristics of the linear material to be estimated based on the second process data of the linear material to be estimated and the regression model; and a quality determination unit that determines whether the estimated value estimated by the characteristic estimation unit satisfies a threshold that defines the quality of the linear material to be estimated.

[0007] In addition, in one embodiment, a characteristic estimation method for estimating the characteristics of a long material manufactured from cast material includes acquiring appearance inspection data indicating the inspection results of the appearance of the long material, acquiring first process data related to the process of manufacturing the long material, acquiring characteristic inspection data indicating the inspection results of the characteristics of the long material, machine learning the relationship between the appearance inspection data, the first process data, and the characteristic inspection data, creating a regression model representing the correlation between the appearance inspection data, the first process data, and the characteristic inspection data, and estimating the characteristics of the long material to be estimated based on the appearance inspection data, first process data, and the regression model of the long material to be estimated.

[0008] In addition, in one embodiment, a quality determination method for determining the quality of the characteristics of a linear material manufactured using a long material produced from a cast material includes acquiring long material data related to the long material, acquiring second process data related to the process of manufacturing the linear material, acquiring characteristic investigation data indicating the results of an investigation into the characteristics of the linear material, performing machine learning on the relationship between the long material data, the second process data, and the characteristic investigation data, creating a regression model representing the correlation between the long material data, the second process data, and the characteristic investigation data, and determining whether an estimated value of the characteristics of the linear material to be estimated based on the second process data of the linear material to be estimated and the regression model satisfies a threshold value that defines the quality of the linear material to be estimated. [Effects of the Invention]

[0009] According to one embodiment, the property estimation system can estimate the properties of a long material manufactured from a cast material. Also, according to one embodiment, the quality determination system can determine the quality of the properties of a linear material manufactured from a long material. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram schematically illustrating an example of the overall configuration of a manufacturing system. [Figure 2] FIG. 2 illustrates an example of a hardware configuration of a computer. [Figure 3] FIG. 2 is a diagram illustrating an example of data included in a file group. [Figure 4] FIG. 2 is a diagram illustrating an example of a functional block. [Figure 5] FIG. 2 is a diagram illustrating an example of a functional block. [Figure 6] FIG. 1 is a schematic side view of a casting system. [Figure 7] 1 is a schematic side view of a wire drawing manufacturing system. FIG. [Figure 8] FIG. 10 is a diagram illustrating an example of management data. [Figure 9] FIG. 10 is a diagram illustrating an example of appearance inspection data. [Figure 10]FIG. 10 is a diagram showing an example of first process data. [Figure 11] FIG. 10 is a diagram illustrating an example of sampling inspection data. [Figure 12] FIG. 10 is a diagram showing an example of test inspection data. [Figure 13] FIG. 2 is a diagram schematically illustrating an example of a crack surface. [Figure 14] FIG. 10 is a diagram showing an example of second process data. [Figure 15] FIG. 10 is a diagram showing an example of second process data. [Figure 16] FIG. 10 is a diagram showing an example of second process data. [Figure 17] FIG. 10 is a diagram illustrating an example of defect investigation data. [Figure 18] FIG. 10 is a diagram illustrating an example of organizational survey data. [Figure 19] 10 is a flowchart illustrating an example of a characteristic estimation process. [Figure 20] 10 is a flowchart illustrating an example of a quality determination process. DETAILED DESCRIPTION OF THE INVENTION

[0011] In all the drawings for explaining the embodiments, the same components are generally designated by the same reference numerals, and repeated explanations thereof will be omitted. In addition, hatching may be used even in plan views to make the drawings easier to understand.

[0012] In the following, we will mainly explain an example in which the characteristic estimation system and the pass / fail determination system are configured from a single computer, but the characteristic estimation system and the pass / fail determination system in this embodiment can also be realized as a distributed system consisting of multiple computers.

[0013] <Overall configuration of the manufacturing system> FIG. 1 is a diagram schematically illustrating an example of the overall configuration of a manufacturing system 1. The manufacturing system 1 is a system that produces long materials from cast materials and uses the produced long materials to produce linear materials. The long materials are long intermediate products, and the linear materials are long end products. For example, the cast material is copper, the long material is coil P10, and the linear material is drawn wire material P50 produced using the coil P10. In addition to producing drawn wire material P50 from coil P10, the end product may also be produced by further processing the drawn wire material P50. In this embodiment, a case will be described in which the cast material is copper, the long material is coil P10, and the linear material is drawn wire material P50, but the cast material, long material, and linear material are not limited to these.

[0014] 1, the manufacturing system 1 includes a casting system 10, a wire drawing manufacturing system 70, and a computer 100. In the manufacturing system 1, a coil P10 is manufactured in a casting process S10, and a drawn wire P50 is manufactured using the manufactured coil P10 in a wire drawing manufacturing process S50. The computer 100 functions as a characteristic estimation system 100A and a quality determination system 100B, the details of which will be described later.

[0015] The casting system 10 produces the coil P10 by performing a casting process in which molten metal is poured into a mold of a desired shape and cooled and solidified to form the coil P10.

[0016] The wire drawing material production system 70 produces a drawn wire material P50 from the coil P10. The drawn wire material P50 is used, for example, as a core wire of an electric wire. The wire drawing material production system 70 may further process the drawn wire material P50.

[0017] The computer 100 has the function of acquiring and managing data related to the coil P10 and the drawn wire material P50 manufactured by the manufacturing system 1, estimating the characteristics of the coil P10, and determining whether the characteristics of the drawn wire material P50 are good or bad. The computer 100 acquires, for example, appearance inspection data D11 and first process data D12 from the casting system 10, and acquires characteristic inspection data (sampling inspection data D13 and test inspection data D14) from a predetermined information processing device. The computer 100 also acquires, for example, second process data D15 from the drawn wire material manufacturing system 70, and acquires characteristic investigation data (defect investigation data D16 and structure investigation data D17) from a predetermined information processing device. Details of these data will be described later with reference to FIG. 3 etc.

