Method, device and program product for generating learning completion model

Abstract

The present invention provides a method, device and program for generating a learning completed model. The shape data of each winding body (204) of a given number or more corresponding to time information, i.e., a shape data group, is obtained, a first defect rate before maintenance is calculated based on the shape data group of each winding body of the given number read before maintenance, a second defect rate after maintenance is calculated based on the shape data group of each winding body of the given number read after maintenance, and an equipment status diagnosis model (M) is generated using the shape data group of each winding body of the given number read before maintenance when the difference between the first defect rate and the second defect rate is greater than a given value.

Description

Technical Field

[0001] The present disclosure relates to a method, device, and program for generating a learned model for displaying information related to maintenance of production equipment. Background Art

[0002] In order to prevent deterioration, failure, etc. and maintain normal operation of a certain device, a maintenance system is generally installed. Patent document 1 discloses a maintenance system that monitors the occurrence of abnormalities such as drainage pump failure and switchboard grounding in a substation, notifies the equipment personnel when an abnormality occurs, and stores information related to the maintenance work performed by the equipment personnel who have received the notification.

[0003] Prior Art Literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Publication No. 2017-167708 Summary of the invention

[0006] The method for generating a learning-completed model disclosed in the present invention is used for the maintenance of a production device, and the production device comprises: a first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; a first bonding roller, arranged on the side of the first electrode sheet; a second bonding roller, arranged on the side of the second electrode sheet, and bonding the first electrode sheet and the second electrode sheet in pairs with the first bonding roller; a winding core, winding the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor, reading the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the given number or more of the winding bodies, and at the given In the method for generating the learned model, first data indicating the position of the first end face corresponding to time information and second data indicating the position of the second end face corresponding to the time information are obtained from the sensor, wherein the time information indicates the time when the winding bodies more than the given number are read or the time when the winding bodies are produced, and when it is judged based on the first data and the second data that the continuous position of the first end face indicated by the first data and the continuous position of the second end face indicated by the second data intersect, it is indicated that the winding body is defective and the cause of the defect is the first supply reel, the second supply reel, The information of the meaning of any one of the first laminating roller, the second laminating roller, the winding core or the sensor is output to the display device for maintenance, and based on the time information, the first data and the second data of each of the given number of winding bodies generated before the maintenance and different from the time of reading or the time of producing the winding body are extracted, and the first defective rate of the given number of winding bodies read before the maintenance is calculated based on the extracted first data and the second data, and based on the time information, the first defective rate of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding body is extracted. The first data and the second data are extracted, and the second defective rate of the winding bodies having a given number or more read after the maintenance is calculated based on the extracted first data and the second data. When it is judged that the first difference between the first defective rate and the second defective rate is less than a given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are not used to generate the learned model. On the other hand, when it is judged that the first difference is greater than the given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are used to generate the learned model.

[0007] The method for generating a learning-completed model disclosed in the present invention is used for the maintenance of a production device, wherein the production device comprises: a first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; a first bonding roller disposed on the side of the first electrode sheet; a second bonding roller disposed on the side of the second electrode sheet and bonding the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core for winding the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor for moving along the radial direction of each of the given number or more of winding bodies. The first end face of the first electrode sheet and the second end face of the second electrode sheet are read, and in the method for generating the learned model, first data indicating the position of the first end face corresponding to time information and second data indicating the position of the second end face read along the radial direction corresponding to the time information are obtained from the sensor, wherein the time information indicates the time of reading along the radial direction of each of the winding bodies of more than the given number or the time of producing the winding body, and when it is determined based on the first data and the second data that the first data indicates the time when the winding body is read. When the continuity of the position of the first end face indicated by the data intersects the continuity of the position of the second end face indicated by the second data, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor is output to the display device for maintenance. Based on the time information, the first data and the second data of the given number of winding bodies whose reading time or production time is different are extracted before and after the maintenance respectively. When it is judged based on the extracted first data and the second data that the second defective rate of the given number of winding bodies read after the maintenance is greater than a given value, the first data and the second data of the given number of winding bodies read before the maintenance are not used to generate the learned model. On the other hand, when it is judged that the second defective rate is less than the given value, the first data and the second data of the given number of winding bodies read before the maintenance are used to generate the learned model.

[0008] The method for generating a learned model disclosed in the present invention is used for the maintenance of a production device, and the production device comprises: a first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; a first bonding roller, arranged on the side of the first electrode sheet; a second bonding roller, arranged on the side of the second electrode sheet, and bonding the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core, winding the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor, reading the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the given number or more of the winding bodies, in the generation method of the learned model. In the method, first data indicating the position of the first end face corresponding to time information and second data indicating the position of the second end face corresponding to the time information are obtained from the sensor, wherein the time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced, and when it is judged based on the first data and the second data that the continuous position of the first end face indicated by the first data and the continuous position of the second end face indicated by the second data intersect, a signal is generated for judging whether the cause of the defect of the winding body is the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor. A learning model of any one of the sensors is used, and based on the judgment using the learned model, information indicating that the cause of the failure is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor is output to the display device for maintenance. When it is judged based on the time information that the second difference between the first probability and the second probability is less than a given value, the first data and the second data of each of the winding bodies of more than the given number read before the maintenance are not used to update the learned model. On the other hand, when it is judged that the second difference is greater than the given value, the first data and the second data of the winding bodies read before the maintenance are used. The learned model is updated based on the first data and the second data of each of the given number of windings or more that are generated before the maintenance and are different from the time of reading or the time of producing the windings, wherein the first probability is the probability that the defect of the windings is improved by inputting the first data and the second data of each of the given number of windings or more that are generated before the maintenance and are different from the time of reading or the time of producing the windings into the learned model, and the second probability is the probability that the defect of the windings is improved by inputting the first data and the second data of each of the given number of windings or more that are generated after the maintenance and are different from the time of reading or the time of producing the windings into the learned model.

[0009] The method for generating a learning-completed model disclosed in the present invention is used for the maintenance of a production device, wherein the production device comprises: a first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; a first bonding roller, arranged on the side of the first electrode sheet; a second bonding roller, arranged on the side of the second electrode sheet, and bonding the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core, winding the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor, reading the first end face of the first electrode sheet along the radial direction of each of the given number or more of the winding bodies to detect the first end face of the first electrode sheet. and the second end face of the second electrode sheet, in the method for generating the learned model, first data indicating the position of the first end face corresponding to time information and second data indicating the position of the second end face corresponding to the time information are obtained from the sensor, wherein the time information indicates the time of reading the winding bodies of the given number or more or the time of producing the winding bodies, respectively, and when it is judged based on the first data and the second data that the position of the first end face represented by the first data is continuous and the position of the second end face represented by the second data is continuous, a judgment is generated for judging A learned model is used to indicate that the cause of the defect of the winding body is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor. Based on the judgment using the learned model, information indicating that the cause of the defect is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor is output to a display device for maintenance. When it is determined based on the time information that the second probability is greater than a given value, the given number or more of the time information read before the maintenance is displayed. The first data and the second data of each winding body are not used to update the learned model. On the other hand, when it is judged that the second probability is less than the given value, the first data and the second data of each winding body of more than the given number read before the maintenance are used to update the learned model, wherein the second probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each winding body of more than the given number generated after the maintenance and different from the time of reading or the time of producing the winding body into the learned model.

[0010] The device disclosed in the present invention is a device for outputting information, which is used to display information related to the maintenance of a production device, and the production device comprises: a first supply reel, which supplies a first electrode sheet; a second supply reel, which supplies a second electrode sheet; a first bonding roller, which is arranged on the side of the first electrode sheet; a second bonding roller, which is arranged on the side of the second electrode sheet and bonds the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core, which winds the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor, which reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of the given number or more of the winding bodies. , the device for outputting information obtains from the sensor first data indicating the position of the first end face corresponding to the time information and second data indicating the position of the second end face corresponding to the time information, wherein the time information indicates the time of reading the winding bodies of more than the given number or the time of producing the winding bodies, and the device for outputting information comprises: a notification judgment unit, based on the first data and the second data, based on whether the continuity of the position of the first end face indicated by the first data intersects with the continuity of the position of the second end face indicated by the second data, to judge whether the winding body is defective, and if it is defective, the reason for the defect is indicated as The information about the meaning of any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the core or the sensor is output to a display device for maintenance; and a model generating unit, based on the time information, extracts the first data and the second data of each of the given number or more of the winding bodies that are generated before the maintenance and are different from the time of reading or the time of producing the winding bodies, and calculates the first defective rate of the given number or more of the winding bodies read before the maintenance based on the extracted first data and the second data, and extracts the first defective rate of the given number of the winding bodies that are generated after the maintenance and are different from the time of reading or the time of producing the winding bodies based on the time information. The first data and the second data of each of the given number of winding bodies having different moments of the body are extracted, and the second defective rate of the given number of winding bodies read after the maintenance is calculated based on the extracted first data and the second data. The learning completion model is generated by using the first data and the second data of each of the given number of winding bodies read before the maintenance when it is judged that the first difference between the first defective rate and the second defective rate is less than a given value, instead of using the first data and the second data of each of the given number of winding bodies read before the maintenance when it is judged that the first difference between the first defective rate and the second defective rate is greater than the given value.

[0011] The device disclosed in the present invention is a device for outputting information, which is used to display information related to the maintenance of a production device, and the production device comprises: a first supply reel, which supplies a first electrode sheet; a second supply reel, which supplies a second electrode sheet; a first bonding roller, which is arranged on the side of the first electrode sheet; a second bonding roller, which is arranged on the side of the second electrode sheet and bonds the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core, which winds the bonded first electrode sheet and the second electrode sheet to respectively generate a given number of winding bodies or more; and a sensor, which is arranged along the given number of winding bodies or more The first end face of the first electrode sheet and the second end face of the second electrode sheet are read in the radial direction of each winding body, and the device for outputting information obtains from the sensor first data corresponding to time information indicating the position of the first end face and second data corresponding to the time information indicating the position of the second end face, wherein the time information indicates the time of reading along the radial direction or the time of producing the winding body, and the device for outputting information comprises: a notification judgment unit, based on the first data and the second data, based on the Whether the continuity of the position of the first end face intersects with the continuity of the position of the second end face represented by the second data, determines whether the winding body is defective, and if it is defective, outputs information indicating that the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor to a display device for maintenance; and a model generating unit, based on the time information, extracts the first data and the second data of each of the given number of winding bodies whose reading time or production time is different before and after the maintenance, and does not use the first data and the second data of each of the given number of winding bodies read before the maintenance when it is judged that the second defective rate of the given number of winding bodies read after the maintenance is greater than the given value based on the extracted first data and the second data, but uses the first data and the second data of each of the given number of winding bodies read before the maintenance when it is judged that the second defective rate is less than the given value to generate a learned model.