[0018] <Computer hardware configuration> Next, the hardware configuration of the computer 100 in this embodiment will be described. Fig. 2 is a diagram showing an example of the hardware configuration of the computer 100 in this embodiment. Note that the configuration shown in Fig. 2 is merely an example of the hardware configuration of the computer 100, and the hardware configuration of the computer 100 is not limited to the configuration shown in Fig. 2 and may be other configurations.

[0019] 2, a computer 100 includes a CPU (Central Processing Unit) 101 that executes a program. This CPU 101 is electrically connected to, for example, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, and a hard disk drive 112 via a bus 113, and is configured to control these hardware devices.

[0020] The CPU 101 is also connected to input devices and output devices via a bus 113. Examples of input devices include a keyboard 105, a mouse 106, a communication port 107, and a scanner 111. Examples of output devices include a display 104, a communication port 107, and a printer 110. The CPU 101 may also be connected to, for example, a removable disk device 108 and a CD / DVD-ROM device 109. For example, sensors (described later) provided in the casting system 10 and the wire drawing manufacturing system 70 are connected to the communication port 107.

[0021] The computer 100 may be connected to, for example, a network. For example, when the computer 100 is connected to other external devices via a network, a communication port 107 constituting a part of the computer 100 is connected to a LAN (Local Area Network), a WAN (Wide Area Network), or the Internet.

[0022] The RAM 103 is an example of volatile memory, and the storage media of the ROM 102, removable disk device 108, CD / DVD-ROM device 109, and hard disk device 112 are examples of non-volatile memory. These volatile and non-volatile memories constitute the storage device of the computer 100.

[0023] The hard disk drive 112 stores, for example, an operating system (OS) 201, a program group 202, and a file group 203. The programs included in the program group 202 are executed by the CPU 101 using the operating system 201. The RAM 103 also temporarily stores at least some of the programs of the operating system 201 and application programs that the CPU 101 executes, as well as various data required for processing by the CPU 101.

[0024] A BIOS (Basic Input Output System) program is stored in the ROM 102, and a boot program is stored in the hard disk drive 112. When the computer 100 is started up, the BIOS program stored in the ROM 102 and the boot program stored in the hard disk drive 112 are executed, and the operating system 201 is started up by the BIOS program and the boot program.

[0025] The program group 202 stores programs that realize the functions of the computer 100, and these programs are read and executed by the CPU 101. The file group 203 stores information, data, signal values, variable values, and parameters indicating the results of processing by the CPU 101 as file items. The program group 202 includes, for example, a characteristic estimation program 202A and a pass / fail determination program 202B, which will be described later. The file group 203 includes files such as data managed by the manufacturing system 1.

[0026] FIG. 3 is a diagram showing an example of data included in the file group 203. As shown in FIG. 3, the file group 203 includes, in addition to the management data D10, the above-mentioned appearance inspection data D11, first process data D12, sampling inspection data D13, test inspection data D14, second process data D15, defect investigation data D16, and organization investigation data D17. The file group 203 also includes a first regression model D18 and a second regression model D19.

[0027] The management data D10 is data used in the manufacturing system 1 to manage the coil P10 and the drawn wire material P50 to be manufactured. The appearance inspection data D11 is data indicating the results of an inspection of the appearance of the coil P10 during manufacturing. The first process data D12 is data related to the manufacturing process of the coil P10. The sampling inspection data D13 and the test inspection data D14 are data indicating the results of inspection of the characteristics of the coil P10. The second process data D15 is data related to the manufacturing process of the drawn wire material P50. The defect investigation data D16 and the structure investigation data D17 are data indicating the results of investigations into the characteristics of the drawn wire material P50. Details of these data will be explained with reference to Figures 8 to 18.

[0028] The first regression model D18 is regression model data used to estimate the characteristics of the coil P10. The second regression model D19 is regression model data used to determine the quality of the characteristics of the drawn wire material P50. Details of the first regression model D18 will be described later with reference to FIG. 19, and details of the second regression model D19 will be described later with reference to FIG. 20.

[0029] The files are stored in a storage medium such as the hard disk drive 112 or memory. The information, data, signal values, variable values, and parameters stored in the storage medium such as the hard disk drive 112 or memory are read into the main memory or cache memory by the CPU 101 and used for the operations of the CPU 101, such as extraction, search, reference, comparison, calculation, processing, editing, output, printing, and display. For example, during the operations of the CPU 101 described above, the information, data, signal values, variable values, and parameters are temporarily stored in the main memory, registers, cache memory, buffer memory, etc.

[0030] The functions of computer 100 may be realized by firmware stored in ROM 102, or may be realized by software alone, hardware alone such as elements, devices, boards, and wiring, a combination of software and hardware, or even a combination of firmware and firmware. Firmware and software are stored as programs in a storage medium such as hard disk drive 112, removable disk drive 108, or CD / DVD-ROM drive 109. The programs are read and executed by CPU 101. For example, the programs cause computer 100 to function as characteristic estimation system 100A or pass / fail determination system 100B.

[0031] Thus, computer 100 is a computer that includes CPU 101, which is a processing device, hard disk drive 112 and memory, which are storage devices, keyboard 105, mouse 106, and communication port 107, which are input devices, and display 104, printer 110, and communication port 107, which are output devices. The functions of computer 100 are realized using the processing device, storage device, input device, and output device.

[0032] <Function block> FIG. 4 is a diagram showing an example of functional blocks when the computer 100 functions as the characteristic estimation system 100A. As shown in FIG. 4, the characteristic estimation system 100A includes a regression model creation unit 310 and a characteristic estimation unit 320.

[0033] The regression model creation unit 310 performs machine learning to determine the relationship between appearance inspection data D11 indicating the inspection results of the appearance of the coil P10, first process data D12 related to the process of manufacturing the coil P10, and characteristic inspection data (sampling inspection data D13, test inspection data D14) indicating the inspection results of the characteristics of the coil P10, and creates a first regression model D18 that represents the correlation between the appearance inspection data D11, the first process data D12, and the characteristic inspection data.

[0034] In addition, the regression model creation unit 310 identifies the appearance inspection data D11, the first process data D12, and the characteristic inspection data (sampling inspection data D13, test inspection data D14) from a plurality of appearance inspection data D11, first process data D12, and characteristic inspection data (sampling inspection data D13, test inspection data D14) based on the management data D10 that manages the coil P10.

[0035] The first process data D12 is received at predetermined intervals from a casting system 10 that manufactures a coil P10 from a casting material, and the regression model creation unit 310 creates a first regression model D18 using the average value of the multiple first process data D12 received at predetermined intervals.

[0036] The sampling inspection data D13 includes, for example, data indicating a determination result as to whether the characteristics of the coil P10 satisfy a predetermined quality. The test inspection data D14 includes, for example, data indicating the quality state of the coil P10.

[0037] The characteristic estimation unit 320 estimates the characteristics of the coil P10 based on the appearance inspection data of the coil P10 to be estimated, the first process data, and the first regression model D18. Details of the functions realized by the regression model creation unit 310 and the characteristic estimation unit 320 will be described later with reference to FIG.

[0038] FIG. 5 is a diagram showing an example of functional blocks when the computer 100 functions as a quality determination system 100B. As shown in FIG. 5, the quality determination system 100B includes a regression model creation section 330, a characteristic estimation section 340, and a quality determination section 350.

[0039] The regression model creation unit 330 performs machine learning on the relationship between the long material data regarding the coil P10 (any data from among the management data D10, visual inspection data D11, first process data D12, sampling inspection data D13, and test inspection data D14), the second process data D15 regarding the process of manufacturing the drawn wire material P50, and the investigation results data (defect investigation data D16, microstructural investigation data D17) showing the investigation results of the characteristics of the drawn wire material P50, and creates a second regression model D19 that represents the correlation between the long material data, the second process data D15, and the microstructural investigation data D17.

[0040] The characteristic estimation unit 340 estimates the characteristic of the coil P10 to be estimated based on the second process data of the drawn wire material P50 to be estimated and the second regression model D19.

[0041] The quality determining unit 350 determines whether or not the estimated value estimated by the characteristic estimating unit 340 satisfies a threshold that defines the quality of the drawn wire material P50 that is the estimation target. Details of the functions realized by the regression model creating unit 330, the characteristic estimating unit 340, and the quality determining unit 350 will be described later with reference to FIG.

[0042] <Casting system configuration> Next, a description will be given of an example of the configuration of the casting system 10. Fig. 6 is a diagram showing the casting system 10 from a side view.

[0043] As shown in Figure 6, the casting system 10 includes a melting furnace 15, a transfer trough 16, a holding furnace 20, a transfer trough 21, a tundish 30, a nozzle 31, a caster 40, a casting wheel 41, a casting belt 42, a cooling water system 43, a mark detector 44, a rolling mill 50, a flaw detector 51, and a coiling machine 60. The tundish 30 is equipped with a temperature sensor 30a. The casting wheel 41 is equipped with a speed sensor 41a. The cooling water system 43 is equipped with a water volume measurement sensor 43a. The rolling mill 50 is equipped with a rolling output sensor 50a and a pressure sensor 50b.

[0044] Copper is introduced into the melting furnace 15 and melted therein. The molten copper flows through the transfer trough 16 into the holding furnace 20. The molten copper held in the holding furnace 20 flows through the transfer trough 21 into the tundish 30. The tundish 30 receives the molten copper and supplies it to the space between the casting wheel 41 and the casting belt 42 of the casting system 10 at a controlled speed via a nozzle 31. In the casting system 10, the rotation of the casting wheel 41 and the casting belt 42 forms copper into a bar shape of a predetermined thickness. The formed copper bar is cooled by pouring cold water onto it using a cooling water device 43. The copper bar thus formed is then rolled into a desired thickness by a rolling mill 50 and wound by a winding device 60 to produce a coil P10.

[0045] The temperature sensor 30a detects the temperature of the molten copper in the tundish 30. The speed sensor 41a detects the rotational speed of the casting wheel 41. The water volume measurement sensor 43a detects the volume of cooling water discharged from the cooling water device 43. The rolling output sensor 50a and the pressure sensor 50b detect the rolling output and pressure of the rolling mill 50. The values ​​detected by the temperature sensor 30a, speed sensor 41a, water volume measurement sensor 43a, rolling output sensor 50a, and pressure sensor 50b are sent to the computer 100 at predetermined intervals, for example, every one or two seconds.

[0046] The flaw detector 44 detects whether or not there are flaws, such as blowholes, on the bar-shaped copper before it is rolled by the rolling mill 50. The flaw detector 44, for example, chamfers the bar-shaped copper and checks for the presence of blowholes on the new surface exposed by the chamfering using an image detection sensor. The flaw detector 51 detects whether or not there are flaws on the bar-shaped copper that has passed through the rolling mill 50 and been formed to the desired thickness. The flaw detector 51 is, for example, an eddy current flaw detector. The flaw detector 51 passes the bar-shaped copper through a coil installed inside the flaw detector 51, counts noise generated when there are flaws on the surface as a detection signal, and determines whether or not there are flaws based on the count. The detection results of the flaw detector 44 and the flaw detector 51 are sent to the computer 100 at predetermined intervals, for example, every one or two seconds.

[0047] <Wire drawing manufacturing system configuration> Next, a description will be given of an example of the configuration of the wire drawn material production system 70. Fig. 7 is a diagram showing the wire drawn material production system 70 from the side.