[0012] The device disclosed in the present invention is a device for outputting information for displaying information related to maintenance of a production device, wherein the production device comprises: a first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; a first bonding roller arranged on the side of the first electrode sheet; a second bonding roller arranged on the side of the second electrode sheet and bonding the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core for winding the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor for reading a first end face of the first electrode sheet and a second end face of the second electrode sheet along the radial direction of each of the given number or more of the winding bodies, the device for outputting information from the sensor Acquire the first data indicating the position of the first end face corresponding to the time information and the second data indicating the position of the second end face corresponding to the time information, wherein the time information indicates the time when the winding bodies more than the given number are read or the time when the winding bodies are produced, and the device for outputting information comprises: a model generating unit, which determines whether the winding body is defective based on the first data and the second data, and whether the continuous position of the first end face indicated by the first data intersects with the continuous position of the second end face indicated by the second data, and generates, if it is defective, a model generating unit for judging whether the defective cause is the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, etc. , a learned model indicating the meaning of any one of the core or the sensor; and a notification judgment unit, based on the judgment using the learned model, outputs information indicating that the defect cause is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the core or the sensor to a display device for maintenance. When the model generating unit judges that the second difference between the first probability and the second probability is less than a given value based on the time information, the first data of each of the winding bodies above the given number and the second data read before the maintenance are not used to update the learned model. On the other hand, when it is judged that the second difference is greater than the given value The learned model is updated using the first data and the second data of each of the given number of windings or more read before the maintenance, wherein the first probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings or more generated before the maintenance and different from the time of reading or the time of producing the winding into the learned model, and the second probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings or more generated after the maintenance and different from the time of reading or the time of producing the winding into the learned model.

[0013] The device disclosed in the present invention is a device for outputting information, which is used to display information related to the maintenance of a production device, and the production device comprises: a first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; a first bonding roller, which is arranged on the side of the first electrode sheet; a second bonding roller, which is arranged on the side of the second electrode sheet and bonds the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core, which winds the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor, which reads the first end face of the first electrode sheet and the second end face of the winding bodies along the radial direction of each of the given number or more of the winding bodies. The second end face of the second electrode sheet, the device for outputting information obtains from the sensor first data indicating the position of the first end face corresponding to time information and second data indicating the position of the second end face corresponding to the time information, wherein the time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced, and the device for outputting information comprises: a model generating unit for judging whether the winding body is defective based on the first data and the second data and based on whether the continuity of the position of the first end face indicated by the first data intersects with the continuity of the position of the second end face indicated by the second data. , in the case of a defect, a learning model is generated for judging that the cause of the defect is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor; and a notification judgment unit, based on the judgment using the learning model, outputs information indicating that the cause of the defect is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor to a display device for maintenance. When the model generation unit judges that the second probability is greater than a given value based on the time information, the model generation unit outputs the information to the display device for maintenance. The first data and the second data of each of the given number of winding bodies read before the maintenance are not used to update the learned model. On the other hand, when it is judged that the second probability is less than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to update the learned model, wherein the second probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding body into the learned model.

[0014] The program disclosed herein is a computer-executed program that generates a learned model for maintenance of a production device, wherein the production device comprises: a first supply reel that supplies a first electrode sheet; a second supply reel that supplies a second electrode sheet; a first bonding roller that is disposed on the side of the first electrode sheet; a second bonding roller that is disposed on the side of the second electrode sheet and bonds the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core that winds the bonded first electrode sheet and the second electrode sheet to generate a given number or more of winding bodies respectively; and a sensor that reads the first end face of the first electrode sheet and the second electrode sheet along the radial direction of each of the given number or more of winding bodies. The program causes the computer to execute the following process: obtaining from the sensor first data corresponding to time information indicating the position of the first end face and second data corresponding to the time information indicating the position of the second end face, wherein the time information indicates the time when the winding bodies more than the given number are read or the time when the winding bodies are produced; and in a case where it is judged based on the first data and the second data that the continuous position of the first end face indicated by the first data intersects the continuous position of the second end face indicated by the second data, indicating that the winding body is defective and the cause of the defect is the first supply reel, the second ... The information about the supply reel, the first laminating roller, the second laminating roller, the winding core or any one of the sensors is output to a display device for maintenance; based on the time information, the first data and the second data of each of the given number of winding bodies generated before the maintenance and different from the time of reading or the time of producing the winding bodies are extracted, and the first defective rate of the given number of winding bodies read before the maintenance is calculated based on the extracted first data and the second data; based on the time information, the first data and the second data of each of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding bodies are extracted. The first data and the second data are extracted, and the second defective rate of the winding bodies having a given number or more read after the maintenance is calculated based on the extracted first data and the second data; and in a case where it is judged that the first difference between the first defective rate and the second defective rate is less than a given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are not used to generate the learned model. On the other hand, in a case where it is judged that the first difference is greater than the given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are used to generate the learned model.

[0015] The program disclosed in the present invention is a computer-executed program that generates a learning-completed model for maintenance of a production device, wherein the production device comprises: a first supply reel that supplies a first electrode sheet; a second supply reel that supplies a second electrode sheet; a first bonding roller that is disposed on the side of the first electrode sheet; a second bonding roller that is disposed on the side of the second electrode sheet and bonds the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core that winds the bonded first electrode sheet and the second electrode sheet to generate a given number or more of winding bodies respectively; and a sensor that moves along the given number of winding bodies. The program causes the computer to execute the following process: obtaining from the sensor first data indicating the position of the first end face and second data indicating the position of the second end face corresponding to the time information, wherein the time information indicates the time when the winding bodies are read along the radial direction of each of the winding bodies having a given number or more or the time when the winding bodies are produced; and determining based on the first data and the second data that the first data indicates the time when the winding bodies are read along the radial direction of each of the winding bodies having a given number or more or the time when the winding bodies are produced; and When the continuity of the position of the first end face represented by the data intersects with the continuity of the position of the second end face represented by the second data, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor is output to the display device for maintenance; and based on the time information, the first data and the second data of each of the given number of winding bodies whose reading time or production time is different are extracted before and after the maintenance, respectively. When it is judged based on the extracted first data and the second data that the second defective rate of the given number of winding bodies read after the maintenance is greater than a given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are not used to generate the learned model. On the other hand, when it is judged that the second defective rate is less than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to generate the learned model.

[0016] The program disclosed in the present invention is a computer-executed program that generates a learning-completed model for maintenance of a production device, wherein the production device comprises: a first supply reel that supplies a first electrode sheet; a second supply reel that supplies a second electrode sheet; a first bonding roller that is disposed on the side of the first electrode sheet; a second bonding roller that is disposed on the side of the second electrode sheet and bonds the first electrode sheet and the second electrode sheet in a pair with the first bonding roller; a winding core that winds the bonded first electrode sheet and the second electrode sheet to generate a given number or more of winding bodies respectively; and a sensor that reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the given number or more of the winding bodies, wherein the program The computer is caused to execute the following process: obtaining from the sensor first data corresponding to time information indicating the position of the first end face and second data corresponding to the time information indicating the position of the second end face, wherein the time information indicates the time of respectively reading the given number or more of the winding bodies or the time of producing the winding bodies; and generating a method for judging that the cause of the defect of the winding body is the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding body, when it is judged based on the first data and the second data that the continuous position of the first end face indicated by the first data and the continuous position of the second end face indicated by the second data intersect. a learned model of the meaning of any one of the core or the sensor; based on the judgment using the learned model, information indicating that the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the core or the sensor is output to the display device for maintenance; and in a case where it is judged based on the time information that the second difference between the first probability and the second probability is less than a given value, the first data and the second data of each of the winding bodies above the given number read before the maintenance are not used to update the learned model. On the other hand, in a case where it is judged that the second difference is greater than the given value, the first data and the second data of the winding bodies read before the maintenance are not used to update the learned model. The learned model is updated based on the first data and the second data of each of the given number of windings read before maintenance, wherein the first probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings generated before the maintenance and different from the time of reading or the time of producing the winding into the learned model, and the second probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings generated after the maintenance and different from the time of reading or the time of producing the winding into the learned model.

[0017] The program disclosed in the present invention is a computer-executed program that generates a learning-completed model for maintenance of a production device, wherein the production device comprises: a first supply reel that supplies a first electrode sheet; a second supply reel that supplies a second electrode sheet; a first bonding roller that is arranged on the side of the first electrode sheet; a second bonding roller that is arranged on the side of the second electrode sheet and bonds the first electrode sheet to the second electrode sheet in a pair with the first bonding roller; a winding core that winds the bonded first electrode sheet and the second electrode sheet to respectively generate a given number or more of winding bodies; and a sensor that reads the first electrode sheet along the radial direction of each of the given number or more of the winding bodies. The first end face of the electrode sheet and the second end face of the second electrode sheet, the program causes the computer to execute the following process: obtaining from the sensor first data corresponding to time information indicating the position of the first end face and second data corresponding to the time information indicating the position of the second end face, wherein the time information indicates the time of reading the winding bodies of the given number or more or the time of producing the winding bodies; in a case where it is judged based on the first data and the second data that the continuous position of the first end face indicated by the first data and the continuous position of the second end face indicated by the second data intersect, Generate a learned model for judging that the cause of the defect of the winding body is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor; based on the judgment using the learned model, output information indicating that the cause of the defect is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor to a display device for maintenance; and when it is judged that the second probability is greater than a given value based on the time information, the given value read before the maintenance is output. The first data and the second data of each of the above winding bodies are not used to update the learned model. On the other hand, when it is judged that the second probability is less than the given value, the first data and the second data of each of the winding bodies above the given number read before the maintenance are used to update the learned model, wherein the second probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each of the winding bodies above the given number generated after the maintenance and different from the time of reading or the time of producing the winding body into the learned model. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a network diagram including a maintenance display device and a production device using the maintenance display device.

[0019] Figure 2This is a flowchart for explaining the overall processing procedure of the maintenance display device.

[0020] Figure 3A The diagram exemplifies the structure of a winding portion for producing a wound body in a production device.

[0021] Figure 3B is a perspective view illustrating a wound body produced in a winding unit.

[0022] Figure 4A This is a schematic diagram illustrating an example of how a winding body is inspected by an inspection machine.

[0023] Figure 4B It is a schematic diagram illustrating the cross-sectional shape along the radial direction of the wound body.

[0024] Figure 4C This is an example of an inspection machine scanning Figure 4B Figure 2 is an image generated by taking a cross section of the wound body shown.

[0025] Figure 5 It is a schematic diagram showing the correspondence between the cross-sectional shape of the wound body and the shape data.

[0026] Figure 6 This is a block diagram illustrating the functional configuration of the maintenance display device according to the first embodiment.

[0027] Fig. 7A This is a diagram illustrating actual production performance data.

[0028] Figure 7B This is a diagram illustrating actual production performance data.

[0029] Figure 8 This is a diagram illustrating an example of maintenance performance data.

[0030] Fig. 9 This is a sequence diagram schematically illustrating the overall flow of processing in the maintenance display device.

[0031] Fig.10 This is a sequence diagram schematically illustrating the overall flow of processing in the maintenance display device.

[0032] Fig.11 This is a flowchart for explaining the processing executed by the maintenance effect determination unit in the learning processing.

[0033] Fig. 12A This is a conceptual diagram for explaining how the effectiveness of maintenance work in the learning process is determined.

[0034] Fig. 12B This is a conceptual diagram for explaining how the effectiveness of maintenance work in the learning process is determined.

[0035] Fig.13A This is a conceptual diagram for explaining how the effectiveness of maintenance work in the learning process is determined.

[0036] Fig. 13B This is a conceptual diagram for explaining how the effectiveness of maintenance work in the learning process is determined.

[0037] Fig.14 This is a flowchart for explaining the processing executed by the device state diagnosis model generation unit in the learning process.

[0038] Fig.15 This is a flowchart for explaining the processing executed by the device status diagnosis unit in the recognition processing.

[0039] Fig.16 This is a flowchart for explaining the processing executed by the notification determination unit in the recognition processing.

[0040] Fig.17A It is a diagram showing a specific example of maintenance group information.

[0041] Fig. 17B It is a diagram showing a specific example of the maintenance plan list.

[0042] Fig.18 This is a flowchart for explaining the processing executed by the maintenance effect determination unit in the updating process.

[0043] Fig.19A This is a conceptual diagram for explaining how the effect of maintenance work in the update process is determined.

[0044] Fig.19B This is a conceptual diagram for explaining how the effect of maintenance work in the update process is determined.

[0045] Fig. 20 This is a flowchart for explaining the processing executed by the device state diagnosis model generation unit in the update processing.

[0046] Fig.21 It is a diagram illustrating the configuration of a maintenance display device according to the second embodiment.

[0047] Fig. 22 This is a flowchart for explaining the processing executed by the maintenance effect determination unit in the second embodiment.

[0048] Fig.23 It is a diagram illustrating the configuration of a maintenance display device according to a third embodiment.

[0049] Fig.24This is a flowchart for explaining the processing executed by the equipment state diagnosis model generation unit in the third embodiment.

[0050] Fig.25 This is a flowchart for explaining the processing executed by the notification determination unit in the third embodiment.

[0051] Fig.26A This is a diagram for explaining a modified example of a method of determining whether or not a maintenance work is effective, which is performed by a maintenance effect determination unit in a learning process.

[0052] Fig.26B This is a diagram for explaining a modified example of a method of determining whether or not a maintenance work is effective, which is performed by a maintenance effect determination unit in a learning process.

[0053] Fig.27A It is a diagram for explaining a modified example of a method of determining whether or not a maintenance work is effective by a maintenance effect determination unit in an update process.

[0054] Fig.27B It is a diagram for explaining a modified example of a method of determining whether or not a maintenance work is effective by a maintenance effect determination unit in an update process.

[0055] Explanation of symbols

[0056] 10, 10A, 10B servers;

[0057] 50 1st supply reel;

[0058] 51 2nd supply reel;

[0059] 100, 100A, 100B maintenance display device;

[0060] 110, 110B storage unit;

[0061] 111 Production results database;

[0062] 112 Equipment status diagnosis model database;

[0063] 113 Maintaining the actual results database;

[0064] 114 No effect equipment status diagnosis model database;

[0065] 120, 120A, 120B control unit;

[0066] 121 Equipment status diagnosis department;

[0067] 122, 122B Notification to the Determination Department;

[0068] 123, 123A, 123B maintenance effect determination unit;

[0069] 124, 124B equipment status diagnosis model generation unit;

[0070] 130 Notification Department;

[0071] 131 Alarm Department;

[0072] 132 Display Department;

[0073] 200 production units;

[0074] 201 winding part;

[0075] 202 Sheet 1;

[0076] 203 Sheet 2;

[0077] 204 winding body;

[0078] 205A: 1st laminating roller;

[0079] 205B: second laminating roller;

[0080] 206 core;

[0081] 206M core rotation drive unit;

[0082] 207 Inspection machine;

[0083] 208 turntable;

[0084] 209 cutting section;

[0085] 210 pressing part;

[0086] 211 joint welding part;

[0087] 212 with adhesive part;

[0088] 213 Roller. DETAILED DESCRIPTION

[0089] In the technology disclosed in Patent Document 1, after an abnormality occurs in the equipment, the personnel involved in the equipment are notified. Therefore, the time for the maintenance by the personnel involved in the equipment is after the abnormality occurs. When the maintenance is performed after the abnormality occurs, the operation of the equipment needs to be stopped, so it is desirable to notify at the time when it is determined that maintenance is required before the abnormality occurs. Therefore, it is necessary to detect the signs of abnormality occurring in the equipment.

[0090] An object of the present disclosure is to provide a method, device, and program for generating a learned model for detecting a sign of abnormality.

[0091] Hereinafter, each embodiment of the present disclosure will be described in detail with reference to the drawings. However, detailed descriptions beyond necessity, for example, detailed descriptions of well-known matters, repeated descriptions of substantially the same structures, etc. may be omitted.

[0092] In addition, the following description and the referenced drawings are provided to enable those skilled in the art to understand the present disclosure, and are not intended to limit the technical solutions of the present disclosure.

[0093] (First embodiment)

[0094] <Maintenance Display Device 100 and Production Device 200>

[0095] Figure 1 1 is a network diagram including the maintenance display device 100 according to the first embodiment of the present invention and the production device 200 to which the maintenance display device 100 is applied. The maintenance display device 100 described in this embodiment is a device for performing maintenance display of the production device 200 for producing lithium-ion secondary batteries. Figure 1 In the example shown, the maintenance display device 100 is applied to one production device 200, but the present invention is not limited thereto, and one maintenance display device may be applied to a plurality of production devices. In addition, in the present embodiment, the maintenance display device 100 is described as a device, but the present invention is not limited thereto, and may be a maintenance display system in which each structure is connected via a network.

[0096] The maintenance display device 100 includes a server 10 having a storage unit 110 and a control unit 120, and a notification unit 130. The server 10 is connected to be able to communicate with the production device 200 via a network NT. The network NT is, for example, a public network such as the Internet, or a local area network such as a company's in-house LAN (Local Area Network).

[0097] The server 10 is, for example, a general-purpose computer such as Figure 1 As shown, a storage unit 110 and a control unit 120 are provided.

[0098] The storage unit 110 includes, for example, a main storage device (not shown) such as a ROM (Read Only Memory) or a RAM (Random Access Memory), and / or an auxiliary storage device (not shown) such as a HDD (Hard Disk Drive), an SSD (Solid State Drive), or a flash memory.

[0099] The control unit 120 is a hardware processor (not shown) such as a CPU (Central Processing Unit), and controls the entire maintenance display device 100 by developing and executing a program stored in the storage unit 110 .

[0100] The storage unit 110 and the control unit 120 may not be configured as an integrated computer. That is, the storage unit 110 and the control unit 120 may be configured as separate units and arranged at separate locations as long as they are configured to communicate with each other. Figure 1 The operation unit (not shown) receives operation input from the outside. The details of the storage unit 110 and the control unit 120 will be described later.

[0101] exist Figure 1 In the example shown, the notification unit 130 is included in the production device 200 and is connected to the server 10 via the network NT. The notification unit 130 notifies the user of the maintenance display device 100 based on the control of the control unit 120. In the present embodiment, the user of the maintenance display device 100 includes the administrator of the maintenance display device 100 or the user who uses the production device 200 to wind the body (see the following description). Figure 3B )’s production operators, etc.

[0102] like Figure 1 As shown, the notification unit 130 includes an alarm unit 131 and a display unit 132. The alarm unit 131 is, for example, a buzzer, a lamp, etc., which warns the user through sound, light, etc. The display unit 132 is, for example, a display device such as a liquid crystal display, an organic EL display, etc., which is a structure for displaying the content of the warning. In addition, the notification unit 130 may include, in addition to the alarm unit 131 and the display unit 132, a sending unit that sends an email containing the content of the warning to, for example, an email address of a pre-registered user.

[0103] In this embodiment, the production device 200 is a production device for lithium-ion secondary batteries. Figure 1 As shown, the production device 200 includes a winding unit 201 and an inspection machine 207. The winding unit 201, which will be described in detail later, winds up the positive electrode sheet and the negative electrode sheet to produce a wound body. The inspection machine 207 inspects the wound body produced by the winding unit 201.

[0104] In addition, Figure 1 In the example shown in FIG. 1 , the notification unit 130 is included in the production device 200, but the present invention is not limited thereto, and the notification unit 130 may also be provided outside the production device 200. Figure 1In the illustrated example, the notification unit 130 is connected to the server 10 via the network NT, but the present invention is not limited thereto, and the server 10 and the notification unit 130 may be directly connected without interposing the network NT.

[0105] In addition, in this embodiment, the case where the production device 200 is a production device for lithium-ion secondary batteries is described, but the present invention is not limited to this. The maintenance display device of the present invention can also be applied to production equipment other than the production device for lithium-ion secondary batteries. Furthermore, the maintenance display device of the present invention can also be applied to various equipment other than production equipment.

[0106] Figure 2 This is a flowchart for explaining the overall processing procedure of the maintenance display device 100 .

[0107] In step S1 , the control unit 120 causes the winding unit 201 of the production device 200 to produce a wound body.

[0108] In step S2, the control unit 120 causes the inspection machine 207 to inspect the produced wound body. Details of the inspection of the wound body by the inspection machine 207 will be described later.

[0109] In step S3, the control unit 120 causes the storage unit 110 to store the inspection result by the inspection machine 207. At the same time, in step S4, the control unit 120 determines whether the winding body is a defective product as a result of the inspection by the inspection machine 207. If it is determined that it is not a defective product (step S4: No), the control unit 120 advances the process to step S5, and if it is determined that it is a defective product (step S4: Yes), the control unit 120 advances the process to step S6.

[0110] When it is determined that the product is not defective, in step S5 , the control unit 120 causes the production device 200 to supply the wound body to the next step.

[0111] If the product is determined to be defective, in step S6, the control unit 120 notifies the notification unit 130 that a defective product has been detected. Details of the notification performed by the notification unit 130 will be described later.

[0112] In step S7 , the control unit 120 causes the production device 200 to discard the wound body determined to be a defective product.