[0048] 7, the wire drawing production system 70 includes a feed device 71, a coiler 72, a size detection device 73, and a wire drawing device 80. The wire drawing device 80 includes a wire drawing die 81 and a take-up device 82. The take-up device 82 is provided with a speed sensor 82a.

[0049] A coil P10 is placed in the let-off device 71. The coil P10 placed in the let-off device 71 passes through a wire drawing device 80 and is wound around a coiler 72. In the wire drawing device 80, the coil P10 passes through a wire drawing die 81 and is taken up by a take-up device 82. At this time, the diameter of the coil P10 is reduced according to the size of the hole of the wire drawing die 81, and a linear coil having a diameter required for the drawn wire material is formed. The coil thus formed is wound around the coiler 72 to produce the drawn wire material P50.

[0050] The size detection device 73 detects the size (diameter) of the copper rod that has passed through the wire drawing device 80. The speed sensor 82a detects the take-up speed of the coil P10 by the take-up device 82. The detection results of the size detection device 73 and the speed sensor 82a are sent to the computer 100 at predetermined intervals, for example, every one or two seconds.

[0051] Here, the manufacturing process of the wire drawn material P50 shown in FIG. 7 is referred to as process α. The process for manufacturing the wire drawn material P50 is not limited to process α. In this embodiment, in addition to process α, there are processes β and γ. Processes β and γ are similar to process α. For example, in this embodiment, processes β and γ are processes in which the diameter size of the hole of the wire drawing die is different from the size of the wire drawing die 81 in process α, and a wire drawn material P50 with a different diameter size is manufactured. The similar processes are not limited to these.

[0052] <Details of data acquired by the computer> Next, details of data acquired by the computer 100 in the manufacturing system 1 will be described with reference to FIGS.

[0053] FIG. 8 is a diagram showing an example of the management data D10. 8, the management data D10 includes manufacturing time data D10A indicating the manufacturing time, coil number data D10B indicating the coil number, size data D10C, product type data D10D, and weight data D10E. However, the data managed as the management data D10 is not limited to these.

[0054] The manufacturing time data D10A is data indicating the time when manufacturing of the coil P10 started. The coil number data D10B is data specifying the coil P10 to be manufactured. In this embodiment, the size data D10C is data indicating, for example, the size of the coil P10. The product type data D10D is data indicating the type of the coil P10. The weight data D10E is data indicating the weight of the coil P10.

[0055] For example, it is managed that a coil P10 with coil number "A001", size "8", product type "X", and weight "1,642" was manufactured at manufacturing time "XX / YY 09:55".

[0056] FIG. 9 is a diagram showing an example of the visual inspection data D11. 9, the visual inspection data D11 is managed so as to correspond to the coil number data D10B. Similarly, other data D12 to D17 described below in FIGS. 10 to 18 are also managed so as to correspond to the coil number data D10B. This makes it possible to identify which coil P10 the data D11 to D17 relate to.

[0057] The appearance inspection data D11 includes data on the appearance inspection in the manufacturing process of the coil P10, such as scratch data D11A indicating scratches, ..., and mark data D11B indicating marks. In Fig. 9, the scratch data D11A and mark data D11B are shown as examples of data included in the appearance inspection data D11, but the data is not limited to these and may also include data related to other inspections of the appearance of the coil P10.

[0058] The flaw data D11A is data transmitted from the flaw detector 51 to the computer 100 at predetermined intervals. The flaw data D11B is data transmitted from the flaw detector 44 to the computer 100 at predetermined intervals. If all the flaw data acquired for one coil P10 indicates that there is no flaw, the computer 100 sets the flaw data D11A to "0." Similarly, if all the flaw data acquired for one coil P10 indicates that there is no flaw, the computer 100 sets the flaw data D11B to "0." Furthermore, the computer 100 associates the time-series flaw data D11A and the flaw data D11B with the coil number data D10B of the coil P10 and stores them in the file group 203.

[0059] For example, coil P10 with coil number "A001" is associated with data of scratch "0", ..., mark "0". In other words, coil P10 with "A001" has no scratches and no marks. On the other hand, coil P10 with coil number "A002" is associated with data of scratch "0", ..., mark "1". In other words, coil P10 with "A001" has no scratches but has marks.

[0060] 10 is a diagram showing an example of the first process data D12. The first process data D12 is data obtained from the casting system 10 in the process in which the casting system 10 manufactures the coil P10.

[0061] 10, the first process step data D12 includes temperature data D12A indicating the temperature, speed data D12B indicating the speed, ..., rolling data D12C indicating the rolling, pressure data D12D indicating the pressure, water volume data D12E indicating the water volume, and determination result data D12F indicating the determination result. In Fig. 10, the temperature data D12A, speed data D12B, rolling data D12C, pressure data D12D, water volume data D12E, and determination result data D12F are shown as examples of data included in the first process step data D12, but are not limited to these, and the first process step data D12 may also include other data in the manufacturing process of the coil P10.

[0062] The temperature data D12A is data transmitted to the computer 100 from the temperature sensor 30a described above at predetermined intervals. The speed data D12B is data transmitted to the computer 100 from the speed sensor 41a described above at predetermined intervals. The rolling data D12C is data transmitted to the computer 100 from the rolling output sensor 50a described above at predetermined intervals. The pressure data D12D is data transmitted to the computer 100 from the pressure sensor 50b described above at predetermined intervals. The water volume data D12E is data transmitted to the computer 100 from the water volume measurement sensor 43a described above at predetermined intervals. In this embodiment, data is acquired at predetermined intervals, for example, at intervals of one second or two seconds, so the computer 100 acquires an average value from each of the multiple pieces of data acquired from one coil P10 and stores the acquired average values ​​as the temperature data D12A, the speed data D12B, the rolling data D12C, the pressure data D12D, and the water volume data D12E. In the manufacturing process of the coil P10, the computer 100 stores each piece of time-series data acquired from the casting system 10 in association with the coil number data D10B of the coil P10 in the file group 203.

[0063] For example, thresholds for temperature, speed, rolling, pressure, and water volume are set in advance for the judgment result data D12F, and if the values ​​indicated by the temperature data D12A, speed data D12B, rolling data D12C, pressure data D12D, and water volume data D12E satisfy the respective thresholds, the computer 100 sets the judgment result data D12F as "OK." If any threshold is not satisfied, the computer 100 sets the judgment result data D12F as "NO."

[0064] For example, it shows that coil P10 with intermediate serial number "A001" was produced at a tundish 30 temperature of "1120", a casting wheel 41 rotation speed of "34.90", rolling pressure of "80", pressure of "5.00", and water volume of "25.00", and the result of the evaluation is "OK". This shows that coil P10 was produced within the range of the preset threshold values.

[0065] FIG. 11 is a diagram showing an example of sampling inspection data D13. After manufacturing, the coil P10 is inspected, for example, by a twisting test. The twisting test involves cutting off an end of the coil P10, twisting the end, and, if any cracks are found, investigating the cracked portion. The sampling inspection data D13 is data showing the results of the twisting test. The sampling inspection data D13 is input, for example, into an information processing device by the operator who performed the twisting test, and the input information from the information processing device is sent to the computer 100.

[0066] 11, the sampling inspection data D13 includes inspection date data D13A, oxygen content data D13B, dimension data D13C, ..., twist data D13D, and judgment result data D13E. In Fig. 11, the sampling inspection data D13 shows inspection date data D13A, oxygen content data D13B, dimension data D13C, twist data D13D, and judgment result data D13E as examples of data included in the sampling inspection data D13, but is not limited to these and may include other data from the twist test.

[0067] The inspection date data D13A is data indicating the inspection date of the twist test. The oxygen content data D13B is data indicating the amount of oxygen contained in the end cut from the coil P10. The dimension data D13C is data indicating the diameter of the end cut from the coil P10. The twist data D13D is data indicating the degree of twist. For example, if the degree of twist at the time of cracking is large, it is indicated as "large," if it is medium, it is indicated as "medium," and if it is small, it is indicated as "small." The thresholds for "large," "medium," and "small" are set appropriately depending on the material of the coil P10, etc. If the coil P10 can withstand the twist test, it is indicated as "OK."

[0068] The judgment result data D13E has preset thresholds for oxygen, dimension, and twist, and if the values ​​indicated by the oxygen amount data D13B, dimension data D13C, and twist data D13D satisfy the respective thresholds, the computer 100 determines the judgment result as OK. If even one piece of data does not satisfy the threshold, the computer 100 determines the judgment result as NO. In other words, the judgment result data D13E is data indicating the judgment result as to whether or not the characteristics of the coil P10 satisfy the preset quality.

[0069] For example, coil P10 with intermediate serial number "A001" had a twist test inspection date of "XX / YY," and the twist test results were oxygen "0.0250," dimension "8.35," twist "small," and the judgment result "OK." This shows that the end of coil P10 meets the required characteristics.

[0070] FIG. 12 is a diagram showing an example of test inspection data D14. The test inspection is an inspection performed on an end portion that is not "OK" in a twist test, for example. The inspection examines the location of the end portion where the crack occurred, the state of the internal structure of the end portion, and where in the casting process the cause of the crack occurred. In this embodiment, a case will be described in which the number of marks on the surface of the cracked end portion and the cause are stored. The test inspection data D14 is input, for example, into an information processing device by the operator who performed the test, and the information input from the information processing device is sent to computer 100.

[0071] As shown in Fig. 12, the test inspection data D14 includes inspection date data D14A, number of marks data D14B, ..., and factor data D14C. In Fig. 12, the test inspection data D14 illustrates the inspection date data D14A, number of marks data D14B, and factor data D14C as examples of data included in the test inspection data D14, but is not limited to these and may include other data related to the investigation. In other words, the test inspection data D14 may be data that indicates the quality state of the coil P10.

[0072] The inspection date data D14A is data indicating the date when the cracked end was inspected. The mark number data D14B is data indicating the number of marks on the cracked surface. The cause data D14C is data indicating the cause of the crack identified by the worker.

[0073] FIG. 13 is a diagram schematically showing an example of a crack surface P10F at an end portion cracked by a twisting test. As shown in Fig. 13, a total of eight marks are shown on the cracked surface P10F at the end of the coil P10, including four marks P1 and one each of marks P2 to P5. For example, mark P1 is a scratch, mark P2 is a foreign matter deposit, mark P3 is a mark containing gas or the like, mark P4 is a blowhole mark, and mark P5 is a coagulation cavity. In this case, for example, the worker inputs "8" as the number of marks into the information processing device. The type of each mark may also be stored.

[0074] As shown in FIG. 12, for example, coil P10 with coil number "A001" has an inspection date of "XX / ZZ", the inspection results show that the number of marks is "8", and the cause is "U".

[0075] 14 to 16 are diagrams showing examples of the second process data D15. The second process data D15 is data obtained according to the manufacturing process of the wire drawn material manufacturing system 70. Therefore, when the manufacturing process differs, the data obtained as the second process data D15 also differs. In this embodiment, the manufacturing process of the wire drawn material manufacturing system 70 includes three processes: process α, process β, and process γ. Therefore, the second process data D15 will be explained by dividing it into three pieces of second process data D151 to D153.

[0076] Fig. 14 is a diagram showing an example of second process data D151 obtained when the manufacturing process is process α. The second process data D151 includes work date data D151A, product lot data D151B, size data D151C, speed data D151D, ..., judgment data D151E. In Fig. 14, the data included in the second process data D151 are exemplified as work date data D151A, product lot data D151B, size data D151C, speed data D151D, and judgment data D151E, but are not limited to these, and other data in the wire drawing manufacturing process S50 may also be included.