[0113] In addition, Figure 2In step S5 and step S7 of the flowchart shown in the figure, the control unit 120 causes the production device 200 to supply the winding body to the next process or to discard the winding body, but the present invention is not limited to this. For example, the user of the maintenance display device 100 may be notified via the notification unit 130 to supply the winding body to the next process or to discard the winding body, thereby causing the user to supply or discard the winding body.

[0114] Next, the winding unit 201 and the inspection machine 207 of the production device 200 will be described in detail.

[0115] <Winding Section 201>

[0116] Figure 3A The diagram illustrates the configuration of the winding unit 201 and the inspection machine 207 .

[0117] like Figure 3A As shown, the winding section 201 includes a first supply reel 50, a second supply reel 51, a first laminating roller 205A, a second laminating roller 205B, a core 206, a core rotation driving section 206M, an inspection machine 207, a turntable (index table) 208, a cutting section 209, a pressing section 210, a joint welding section 211, a tape sticking section 212, and a cylinder 213. The winding section 201 is a device for producing a winding body 204 by laminating the first sheet 202 supplied from the first supply reel 50 and the second sheet 203 supplied from the second supply reel 51 by the first laminating roller 205A and the second laminating roller 205B and winding them on the core 206. The core rotation driving section 206M drives the core 206 at a desired rotation speed.

[0118] The first sheet 202 is, for example, a sheet-shaped member (positive electrode sheet) coated with a positive electrode material, and the second sheet 203 is, for example, a sheet-shaped member (negative electrode sheet) coated with a negative electrode material. The first sheet 202 is an example of the first electrode sheet of the present invention, and the second sheet 203 is an example of the second electrode sheet of the present invention. In addition, in the above example, the first sheet 202 is a positive electrode sheet and the second sheet 203 is a negative electrode sheet, but the present invention is not limited thereto, and the first sheet 202 may be a negative electrode sheet and the second sheet 203 may be a positive electrode sheet.

[0119] The turntable 208 rotates the winding cores 206 along a circular track while gradually rotating at a predetermined angle. Thus, one of the three winding cores 206 (206α, 206β, 206γ) is placed at a winding position. The winding position is a position where the winding core 206 can be rotated by the winding core rotation drive unit 206M.

[0120] In addition, Figure 3AIn the illustrated example, the turntable 208 sequentially switches three winding cores, but the present invention is not limited thereto, and the number of winding cores 206 held by the turntable 208 may be more than one as long as it is two or more.

[0121] The cutting unit 209 cuts the first sheet 202 and the second sheet 203 when winding on one winding core 206 is completed. At this time, the pressing unit 210 presses the winding body 204 wound on the winding core 206 and presses the vibration of the end portions of the first sheet 202 and the second sheet 203 after being cut. Figure 3A In the example shown, the cutting unit 209 is arranged at a position where the first sheet 202 and the second sheet 203 are cut before being bonded together, but may be arranged at a position where the first sheet 202 and the second sheet 203 are cut after being bonded together.

[0122] The joint welding section 211 welds the current collecting joint to the first sheet 202. The tape sticking section 212 fixes the winding body 204 with a tape so that the winding body 204 does not vibrate when the winding body 204 is cut by the cutting section 209 after winding in the winding core 206 is completed. The roller 213 adjusts the tension applied to the first sheet 202 and the second sheet 203 via the second laminating roller 205B.

[0123] Figure 3B 2 is a perspective view illustrating a wound body 204 produced in the winding unit 201. Figure 3B , the first sheet 202 and the second sheet 203 constituting the winding body 204 are shown in a state where the terminal ends (the ends cut by the cutting portion 209) are not wound. Figure 3B As shown, the width (length along the axial direction of the wound body 204 ) of the second sheet 203 is formed larger than that of the first sheet 202 .

[0124] <Inspection Machine 207>

[0125] The inspection machine 207 inspects the produced wound body 204. The inspection machine 207 is, for example, a SS-OCT (Swept Source-Optical Coherence Tomography) device. The inspection machine 207 is an example of the sensor of the present invention.

[0126] Figure 4A 2 is a schematic diagram illustrating the inspection machine 207 inspecting the winding body 204. Figure 4A As shown, the inspection machine 207 scans the light L irradiated on the wound body 204 to be inspected from the radial inside to the outside of the wound body 204 , and generates an image (shape data) representing the shape of the internal structure of the wound body 204 using the coherence of the light L.

[0127] Figure 4B 2 is a schematic diagram illustrating a cross-sectional shape along the radial direction of the wound body 204. Figure 4C This is an example of the inspection machine 207 Figure 4B FIG. 2 is a diagram showing shape data I generated by scanning a cross section of the wound body 204 shown in FIG. Figure 4B as well as Figure 4C In the figure, the up-down direction corresponds to the axial direction of the wound body 204, and the left-right direction corresponds to the radial direction of the wound body 204, respectively.

[0128] like Figure 4B As shown, in the cross section along the radial direction of the roll 204, the first sheet 202 and the second sheet 203 having a width greater than the first sheet 202 are alternately stacked. The inspection machine 207 extracts the positions of the two ends in the radial direction of the roll 204, the two ends in the axial direction of the first sheet 202, and the two ends in the axial direction of the second sheet 203, and images them. Figure 4C In the example shown (shape data I), the diamond α corresponds to the first data indicating the positions of both ends of the first sheet 202 , and the black circle β corresponds to the second data indicating the positions of both ends of the second sheet 203 .

[0129] When the winding body 204 is produced in the winding unit 201, defective products may be produced. The inspection machine 207 generates shape data representing the cross-sectional shape along the radial direction of the winding body 204 as described above, and stores it in the storage unit 110 as shape data. In addition, the inspection result determined based on the shape data is also stored in the storage unit 110. In addition, the determination based on the shape data can be performed by the inspection machine 207 or by the storage unit 110. Figure 1 The control unit 120 shown in FIG. Figure 1 or Figure 3A Other structures not shown in the figure are carried out.

[0130] Figure 5 Schematic diagram showing the correspondence between the cross-sectional shape of the wound body 204 and the shape data. Figure 5 A cross-sectional view of the winding body 204 is illustrated in the upper part of FIG. Figure 5 In the example shown, the deviation of the first sheet 202 increases as the process proceeds to the right.

[0131] exist Figure 5 The lower part of the example shows the Figure 5 The shape data I1 to I6 are generated based on the cross-sectional shape of the wound body 204 illustrated in the upper part.

[0132] In the case where the displacement of the first sheet 202 is large, as in the shape data I5 and I6, the continuous position of the upper end face of the first sheet 202 (the first end face of the present invention) represented by the first data may intersect with the continuous position of the upper end face of the second sheet 203 (the second end face of the present invention) represented by the second data. In other words, the first imaginary line segment connecting the plurality of positions of the upper end face of the first sheet 202 (the first end face of the present invention) represented by the first data may intersect with the second imaginary line segment connecting the plurality of positions of the upper end face of the second sheet 203 (the second end face of the present invention) represented by the second data. In this way, the winding body 204 having a cross-sectional shape in which the continuous position of the upper end face of the first sheet 202 and the continuous position of the upper end face of the second sheet 203 intersect is judged as "defective" by the inspection machine 207 as an inspection result.

[0133] exist Figure 5 In the shape data I3 and I4, although the position of the first sheet 202 is offset, the continuity of the position of the upper end surface of the first sheet 202 does not intersect the continuity of the position of the upper end surface of the second sheet 203. The winding body 204 having such a cross-sectional shape is judged as "passed" by the inspection machine 207 as an inspection result.

[0134] Figure 5 The shape data I1 and I2 shown show that the positional deviation of the first sheet 202 is smaller than that of the shape data I3 to I6. The wound body 204 having such a cross-sectional shape is judged as "good" by the inspection machine 207 as a result of inspection.

[0135] For example, "good", "pass", and "bad" may be determined according to the following criteria. The inspection machine 207 determines the shortest distance between the position of the upper end surface of the first sheet 202 and the position of the upper end surface of the second sheet 203. If the determined distance is "0", the inspection machine 207 determines it as "bad", if the determined distance is greater than 0 and below a given threshold, the inspection machine 207 determines it as "pass", and if the determined distance is greater than a given threshold, the inspection machine 207 determines it as "good". The given threshold can be appropriately set according to the balance of battery design, required safety, or cost.

[0136] exist Figure 2 In step S4 of the flowchart shown, it is determined whether the winding body 204 is a defective product. In the determination of this step, the winding body 204 determined to be a defective product is the winding body 204 whose inspection result is "defective". On the other hand, in the determination of this step, the winding body 204 whose inspection result is "good" or "pass" is determined to be not a defective product.

[0137] Thus, when the continuity of the position of the upper end surface indicated by the first data and the continuity of the position of the upper end surface indicated by the second data intersect, the winding body 204 is determined to be defective. Figure 3A The deviation may be caused by a defect in each structure of the winding section 201. Specifically, at least one of the first supply reel 50, the second supply reel 51, the first laminating roller 205A, the second laminating roller 205B, the winding core 206, the inspection machine 207, the cutting section 209, the pressing section 210, the joint welding section 211, the tape pasting section 212, and the roller 213 may be the cause of the deviation.

[0138] It is known from experience that the shape data obtained by the inspection machine 207 may differ slightly depending on which structural defect of the winding unit 201 is the cause of the defect in the winding body 204. Figure 5 In the example shown, the differences between the shape data are exaggerated, but in reality the differences are minute and it is difficult to tell whether there are differences by human eyes. Therefore, it is almost impossible for a human to tell which structure of the winding unit 201 has a problem by visually observing a plurality of shape data.

[0139] In the maintenance display device 100 according to the embodiment of the present invention, a large amount of shape data is collected, and machine learning is performed based on the data to determine which structure of the winding unit 201 is most likely to have a bad condition, and based on the result, the structure recommended for maintenance is notified. As a result, it is expected that the structure to be maintained can be maintained at the stage of a sign before the winding body 204 actually has an abnormality. The various processes performed by the maintenance display device 100 according to the embodiment of the present invention are described in detail below.

[0140] In addition, Figure 5 In the example shown, "good", "pass", and "bad" are determined based on the relative positions (offset from parallel) of the line segments (first imaginary line segment and second imaginary line segment) formed by the continuous positions of the upper end surfaces of the first sheet 202 and the second sheet 203. However, the present invention is not limited to this, and for example, "good", "pass", and "bad" may be determined based on the line segments formed by the continuous positions of the lower end surfaces.

[0141] In addition, in this specification, for easy understanding of the description, Figure 5 Although the defect of the winding body 204 is described as follows, such as the first sheet 202 being offset relative to the second sheet 203 over time as shown in the figure, the defect of the winding body in the present invention is not limited to this. The defect of the winding body includes not only the case where the first sheet is offset, but also the case where the second sheet is offset, and the case where both the first sheet and the second sheet are offset at the same time. In addition, it includes not only the case where Figure 5 The above-mentioned case includes a continuous skewed deviation of the position of the upper end of the first sheet as shown in the example, and also includes a defect that the first sheet or the second sheet crosses due to fluctuation or only a part of the position is deviated from the continuous position. Furthermore, in the case where all or part of the position of the upper end or the lower end of the first sheet or the second sheet cannot be obtained in the shape data, it is impossible to determine whether the first sheet and the second sheet cross, and therefore, it also includes the following case, that is, when the number of positions of the upper end or the lower end of the first sheet or the second sheet obtained is less than a certain value, it is determined to be defective.

[0142] <Maintenance Display Device 100>

[0143] Next, the functional structure and operation of the maintenance display device 100 for displaying the maintenance work of the above-mentioned production device 200 are described in detail. In addition, the so-called maintenance work in this embodiment means that the production device 200 is appropriately adjusted in various structures, parts are replaced, etc., so that no defect occurs in the winding body 204 produced by the production device 200. The maintenance work is performed by the operator who actually operates the production device 200.

[0144] <Storage Unit 110>

[0145] Figure 6 1 is a block diagram illustrating the functional structure of the maintenance display device 100 according to the first embodiment. As described above, the maintenance display device 100 includes a storage unit 110, a control unit 120, and a notification unit 130 (see Figure 1 ).

[0146] like Figure 6 As shown, the storage unit 110 includes a production performance database 111 , a facility status diagnosis model database 112 , and a maintenance performance database 113 .

[0147] The production result database 111 is a database in which production result data related to the production results of the production device 200 are registered. The production result data includes the production date and time of the produced wound body 204 and the shape data of the wound body 204.

[0148] Fig. 7A as well as Figure 7B This is a diagram illustrating an example of actual production performance data PD. Fig. 7A In the example, part of the actual production performance data PD is shown in a table format. Fig. 7A As shown, the production actual result data PD includes various data such as “production date and time (reading date and time)”, “equipment”, “inspection result”, “first sheet”, “second sheet”, and “shape data ID”.

[0149] The "production date (read date)" data is data related to the production date when the winding body 204 was produced, and the date (read date) when the shape data was inspected and read by the inspection machine 207. The production date and the read date may be the same or different, and either one may be included in the actual production results data PD. The "production date (read date)" data is an example of time information of the present invention. The "equipment" data is data used to identify the equipment that gives the actual production results when there are multiple production devices 200. Fig. 7A , as an example, identifiers "A", "B", and "C" of different production devices 200 are shown.

[0150] The “inspection result” data indicates the inspection result of the wound body 204 produced in the production device 200 (refer to Figure 5 ) data. Fig. 7A In FIG. 1 , as an example, three levels of inspection results, namely, “good”, “pass”, and “poor”, are shown.

[0151] The "first sheet material" data and the "second sheet material" data are data related to the material used to produce the wound body 204. As the "first sheet material" data and the "second sheet material" data, identifiers for identifying the respective materials are stored.

[0152] The “shape data ID” is shape data indicating the cross-sectional shape of the wound body 204 (see Figure 5 ) has established a corresponding identification number. Figure 7B , the correspondence between the shape data ID and the shape data is illustrated.

[0153] Each data except the shape data in the production actual result data PD is automatically or manually input by the operator and registered in the production actual result database 111 each time the wound body 204 is produced in the production device 200. The shape data is input by the inspection machine 207 (see Figure 1 or Figure 4A ) is generated during the inspection and registered in correspondence with the shape data ID. That is, the production actual result data PD substantially includes the shape data of the winding body 204. Thus, in the production actual result database 111, the production actual result data PD of the produced winding body 204 is registered every time the winding body 204 is produced.

[0154] The equipment state diagnosis model database 112 is a database in which a plurality of equipment state diagnosis models M are registered. The so-called equipment state diagnosis model M is a learned model that is used to diagnose whether maintenance work is required on the production device 200 and serves as a basis for diagnosis. The equipment state diagnosis model M is a learned model that has learned what kind of maintenance work is effective for what kind of defect when the production device 200 that produces defective products is improved through maintenance work (the production ratio of defective products is reduced). More specifically, the equipment state diagnosis model M is a collection of data including shape data of a winding body containing a plurality of defective products and the content of maintenance work performed to improve the defect of the defective product. The equipment state diagnosis model M is generated by the equipment state diagnosis model generation unit 124 described later.

[0155] The equipment status diagnosis model M is generated each time a maintenance operation is performed that reduces the production ratio of defective products in the production of winding bodies thereafter. That is, for example, the equipment status diagnosis model M related to the maintenance operation performed yesterday and the equipment status diagnosis model M related to the maintenance operation performed today are generated independently.

[0156] In addition, the form of the equipment state diagnosis model M is not particularly limited, but in order to further improve the diagnosis accuracy, it is desirable to adopt a machine learning model such as a neural network model. The selection of the model used in the equipment state diagnosis model M can be performed by the user of the maintenance display device 100 via an operation unit not shown, etc., or by the equipment state diagnosis model generation unit 124.

[0157] The maintenance actual result database 113 is a database in which maintenance actual result data MD related to the maintenance work actually performed on the production device 200 is registered. The maintenance actual result data MD includes, for example, equipment data for identifying the production device 200, data related to the date and time of performing the maintenance work (maintenance date and time), data indicating the content of the maintenance work performed, etc. For example, in the case of a maintenance work that ends in a short time of a few minutes, the maintenance date and time can be either the start time or the end time of the maintenance work. In addition, in the case of a maintenance work that takes a long time, such as several hours, the maintenance date and time is preferably set to the central time of the maintenance work. Figure 8 The diagram exemplifies the maintenance actual result data MD. The maintenance actual result data MD is obtained by, for example, a worker who actually performed the maintenance work on the production device 200 immediately after the maintenance work is performed. Figure 1 The information is input into the maintenance display device 100 through an operation unit (not shown).

[0158] <Control Unit 120>

[0159] like Figure 6As shown, the control unit 120 includes a device state diagnosis unit 121 , a notification determination unit 122 , a maintenance effect determination unit 123 , and a device state diagnosis model generation unit 124 .

[0160] The equipment state diagnosis unit 121 diagnoses the state of the production device 200 using the shape data of the newly produced winding body 204 in the production device 200 and the equipment state diagnosis model M. The diagnosis result is calculated as a consistency C indicating the degree of consistency between the shape data of the newly produced winding body 204 and the past shape data included in the equipment state diagnosis model M. Here, the equipment state diagnosis model M includes the content of the maintenance work and the shape data before the time point when the maintenance work is performed. Therefore, the consistency C is a value indicating the degree to which the defect of the winding body 204 is reduced by performing the maintenance work included in the equipment state diagnosis model M when a defect of the winding body 204 having the shape data included in the equipment state diagnosis model M occurred in the past. In other words, the consistency C between the shape data of the newly produced winding body 204 and the shape data included in the equipment state diagnosis model M indicates the probability that the defect of the winding body 204 will be improved by performing the maintenance included in the equipment state diagnosis model M.

[0161] In addition, as for the method of calculating the consistency C by comparing the multiple (m) shape data of the newly produced winding body 204 with the multiple (n) shape data in the past included in the equipment status diagnosis model M, pattern matching or deep learning using the feature amount of multiple shape data with dimension compression can be appropriately used. In addition, the consistency can also be calculated based on the distance between the vectors obtained from each shape data.

[0162] The notification determination unit 122 determines whether to perform a notification related to the maintenance work of the production device 200 based on the consistency C. The notification determination unit 122 determines to perform a notification indicating that the maintenance work should be performed when the consistency C is greater than a given threshold, and determines not to perform a notification when the consistency C is less than a given threshold. The notification related to the maintenance work includes an alarm for attracting the user's attention, a display notifying the content of the maintenance work whose effect can be foreseen by executing it, and the like.

[0163] The maintenance effect determination unit 123 determines whether there is an effect of the maintenance work on the production device 200. ... based on, for example, the defective rate (the ratio of defective products to the total number of products produced) before and after the maintenance work or the shape data of the winding body 204 before and after the maintenance work (see Figure 5 ), to determine whether there is any effect of maintenance work.

[0164] The equipment state diagnosis model generator 124 generates an equipment state diagnosis model M based on the maintenance actual achievement data MD determined to be effective and the shape data of defective products produced before the maintenance work is performed. The equipment state diagnosis model M generated by the equipment state diagnosis model generator 124 is registered in the equipment state diagnosis model database 112 described above.

[0165] <Overall Flow of Processing in Maintenance Display Device 100>

[0166] Below, refer to Fig. 9 as well as Fig.10 For Figure 6 The overall flow of processing in the maintenance display device 100 having the functional configuration shown will be described. Fig. 9 as well as Fig.10 It is a sequence diagram schematically illustrating the overall flow of processing in the maintenance display device 100 .

[0167] exist Fig. 9 , an overview of the learning process in the maintenance display device 100 and the recognition process using the learned model generated by the learning process is shown.

[0168] [Learning Process]

[0169] The learning process in the maintenance display device 100 is a process for generating a learning model (equipment status diagnosis model M) that learns what kind of shape data the defective product is improved by what kind of maintenance work when a defective product is produced by the production device 200. Therefore, the premise of the learning process is to perform maintenance work before starting the learning process.

[0170] In step S11, the maintenance effect determination unit 123 obtains the maintenance work performed before (before) the start of the learning process. Figure 8 The shape data (see FIG. 7 ) included in the production performance data PD (see FIG. 7 ) of the plurality of wound bodies 204 produced (with the maintenance date and time shown earlier) Figure 5 ), and based on this data, calculate the defect rate Nf before maintenance work before . Defective rate Nf before For example, it is calculated by dividing the number of wound bodies 204 judged to be defective among the wound bodies 204 produced before the maintenance work by the total number of production before the maintenance work. before This is an example of the first defective rate of the present invention.

[0171] In step S12, the maintenance effect determination unit 123 obtains the maintenance result after the maintenance operation ( Figure 8The shape data included in the production actual results data PD of the plurality of winding bodies 204 produced after the maintenance date and time shown in the figure is used to calculate the defect rate Nf after the maintenance work based on the data. after . Defective rate Nf after For example, it is calculated by dividing the number of wound bodies 204 judged to be defective among the wound bodies 204 produced after the maintenance work by the total number of production after the maintenance work. after This is an example of the second defective rate of the present invention.

[0172] In step S13, the maintenance effect determination unit 123 calculates the defect rate Nf before and after the maintenance work. before and f after The comparison is made to determine whether the maintenance work is effective. Details of the process of determining the effect of the maintenance work performed by the maintenance effect determination unit 123 in the learning process will be described later.

[0173] If it is determined in step S13 that the maintenance work is effective, the maintenance effect determination unit 123 in step S14 updates the maintenance actual result data MD (see Figure 8 ) is output to the equipment status diagnosis model generation unit 124.