[0077] The work date data D151A is data indicating the date on which the drawn wire material P50 was manufactured. The product lot data D151B is data indicating the lot number of the drawn wire material P50. The size data D151C is data indicating the diameter of the drawn wire material P50. The speed data D151D is data indicating the take-up speed of the take-up device 82. The judgment data D151E is data indicating the judgment result as to whether or not the manufacturing was performed as specified. If the size data D151C, the speed data D151D, etc. are within the range of predefined thresholds, the computer 100 judges the result as OK. If even one piece of data does not satisfy the threshold, the computer 100 judges the result as NG. There are multiple NG types, and the NG corresponding to the type becomes the judgment data D151E. For example, if one threshold is not satisfied, it becomes NG1, and if two thresholds are not satisfied, it becomes NG2. NG1 and NG2 may be distinguished depending on the NG type.

[0078] For example, it is shown that coil P10 with intermediate serial number "A002" is wire drawn material P50 manufactured on the production date "XX / YY" as product lot "α001" with a size of "2.6" and a take-up speed of "500", and that the judgment result is "OK." Furthermore, it is shown that two wire drawn materials P50 with coil numbers "A002" and "B021" are wire drawn materials P50 manufactured by process α.

[0079] FIG. 15 is a diagram showing an example of second process data D152 for process β. As shown in FIG. 15, the second process data D152 includes work date data D152A, product lot data D152B, size data D152C, speed data D152D, ..., and judgment data D152E. FIG. 16 is a diagram showing an example of second process data D153 for process γ. As shown in FIG. 16, the second process data D153 includes work date data D153A, product lot data D153B, size data D153C, speed data D153D, ..., and judgment data D153E. Details of these data are the same as those described for process α in FIG. 14, and therefore will not be described here.

[0080] As shown in Fig. 15, coil numbers "A030", "A031", and "B020" indicate that the wire drawn material P50 was produced in process β. As shown in Fig. 16, coil numbers "A030", "A031", and "B021" indicate that the wire drawn material P50 was produced in process γ. A plurality of the same or different types of wire drawn material P50 are produced from the coil P10, which is an intermediate product. That is, the wire drawn material P50 produced from the coil P10 having the coil number "A001" is only the wire drawn material P50 produced by the process α, the wire drawn material P50 produced from the coils P10 having the coil numbers "A030" and "A031" is the wire drawn material P50 produced by the process β and the process γ, the wire drawn material P50 produced from the coil P10 having the coil number "B020" is the wire drawn material P50 produced by the process β only, and the wire drawn material P50 produced from the coil P10 having the coil number "B021" is the wire drawn material P50 produced by the process α and the process γ. Note that, for example, for the coil number "A001," data for any of the processes α to γ ​​is not stored. In this case, it can be seen that the wire drawn material P50 has not yet been produced from the coil number "A001."

[0081] 17 is a diagram showing an example of the defect investigation data D16. The defect investigation data D16 is data that summarizes the judgment result data in the processes in the wire drawing material production system 70. The judgment result indicates whether or not a defect has been detected in the produced wire drawing material P50. When multiple drawn wire materials P50 are produced from one coil P10, the defect investigation data D16 for the multiple drawn wire materials P50 are associated with the coil number D10B.

[0082] As shown in FIG. 17, the defect investigation data D16 is configured by associating coil number data D10B with determination result data D16A, D16B, and D16C for process α, process β, and process γ.

[0083] For example, the judgment result data for process α, "OK," is stored in association with coil number "A002." That is, it indicates that the wire drawn material P50 with coil number "A002" was manufactured in process α, and the judgment result for whether or not it was defective was "OK." Furthermore, it indicates that the wire drawn material P50 with coil number "A030" was manufactured in process β, and includes one with the judgment result "NG1," and that the wire drawn material P50 with coil number "B021" was manufactured in process γ, and includes one with the judgment result "NG2." The defect investigation data D16 is input into an information processing device or the like by the worker who conducted the defect investigation or by a person who received information from the worker, and the input information from the information processing device is transmitted to the computer 100.

[0084] 18 is a diagram showing an example of the structure investigation data D17. The structure investigation data D17 is data showing the results of investigating the structure of the drawn wire material P50 that was determined to be defective in the steps α to γ.

[0085] 18, the structure investigation data D17 is configured by associating investigation results D17A with coil number data D10B. In the present embodiment, the judgment result of the wire drawn material P50 with coil number "A030" is "NG1", and the judgment result of the wire drawn material P50 with coil number "B021" is "NG2", so the results of the structure investigation are stored for these two wire drawn materials P50.

[0086] For example, it is stored that coil number "A030" is a "wire break" and coil number "B021" is a "foreign matter." In this way, the results of the microstructural investigation are stored for the drawn wire material P50 for which the judgment result was NG. The microstructural investigation data D17 is input into an information processing device or the like by the worker who performed the microstructural investigation or by a person who received the information from the worker, and the information input from the information processing device is transmitted to the computer 100. The microstructural investigation may be classified into more detailed categories rather than just "wire break" and "foreign matter." For example, the results of the microstructural investigation may be classified and stored as "foreign matter missing," "foreign matter attached," "cuppy wire break," and "cuppy recess."

[0087] 8 to 18, in the manufacturing system 1, the computer 100 can manage data obtained through the process of manufacturing the coil P10 in the casting process S10, inspections such as a twisting test, the process of manufacturing the wire drawn material P50 in the wire drawing process S50, and inspections such as a defect inspection, in association with the coil number P10B. That is, the computer 100 can manage the appearance inspection data D11, the first process data D12, the sampling inspection data D13, the test inspection data D14, the second process data D15, the defect investigation data D16, and the structure investigation data D17 in association with the management data D10 including the coil P10.

[0088] <Characteristics estimation processing> Next, the characteristic estimation process executed by the computer 100 will be described. 19 is a flowchart showing an example of characteristic estimation processing executed by CPU 101. The characteristic estimation processing is realized, for example, by CPU 101 reading and executing characteristic estimation program 202A from program group 202 in hard disk drive 112. In this case, computer 100 functions as characteristic estimation system 100A. Before executing this processing, data related to coil P10 must be prepared in advance for each type data D10D of coil P10. It is desirable to prepare as many pieces of data as possible, taking into consideration the calculation load on CPU 101.