[0174] In step S15, the equipment state diagnosis model generation unit 124 generates the equipment state diagnosis model M using the maintenance achievement data MD determined to be effective. The equipment state diagnosis model M will be described in detail later.

[0175] In step S16, the device state diagnosis model generation unit 124 registers the generated device state diagnosis model M in the device state diagnosis model database 112 (see Figure 6 ).

[0176] The processing from step S11 to step S16 described above is a learning process of the maintenance display device 100 .

[0177] [Recognition Processing]

[0178] The identification process described below is a process for identifying whether an abnormality or a sign of an abnormality has occurred in the newly produced plurality of wound bodies 204 using the equipment state diagnosis model M generated by the learning process when the plurality of wound bodies 204 are newly produced in the production device 200.

[0179] In step S17 , the equipment state diagnosis unit 121 acquires shape data of a plurality of newly produced wound bodies (hereinafter referred to as new shape data).

[0180] In step S18, the equipment state diagnosis unit 121 calculates the consistency C using the new shape data and the equipment state diagnosis model M. The consistency C is a value indicating the degree of consistency between the new shape data and the past shape data included in the equipment state diagnosis model M. That is, the greater the consistency C, the higher the probability that an abnormality or a sign of an abnormality occurs in the production device 200, and the newly produced wound body 204 will become a defective product.

[0181] In step S19, when the degree of consistency C is equal to or greater than a given threshold, the notification determination unit 122 determines that notification is necessary for the user of the maintenance display device 100. When the degree of consistency C is equal to or greater than a given threshold, it means that an abnormality or a sign of an abnormality has occurred in the production device 200 and maintenance work needs to be re-performed.

[0182] In step S110, the notification determination unit 122 outputs the content of the maintenance work to be notified to the user to the notification unit 130. The content of the maintenance work to be notified to the user is determined based on the equipment state diagnosis model M for which the degree of consistency C is equal to or greater than a given threshold.

[0183] In steps S111 and S112, the notification unit 130 notifies the user that maintenance work should be performed. In step S111, the alarm unit 131 issues an alarm. In step S112, the display unit 132 displays the content of the maintenance work to be notified to the user. Fig. 9 Although an example is shown in which both the alarm in step S111 and the content display of the maintenance work in step S112 are performed, for example, the alarm may not be issued and only the content display of the maintenance work may be performed.

[0184] In this way, through the notification in steps S111 and S112, the operator who has received the notification performs the maintenance work on the production device 200 based on the content of the notified maintenance work.

[0185] The processes from step S17 to step S112 described above are recognition processes of the maintenance display device 100 using the learned model generated by the learning process.

[0186] exist Fig.10 , an overview of the update process in the maintenance display device 100 and the recognition process using the learned model updated by the update process is shown.

[0187] [Update Process]

[0188] The update process in the maintenance display device 100 is a process for updating the learned model (equipment state diagnosis model M) based on the maintenance results of the maintenance work when maintenance work is newly performed after the above-mentioned learning process. That is, the premise of the update process is that maintenance work is performed before the start of the update process.

[0189] In step S21, the maintenance effect determination unit 123 uses the actual production performance data PD (see Fig. 7A ) contains shape data (see Figure 5 ), and the device state diagnosis model M registered in the device state diagnosis model database 112, calculate the consistency C before the maintenance work before . Consistency before maintenance work C before This is an example of the first probability of the present invention.

[0190] In step S22, the maintenance effect determination unit 123 calculates the consistency C after the maintenance work by using the shape data included in the actual production results data of the plurality of winding bodies 204 produced after the maintenance work and the past shape data included in the equipment state diagnosis model M registered in the equipment state diagnosis model database 112. after . Consistency after maintenance work C after This is an example of the second probability of the present invention.

[0191] In step S23, the maintenance effect determination unit 123 determines the degree of consistency C before and after the maintenance work. before and C after The comparison is performed to determine whether the maintenance work is effective. Details of the maintenance work effect determination process performed by the maintenance effect determination unit 123 in the update process will be described later.

[0192] When it is determined in step S23 that the maintenance work is effective, the maintenance effect determination unit 123 outputs maintenance achievement data MD indicating the content of the maintenance work performed before the start of the update process to the equipment state diagnosis model generation unit 124 in step S24 .

[0193] In step S25, the facility state diagnosis model generation unit 124 updates the facility state diagnosis model M using the maintenance achievement data MD determined to be effective. Details of the update process of the facility state diagnosis model M will be described later.

[0194] In step S26, the device state diagnosis model generation unit 124 updates the device state diagnosis model database 112 (see Figure 6 ).

[0195] The processes from step S21 to step S26 described above are update processes of the maintenance display device 100 .

[0196] [Recognition Processing]

[0197] The identification process described below is a process for identifying whether an abnormality or a sign of an abnormality has occurred in the newly produced plurality of wound bodies 204 using the equipment state diagnosis model M updated by the update process when a plurality of wound bodies 204 are newly produced in the production device 200.

[0198] In step S27 , the equipment state diagnosis unit 121 acquires shape data of a plurality of newly produced wound bodies (hereinafter referred to as new shape data).

[0199] In step S28 , the device state diagnosis unit 121 calculates the degree of consistency C using the new shape data and the device state diagnosis model M. The degree of consistency C is a value indicating the degree to which the new shape data and the past shape data included in the device state diagnosis model M are consistent.

[0200] In step S29, when the degree of consistency C is equal to or greater than a given threshold, the notification determination unit 122 determines that it is necessary to notify the user of the maintenance display device 100. When the degree of consistency C is equal to or greater than a given threshold, it means that an abnormality or a sign of an abnormality has occurred in the production device 200 and maintenance work needs to be re-performed.

[0201] In step S210, the notification determination unit 122 outputs the content of the maintenance work to be notified to the user to the notification unit 130. The content of the maintenance work to be notified to the user is determined based on the equipment state diagnosis model M for which the degree of consistency C is equal to or greater than a given threshold.

[0202] In steps S211 and S212, the notification unit 130 notifies the user that maintenance work should be performed. In step S211, the alarm unit 131 issues an alarm. In step S212, the display unit 132 displays the content of the maintenance work to be notified to the user. Fig.10 Although an example is shown in which both the alarm in step S211 and the content of the maintenance work in step S212 are displayed, for example, the alarm may not be issued and only the content of the maintenance work may be displayed.

[0203] The operator who has received the notification in steps S211 and S212 performs the maintenance work on the production device 200 based on the content of the notified maintenance work.

[0204] The processing from step S27 to step S212 described above is the recognition processing of the maintenance display device 100. Fig.10The identification process from step S27 to step S212 shown is the same as Fig. 9 The identification processing from step S17 to step S112 shown is substantially the same processing.

[0205] <Details of each process>

[0206] The following, Fig. 9 as well as Fig.10 The learning process, recognition process, and update process shown are described in detail respectively.

[0207] [Learning Process]

[0208] First, the learning process executed by the maintenance effect determination unit 123 and the equipment state diagnosis model generation unit 124 will be described.

[0209] (Processing of the Maintenance Effect Determination Unit 123)

[0210] Next, the processing ( Fig. 9 The processing of steps S11 to S14 will be described below. Fig.11 This is a flowchart for explaining the process executed by the maintenance effect determination unit 123 in the learning process.

[0211] In step S31, the maintenance effect determination unit 123 reads out from the actual production result database 111 the actual production result data registered in the actual production result database 111, including the time when the maintenance work was performed before the learning process ( Figure 8 The production actual result data list PL includes all the production actual result data of the wound body 204 produced from the maintenance date and time shown to the previous given time. before The given time is a time of a predetermined length, which is the time required to produce a certain number of wound bodies 204. The production actual result data PL read out before The production performance data PD of each of the wound bodies 204 having a predetermined number or more of different production dates and times (reading dates and times) are included.

[0212] In step S32, the maintenance effect determination unit 123 determines the maintenance effect based on the production actual performance data list PL before The actual production performance data contained in the calculation is used to calculate the pre-maintenance defect rate Nf before As mentioned above, the defect rate before maintenance Nf before It is based on the production actual results data list PL before It is calculated by dividing the shape data of the actual production results data included in and the number of wound bodies 204 determined to be defective as a result of the inspection by the total number of productions before the maintenance work.

[0213] In step S33, the maintenance effect determination unit 123 reads out the data including the time when the maintenance was performed ( Figure 8 The production actual result data list PL includes all the production actual result data of the wound body 204 produced from the maintenance date and time shown in the figure to the given time thereafter. after The actual production data PL read out after In the production results data PL before Similarly, the production performance data PD of each of the wound bodies 204 having a predetermined number or more of different production dates and times (reading dates and times) is included.

[0214] In step S34, the maintenance effect determination unit 123 determines the maintenance effect based on the production actual performance data list PL after The actual production performance data contained in the maintenance defect rate Nf is calculated after As mentioned above, the defect rate after maintenance Nf after It is based on the production actual results data list PL after It is calculated by dividing the shape data of the actual production results data included in and the number of wound bodies 204 determined to be defective as a result of the inspection by the total number of production after the maintenance work.

[0215] In step S35, the maintenance effect determination unit 123 calculates the pre-maintenance failure rate Nf before and post-maintenance defect rate Nf after The difference (first difference) is determined to determine whether the difference is greater than a given threshold Th N When the difference is greater than the threshold Th N If (step S35: Yes), the maintenance effect determination unit 123 advances the process to step S36, and if not (step S35: No), the maintenance effect determination unit 123 advances the process to step S37.

[0216] In step S36, due to the defect rate Nf before maintenance before Compared with the defect rate after maintenance Nf after The maintenance effect determination unit 123 determines that the maintenance work is effective. The maintenance work in step S36 and step S37 described later refers to the maintenance work before the learning process, that is, before the learning process. Fig. 9 The maintenance work is performed in advance in step S11.

[0217] On the other hand, in step S37, due to the difference between the defect rate Nf before maintenance before Compared with the defect rate after maintenance Nf after Since the maintenance effect determination unit 123 does not decrease, it determines that the maintenance work has no effect or has a very small effect.

[0218] In this way, the maintenance effect determination unit 123 determines in the learning process whether or not the maintenance work performed before the learning process is effective.

[0219] Fig. 12A , Fig. 12B as well as Fig.13A , Fig. 13B This is a conceptual diagram for explaining how the effectiveness of maintenance work in the learning process is determined.

[0220] exist Fig. 12A as well as Fig. 12B , an example is shown in which a defect occurs in the wound body 204 due to the first laminating roller 205A being tilted due to deterioration over time. Fig. 12A as well as Fig. 12B The probability of producing defective wound bodies 204 increases with the passage of time.