[0089] In step ST101, CPU 101 acquires management data D10 and visual inspection data D11 for coil P10 of product type data D10D. Next, in step ST102, CPU 101 acquires first process data D12. Next, in step ST103, CPU 101 acquires characteristic inspection data (sampling inspection data D13 and test inspection data D14). Next, in step ST104, CPU 101 creates a first regression model D18 based on the data acquired by the processing of steps ST101 to ST103. The processing of steps ST101 to ST104 is realized by the function of regression model creation unit 310.

[0090] The regression model creation unit 310 uses, for example, data indicating the characteristics of the coil P10 as a dependent variable and any data from among the visual inspection data D11, first process data D12, sampling inspection data D13, and test inspection data D14 of the coil P10 as an explanatory variable to create a first regression model D18 by machine learning, in which the visual inspection data D11, first process data D12, sampling inspection data D13, and test inspection data D14 are correlated. The dependent variable may be, for example, data D14B on the number of marks after a twisting test, and the explanatory variables may be, for example, scratch data D11A, scratch data D11B, temperature data D12A to water volume data D12E. The regression model creation unit 310 creates the first regression model D18 for each type of coil P10.

[0091] For example, multivariate analysis is used to create the first regression model D18. There are many methods for performing multivariate analysis. For example, the CPU 101 performs multiple regression analysis using the "sklearn.Linear model.Linear Regression" class of the Python machine learning library "scikit-learn," which allows the regression coefficients of each data to be calculated relatively easily. In this manner, the first regression model D18 is created using machine learning. In this embodiment, the temperature data D12A to the water volume data D12E included in the first process data D12 are stored as average values, thereby reducing the calculation load on the CPU 101. The computer 100 stores the first regression model D18 created in this manner in the file group 203 of the hard disk drive 112.

[0092] Next, in step ST105, CPU 101 calculates an estimate of the characteristics of coil P10 based on first regression model D18. More specifically, when manufacturing a coil P10 of the same type as the coil P10 for which the first regression model D18 to be estimated was created, CPU 101 estimates the characteristics of coil P10 using appearance inspection data D11, first process data D12, and first regression model D18 obtained during the manufacturing of the coil P10. The processing of step ST105 is realized by the function of characteristic estimation unit 320.

[0093] For example, in the above-described example, the CPU 101 can estimate the number of marks on the cross section in the twist test over the entire length of the coil P10 using the flaw data D11A, the mark data D11B, the temperature data D12A to the water volume data D12E, and the first regression model D18. That is, the CPU 101 can estimate the characteristics of the coil P10 over the entire length, which are explanatory variables, based on the data at the time of manufacturing the coil P10, which are explanatory variables, and the first regression model D18. Therefore, the characteristic estimation system 100A can estimate the characteristics of the coil P10 over the entire length.

[0094] <Good / bad judgment process> Next, the quality determination process executed by the computer 100 will be described. 20 is a flowchart showing an example of a quality determination process executed by the CPU 101. The quality determination process is realized, for example, by the CPU 101 reading and executing the quality determination program 202B from the program group 202 in the hard disk drive 112. In this case, the computer 100 functions as a quality determination system 100B. Before executing this process, data on the drawn wire material P50 is prepared in advance for each type data D10D of the coil P10. As with the characteristic estimation process described above, it is desirable to prepare as many pieces of data as possible, taking into account the calculation load on the CPU 101.

[0095] In step ST201, CPU 101 acquires long material data related to a coil P10, which is a long material. For example, CPU 101 acquires data related to the coil P10, which is the material for the drawn wire material P50, as long material data. The long material data acquired in this manner is any data included in the management data D10, the appearance inspection data D11, the first process data D12, the sampling inspection data D13, and the test inspection data D14. Next, in step ST202, CPU 101 acquires second process data D15 (D151, D152, D153). Next, in step ST203, CPU 101 acquires defect investigation data D16 and structure investigation data D17. Next, in step ST204, CPU 101 creates a second regression model D19 based on the data acquired by the processing of steps ST201 to ST203. The processing of steps ST201 to ST204 is realized by the function of the regression model creation unit 330.

[0096] The regression model creation unit 330 uses, for example, data indicating the characteristics of the drawn wire material P50 as a dependent variable and any data from among long material data on the coil P10 (any data from among the appearance inspection data D11, the first process data D12, the sampling inspection data D13, and the test inspection data D14), second process data D15 on the drawn wire material P50, defect investigation data D16, and texture investigation data D17 as explanatory variables, and creates, by machine learning, a second regression model D19 that associates the correlations between the data on the coil P10, the second process data D15, the defect investigation data D16, and the texture investigation data D17. The dependent variable can be, for example, the defect investigation data D16, and the explanatory variables can be, for example, the scratch data D11A, the mark data D11B, the temperature data D12A to the water amount data D12E, and the second process data D151 to D153.

[0097] To create the second regression model D19, for example, multivariate analysis is used, as in the case of the characteristic estimation process described above. In this way, the second regression model D19 is created using machine learning. In this embodiment, the data included in the first process data D12 and the data included in the second process data D15 (D151 to D153) are stored as average values, which can reduce the calculation load on the CPU 101. The computer 100 stores the second regression model D19 created in this way so that it is included in the file group 203 of the hard disk device 112.

[0098] Next, in step ST205, CPU 101 calculates an estimate of the properties of the drawn wire material P50 based on the second regression model D19. More specifically, when manufacturing the drawn wire material P50 for which the second regression model D19 to be estimated has been created, CPU 101 estimates the properties of the drawn wire material P50 by using second process data D151 to D153 obtained during the manufacturing of the drawn wire material P50 and the second regression model D19. The processing of step ST205 is realized by the function of the property estimation unit 340.