[0221] On the other hand, Fig.13A as well as Fig. 13B In the above, the first laminating roller 205A is defective. Fig. 12A as well as Fig. 12B Similarly, the example in which the wound body 204 is defective due to dirt adhering to the first laminating roller 205A is shown, instead of deteriorating over time. Fig.13A as well as Fig. 13B In the embodiment, the probability of producing defective wound body 204 is not related to the passage of time.

[0222] Moreover, in Fig. 12A as well as Fig.13A , the shape data after the inclination adjustment operation of the first laminating roller 205A is performed as a maintenance operation (an operation to restore the rotation axis to a state of being inclined due to deterioration over time) is shown. Fig. 12B as well as Fig. 13B , the shape data after the cleaning operation (the operation of removing dirt attached to the first laminating roller 205A) of the first laminating roller 205A is performed as the maintenance operation is shown.

[0223] exist Fig. 12A In the figure, the shape data of the five winding bodies produced before maintenance (the five shape data on the left side of the figure) change with the passage of time, and only the last data is judged as defective. Fig. 12A In the example shown, the defect rate before maintenance Nf before In the following description, the given number or more (in Fig. 12A The shape data of each of the winding bodies 204 (5 in total) is recorded as a shape data group.

[0224] In addition, Fig. 12A In the shape data group of the five winding bodies produced after maintenance (the five shape data on the right side of the figure), there is almost no change with the passage of time, and the shape data judged as defective is 0 (defective rate after maintenance Nf after = 0). This is because the deterioration of the first laminating roller 205A over time, which is the cause of defective products in the production of the winding body 204, is eliminated by appropriate maintenance, so that the defect rate of the winding body 204 after maintenance is improved. Fig. 12A In the example shown, it can be said that the maintenance work was effective. Fig. 12A In the example shown, the defect rate before maintenance Nf before and post-maintenance defect rate Nf after The difference (the first difference of the present invention) is 20%.

[0225] On the other hand, Fig. 12B In the figure, the shape data sets of the five winding bodies produced before maintenance are compared with Fig. 12A Similarly, as time passes, only the last data is judged as defective. That is, the defect rate before maintenance Nf before is 20%.

[0226] In addition, Fig. 12B In the shape data set of the five wound bodies produced after maintenance, there was almost no change with the passage of time, and all the shape data were judged to be defective (defective rate after maintenance Nf after =100%). This is because, although the first laminating roller 205A has deteriorated over time, an inappropriate maintenance operation such as cleaning the first laminating roller 205A has been performed, and therefore all the wound bodies 204 produced thereafter are affected by the deterioration of the first laminating roller 205A over time and become defective. Fig. 12B In the example shown, the content of the maintenance work performed is inappropriate and it can be said that the maintenance work is ineffective. Fig. 12B In the example shown, the defect rate before maintenance Nf before and post-maintenance defect rate Nf after The difference (the first difference of the present invention) is -80%.

[0227] On the other hand, Fig.13A In the above, the cause of the failure was not the deterioration of the first laminating roller 205A over time, so the shape data set of the five winding bodies produced before the maintenance did not change with the passage of time, and sporadic failures occurred. Fig.13A In the example shown, only the last data is defective. Fig.13A In the example of before is 20%.

[0228] In addition, Fig.13A In the shape data set of the five winding bodies produced after maintenance, the number of defective products is 0. This is because the dirt on the first laminating roller 205A, which is the cause of the defective products of the winding body 204, is eliminated by appropriate maintenance, so that the defective rate of the winding body 204 after maintenance is improved. Fig.13A In the example shown, it can be said that the maintenance work was effective. Fig.13A In the example shown, the defect rate before maintenance Nf before and post-maintenance defect rate Nf after The difference (the first difference of the present invention) is 20%.

[0229] On the other hand, Fig. 13B In the shape data set of the five winding bodies produced after maintenance, sporadic defects occurred, as before maintenance. This is because, although defects occurred in the winding body 204 due to dirt on the first laminating roller 205A, an inappropriate maintenance operation such as adjusting the inclination of the first laminating roller 205A was performed, and the cause of the defect was not eliminated. That is, in Fig. 13B In the example shown, the content of the maintenance work performed is inappropriate and it can be said that the maintenance work is ineffective. Fig. 13B In the example shown, the defect rate before maintenance Nf before and post-maintenance defect rate Nf after The difference (the first difference of the present invention) becomes 0.

[0230] like Fig. 12A , Fig. 12B as well as Fig.13A , Fig. 13B As shown, the content of the maintenance work to be performed (effective in eliminating the defect) is different when the deterioration of the structure of the winding body 204 over time is the cause of the defect and when it is not. In the maintenance display device 100 according to the embodiment of the present invention, as shown in FIG. Fig. 7A As shown, the shape data of the winding body 204 is associated with the production time of the winding body 204 or the reading time (time information) of the shape data. In the maintenance display device 100 according to the embodiment of the present invention, by using the shape data associated with the time information, it is possible to perform each process taking into account the change of the shape data with the passage of time.

[0231] In addition, Fig. 12A , Fig. 12B as well as Fig.13A , Fig. 13B In the example shown, the difference in the defect rate of maintenance work that is determined to be effective is 20% ( Fig. 12A as well as Fig.13AFor example, the difference in the defect rate of maintenance work that was judged to be ineffective is -80% ( Fig. 12B example) or 0 ( Fig. 13B Thus, for example, by setting the threshold Th N By setting it to 10%, it is possible to determine whether the maintenance work is effective.

[0232] In addition, Fig. 12A , Fig. 12B as well as Fig.13A , Fig. 13B In the embodiment of the present invention, the reason for producing the defective winding body 204 is the first laminating roller 205A, but the present invention is not limited to this. As described above, in the present invention, the reason for producing the defective winding body 204 is Figure 3A At least one of the structures of the winding section 201 shown (the first supply reel 50, the second supply reel 51, the first laminating roller 205A, the second laminating roller 205B, the winding core 206, the inspection machine 207, the cutting section 209, the pressing section 210, the joint welding section 211, the tape laminating section 212, and the roller 213). Furthermore, when a structure other than the first laminating roller 205A becomes a cause of a defect, a signal indicating the defect is obtained. Fig. 12A , Fig. 12B as well as Fig.13A , Fig. 13B Shape data of a shape different from the shape illustrated in FIG.

[0233] (Processing of the Equipment State Diagnosis Model Generator 124)

[0234] Next, the processing performed by the device state diagnosis model generation unit 124 in the learning process ( Fig. 9 The processing of steps S15 and S16 will be described below. Fig.14 This is a flowchart for explaining the process executed by the device state diagnosis model generation unit 124 in the learning process.

[0235] In step S41 , the equipment state diagnosis model generation unit 124 reads out the maintenance actual result data MD of the maintenance work determined to be effective by the maintenance effect determination unit 123 .

[0236] In step S42, the equipment state diagnosis model generation unit 124 reads the pre-maintenance production performance data list PL from the production performance database 111. before Here, the pre-maintenance production actual performance data list PL read by the equipment state diagnosis model generation unit 124 is before The production actual result data list PL before maintenance read in the process of the maintenance effect determination unit 123 before Same (refer to Fig.11 Step S31).

[0237] In step S43, the equipment state diagnosis model generation unit 124 uses the read maintenance actual performance data MD and production actual performance data list PL before The actual production results data PD contained in the equipment status diagnosis model M is learned new .

[0238] In step S44, the device state diagnosis model generation unit 124 generates the newly generated device state diagnosis model M new Registered in the equipment status diagnosis model database 112.

[0239] In this way, in the learning process, a new equipment state diagnosis model M is generated by learning how the shape data changes with the passage of time and how the defect of the wound body 204 is improved by what kind of maintenance work. new , and registered in the equipment status diagnosis model database 112.

[0240] [Recognition Processing]

[0241] Next, the recognition process performed by the device state diagnosis unit 121 and the notification determination unit 122 will be described.

[0242] (Processing of the Device Status Diagnostic Unit 121)

[0243] Next, the processing performed by the device state diagnosis unit 121 in the identification processing ( Fig. 9 The processing of step S17 and step S18 will be described below. Fig.15 This is a flowchart for explaining the processing executed by the device state diagnosis unit 121 in the recognition processing.

[0244] In step S51, the equipment state diagnosis unit 121 determines whether new production performance data PD is registered in the production performance database 111. new . In the case of unregistered new production actual results data PD new (Step S51: No), the equipment state diagnosis unit 121 repeatedly executes Step S51. new (step S51 : Yes), the device state diagnosis unit 121 advances the process to step S52 .

[0245] In step S52, the equipment state diagnosis unit 121 performs the following operations based on the newly registered production actual performance data PD. new The actual production data list PL is extracted from the actual production data database 111. The actual production data list PL is a list of newly registered actual production data PD from among the actual production data PD registered in the actual production data database 111.new The actual production data PD of the winding body 204 produced from the production date to the previous given time is extracted and listed. That is, the actual production data list PL contains at least the newly registered actual production data PD new .

[0246] In step S53 , the equipment state diagnosis unit 121 generates a degree of consistency C using shape data corresponding to time information included in the actual production results data list PL and past shape data corresponding to time information included in the equipment state diagnosis model M read from the equipment state diagnosis model database 112 .

[0247] More specifically, the equipment state diagnosis unit 121 extracts a shape data group arranged in time series from a given number or more of the actual production result data included in the actual production result data list PL (see Figure 5 ). On the other hand, the device state diagnosis unit 121 extracts a plurality of device state diagnosis models M registered in the device state diagnosis model database 112. The plurality of device state diagnosis models M correspond to different maintenance tasks, respectively.

[0248] The equipment state diagnosis unit 121 calculates a plurality of degrees of consistency C for all combinations of shape data sets of a predetermined number or more of the wound bodies 204 extracted from the actual production data and shape data sets included in the plurality of equipment state diagnosis models M.

[0249] (Notification of Processing of the Determination Unit 122)

[0250] Next, the processing performed by the notification determination unit 122 in the recognition processing ( Fig. 9 The processing from step S19 to step S112 is described below. Fig.16 This is a flowchart for explaining the process executed by the notification determination unit 122 in the recognition process.

[0251] In step S61, the notification determination unit 122 aggregates the consistency C for each maintenance group based on the plurality of consistency C generated by the device state diagnosis unit 121. The so-called maintenance group is a group corresponding to the content of the maintenance work. Fig.17A The maintenance group information shown is used to group the maintenance work to be performed. Fig.17A As shown in FIG. 1 , the maintenance group information is information that establishes a correspondence between the maintenance group and the maintenance work performed by the maintenance group. Fig.17AAs shown, the maintenance group information may further include the maintenance plan to be implemented corresponding to the maintenance group. In addition, in this embodiment, the group divided according to each component of the maintenance operation object is described as the maintenance group, but the present invention is not limited to this. The maintenance group may also be a group divided according to each content of the maintenance operation, or a group divided according to each model of the parts replaced in the maintenance operation.