[0099] For example, in the above-described example, the CPU 101 can estimate the defect investigation data using the scratch data D11A, the mark data D11B, the temperature data D12A to the water volume data D12E, the second process data D151 to D153, and the second regression model D19.

[0100] Next, in step ST206, the CPU 101 executes a pass / fail judgment of the drawn wire material P50. The processing of this step ST206 is realized by the function of the pass / fail judgment unit 350. The CPU 101 judges, for example, based on a preset threshold value, whether the estimated defect investigation data is data indicating that the drawn wire material P50 is a defective drawn wire material or data indicating that the drawn wire material satisfies the regulations. If the drawn wire material P50 is defective, the judgment result is NG, and if the threshold value is satisfied, the judgment result is OK. If the judgment result is NG, the NG may be classified as "NG1" or "NG2" based on the value of the estimated defect investigation data and a level of a preset threshold value.

[0101] Therefore, the CPU 101 can judge the quality of the properties of the drawn wire material P50 by using the properties of the drawn wire material P50 estimated by the property estimation unit 340. Furthermore, the quality determination system 100B can estimate the properties of the drawn wire material P50 by substituting parameters such as visual inspection data into the second regression model D19. Therefore, the quality determination system 100B can execute a simulation to improve the quality of the drawn wire material P50 by changing the parameters used when manufacturing the drawn wire material P50 and calculating estimated values ​​of the properties of the drawn wire material P50.

[0102] The invention made by the inventor has been specifically described above based on the embodiments thereof, but it goes without saying that the present invention is not limited to the above-described embodiments and can be modified in various ways without departing from the spirit of the invention. [Explanation of symbols]

[0103] 1. Manufacturing System 10 Casting System 70 Wire drawing manufacturing system 100 computers 100A Characteristics Estimation System 100B Good / bad judgment system 112 Hard disk drive 202A Characteristics Estimation Program 202B Good / bad judgement program 310 Regression Model Creation Department 320 Characteristics Estimation Unit 330 Regression Model Creation Department 340 Characteristics Estimation Unit 350 Good / bad judgement section D10 Management Data D11 Visual inspection data D12 1st process data D13 Sampling inspection data D14 Test and inspection data D15, D151, D152, D153 2nd process data D16 Defect investigation data D17 Organizational Survey Data D18 First regression model D19 Second regression model P10 coil P50 Wire drawing material P10F cracked edge S10 Casting Process S50 Wire drawing manufacturing process α~γ process

Claims

1. A property estimation system for estimating properties of a long material manufactured from a cast material, comprising: a regression model creation unit that performs machine learning on the relationship between appearance inspection data showing the inspection results of the appearance of the long material, first process data related to the process of manufacturing the long material, and characteristic inspection data showing the inspection results of the characteristics of the long material, and creates a regression model that represents the correlation between the appearance inspection data, the first process data, and the characteristic inspection data; a characteristic estimation unit that estimates a characteristic of the long material to be estimated based on visual inspection data of the long material to be estimated, first process data, and the regression model; A characteristic estimation system comprising:

2. The characteristic estimation system according to claim 1, the regression model creation unit identifies the appearance inspection data, the first process data, and the characteristic inspection data from a plurality of the appearance inspection data, the first process data, and the characteristic inspection data based on management data that manages the long material; Trait estimation system.

3. The characteristic estimation system according to claim 1, the first process data is received at predetermined intervals from a casting system for manufacturing the elongated material from the casting material; the regression model creation unit creates the regression model using an average value of the plurality of first process data received at the predetermined intervals. Trait estimation system.

4. The characteristic estimation system according to claim 1, The characteristic inspection data includes data indicating a determination result as to whether or not the characteristics of the long material satisfy a predetermined quality. Trait estimation system.

5. The characteristic estimation system according to claim 1, The characteristic inspection data includes data indicating the quality state of the long material. Trait estimation system.

6. A quality assessment system for assessing the quality of a linear material manufactured using a long material manufactured from a cast material, comprising: a regression model creation unit that performs machine learning on the relationship between long material data related to the long material, second process data related to the process of manufacturing the linear material, and characteristic investigation data that indicates the investigation results of the characteristics of the linear material, and creates a regression model that represents the correlation between the long material data, the second process data, and the characteristic investigation data; a characteristic estimation unit that estimates characteristics of the linear material to be estimated based on second process data of the linear material to be estimated and the regression model; a quality determination unit that determines whether or not the estimated value estimated by the characteristic estimation unit satisfies a threshold that defines the quality of the linear material to be estimated; A quality determination system comprising:

7. A method for estimating properties of a long material manufactured from a cast material, comprising: Obtaining appearance inspection data indicating the results of an inspection of the appearance of the long material; Acquire first process data relating to a process for manufacturing the elongated material; Acquire characteristic inspection data indicating the inspection results of the characteristics of the long material; machine learning the relationship between the appearance inspection data, the first process data, and the characteristic inspection data to create a regression model that represents the correlation between the appearance inspection data, the first process data, and the characteristic inspection data; Estimating the characteristics of the elongated material based on the visual inspection data of the elongated material, the first process data, and the regression model; Characteristic estimation method.

8. A method for determining whether a linear material manufactured using a long material manufactured from a cast material is good or bad, comprising: Acquire long material data regarding the long material; acquiring second process data relating to a process for manufacturing the linear material; acquiring characteristic investigation data indicating the investigation results of the characteristics of the linear material; The relationship between the long material data, the second process data, and the characteristic survey data is machine-learned, and a regression model is created that represents the correlation between the long material data, the second process data, and the characteristic survey data; determining whether or not an estimated value of the characteristics of the linear material to be estimated based on second process data of the linear material to be estimated and the regression model satisfies a threshold value that defines the quality of the linear material to be estimated; Method of determining pass / fail.

Images