[0252] In the following description, the result of summing up the consistency C for each maintenance team is set as the total value A. The method of generating the total value A can be appropriately determined from a plurality of types of summing methods. As specific examples of the plurality of types of summing methods, for example, there can be listed a method of summing up the consistency C, a method of averaging the consistency C, a method of selecting the maximum value from the consistency C, a method of extracting a given number of consistency Cs from a superior and averaging them, and the like.

[0253] In step S62, the notification determination unit 122 generates a maintenance plan list ML. The maintenance plan list ML is a list of maintenance groups, and the maintenance groups are arranged in descending order of the total value A, for example. Fig. 17B It is a diagram showing a specific example of the maintenance plan list ML.

[0254] like Fig. 17B As shown, the maintenance plan list ML includes data such as "maintenance plan ID", "equipment", "maintenance plan" and "total value". The "maintenance plan ID" data is an identifier assigned to each maintenance team that is rearranged according to the size of the total value. As the "maintenance plan ID" data, for example, the larger the total value, the smaller the number assigned. The "maintenance plan" data is data indicating the content of the maintenance to be performed corresponding to the maintenance team. The notification determination unit 122 refers to Fig.17A The maintenance plan corresponding to the maintenance group is determined by the maintenance group information shown. The "total value" data is data indicating the value of the total value A totaled for each maintenance group.

[0255] exist Fig. 17B In the embodiment, as an example of a maintenance plan, a case of maintaining the first laminating roller 205A, the second laminating roller 205B, or the first supply reel 50 is illustrated. However, the present invention is not limited to this. Figure 3A Any of the structures of the winding portion 201 shown can be the subject of maintenance.

[0256] also, Fig. 17BThe "maintenance of the first laminating roller" in the maintenance scheme shown means at least one of the maintenance operations on the first laminating roller 205A, such as adjusting the inclination of the first laminating roller 205A, cleaning the first laminating roller 205A, and replacing the first laminating roller 205A. The same is true for the "maintenance of the second laminating roller" and the "maintenance of the first supply reel".

[0257] The total value A is a value obtained by summing up the consistency C, and therefore has the same properties as the consistency C. Therefore, the larger the total value A is, the more necessary it is for the maintenance team to perform the maintenance content on the target production device 200. In addition, the maintenance plan list ML is a list of maintenance teams arranged in descending order of the total value A, and therefore the higher the maintenance team in the maintenance plan list ML is, the more necessary it is for the maintenance of the target production device 200.

[0258] In step S63, the notification determination unit 122 determines whether the total value A is greater than a given sign threshold value Th for each maintenance team. f The so-called given warning threshold Th f , is the minimum value of the total value of the sign of abnormality assumed to have occurred in the production device 200. In addition, in the present embodiment, the so-called abnormality of the production device 200 means, for example, that the production device 200 produces more than a given proportion of winding bodies 204 with an inspection result of "bad". In addition, the so-called sign of abnormality of the production device 200 means, for example, that the production device 200 produces more than a given proportion of winding bodies 204 with an inspection result of "pass". The given sign threshold Th f For example, it may be determined empirically based on past maintenance achievement data MD or the like.

[0259] Even if the maintenance plan list ML contains only one total value A greater than the warning threshold Th f When the maintenance team is notified (step S63: Yes), the notification determination unit 122 also advances the process to step S64. f When none of the maintenance groups is included in the maintenance plan list ML (step S63 : No), the notification determination unit 122 determines that notification is unnecessary and ends the process.

[0260] In step S64, the notification determination unit 122 determines whether there is a maintenance group included in the maintenance plan list ML whose total value A is greater than a given abnormality threshold value Th. a The so-called given abnormal threshold Th a , is the minimum value of the total value assuming that an abnormality has occurred in the production device 200 beyond the warning stage. aFor example, it is determined empirically based on past maintenance actual performance data MD, etc. to be greater than the warning threshold value Th f When the total value A is greater than the abnormal threshold value Th a If the maintenance group is included in the maintenance plan list ML (step S64: Yes), the notification determination unit 122 advances the process to step S66. a When the maintenance team is not included in the maintenance plan list ML (step S64 : No), the notification determination unit 122 advances the process to step S65 .

[0261] In step S65, the notification determination unit 122 notifies the display unit 132 of the notification unit 130 that the total value A is greater than the warning threshold value Th in step S63. f More specifically, the notification determination unit 122 causes the display unit 132 to display not only a message such as "Please perform the following maintenance content." but also the content of the recommended maintenance work. In addition, the content of the recommended maintenance work is Fig. 17B The maintenance plan list ML shown corresponds to the content of the "maintenance plan" data.

[0262] Here, the notification determination unit 122 notifies the user that when there are a plurality of total values ​​A that are greater than the sign threshold value Th f In the case of a maintenance group, the contents of multiple maintenance operations can also be sorted and displayed by the total value. In this case, in more detail, the notification determination unit 122 displays a message such as "Please perform the following maintenance contents. When the condition does not improve even if the upper-level maintenance contents are performed, sometimes the condition can be improved by simply performing the lower-level maintenance contents." and displays the contents of the recommended maintenance operations in order from upper to lower levels.

[0263] Furthermore, the notification determination unit 122 notifies not only the maintenance work content but also the maintenance plan ID corresponding to the maintenance content. When the operator who performs the maintenance work inputs the maintenance actual result data MD, by inputting the maintenance actual result data MD and the maintenance plan ID that became the opportunity for maintenance, it is possible to easily determine whether the input maintenance actual result data MD is data corresponding to the maintenance work performed based on the notification of the maintenance display device 100.

[0264] In step S66, the notification determination unit 122 causes the display unit 132 to display the content of the maintenance work, similarly to step S65, and causes the alarm unit 131 to issue an alarm for notifying that an abnormality has occurred in the target production device 200. When an abnormality has occurred in the target production device 200 but it is not a sign of abnormality, it is a situation that requires urgent maintenance work. Therefore, the notification determination unit 122 not only displays the content of the maintenance work on the display unit 132, but also issues an alarm through the alarm unit 131 to quickly notify the user of the maintenance display device 100 of the occurrence of the abnormality.

[0265] In this way, in the recognition process, the production actual result data PD (especially the shape data and time information) of the newly produced winding body 204 and the equipment status diagnosis model M are used to determine whether an abnormality (a situation where a given ratio of defective products or more is produced) or a sign of an abnormality has occurred in the production device 200. Furthermore, if it is determined that an abnormality or a sign of an abnormality has occurred, the user is notified. In this way, the user can quickly learn about the abnormality when an abnormality has occurred in the production device 200, and can learn the content of the maintenance work to be performed to improve the abnormality.

[0266] [Update Process]

[0267] Next, the updating process performed by the maintenance effect determination unit 123 and the equipment state diagnosis model generation unit 124 will be described.

[0268] (Processing of the Maintenance Effect Determination Unit 123)

[0269] Next, the processing performed by the maintenance effect determination unit 123 in the update processing ( Fig.10 The processing of steps S21 to S24 is described below. Fig.18 This is a flowchart for explaining the process executed by the maintenance effect determination unit 123 in the update process.

[0270] In step S71, the maintenance effect determination unit 123 determines whether new maintenance result data MD is registered in the maintenance result database 113 of the storage unit 110. new If it is determined that no new maintenance actual performance data MD has been registered new If the maintenance result determination unit 123 determines that the new maintenance result data MD is registered, the maintenance result determination unit 123 executes step S71 repeatedly. new (step S71 : Yes), the maintenance effect determination unit 123 advances the process to step S72 .

[0271] In step S72, the maintenance effect determination unit 123 determines the maintenance result based on the newly registered maintenance actual result data MD. newThe "maintenance date and time" data contained in Figure 8 ), determine the maintenance results data MD from the new registration new Whether a given time has passed since the corresponding maintenance.

[0272] If it is determined that the given time has passed since the execution time of the maintenance work (step S72: Yes), the maintenance effect determination unit 123 advances the process to step S73. If it is determined that the given time has not passed since the execution time of the maintenance work (step S72: No), the maintenance effect determination unit 123 repeatedly executes the process of step S72.

[0273] In step S73, the maintenance effect determination unit 123 reads out the pre-maintenance production result data list PL including all the production result data PD of the wound body 204 produced from the maintenance work to the previous predetermined time from the production result database 111. before .

[0274] In step S74, the maintenance effect determination unit 123 reads out the new maintenance actual achievement data MD from the equipment state diagnosis model database 112. new The maintenance team of the corresponding maintenance content has a device status diagnosis model M, and based on the read device status diagnosis model M and the production actual performance data list PL before To generate the consistency before maintenance C before About the consistency before maintenance C before The method of generating is Fig.15 The method of generating the degree of consistency C executed by the device state diagnosis unit 121 in step S53 is the same.

[0275] In step S75, the maintenance effect determination unit 123 reads out from the production record database 111 a production record data list PL including all the production record data PD of the wound body 204 produced from the maintenance work to a predetermined time thereafter. after .

[0276] In step S76, the maintenance effect determination unit 123 reads out the new maintenance actual achievement data MD from the equipment state diagnosis model database 112. new The maintenance team of the corresponding maintenance content has a device status diagnosis model M, and based on the read device status diagnosis model M and the production actual performance data list PL after To generate the maintained consistency C after About consistency C after The method of generating is Fig.15 The method of generating the degree of consistency C executed by the device state diagnosis unit 121 in step S53 is the same.

[0277] In step S77, the maintenance effect determination unit 123 takes the consistency C before maintenance as before Consistency with maintenance C after The difference (an example of the second difference of the present invention) is determined to determine whether the difference is greater than a given threshold value Th D When the difference is greater than the threshold Th D If (step S77: Yes), the maintenance effect determination unit 123 advances the process to step S78, otherwise (step S77: No), the process advances to step S79. D It may be appropriately determined based on the actual results of past maintenance work, etc.

[0278] In step S78, due to the consistency C before maintenance before Compared with the consistency after maintenance C after Since the maintenance effect determination unit 123 determines that the maintenance work performed based on the maintenance content notified by the notification determination unit 122 is effective, the maintenance effect determination unit 123 determines that the maintenance work performed based on the maintenance content notified by the notification determination unit 122 is effective.

[0279] In step S79, due to the consistency C before maintenance before Compared with the consistency after maintenance C after Since the maintenance effect determination unit 123 determines that the maintenance work performed based on the maintenance content notified by the notification determination unit 122 is ineffective or has a very small effect, the maintenance effect determination unit 123 determines that the maintenance work performed based on the maintenance content notified by the notification determination unit 122 is ineffective or has a very small effect.

[0280] Fig.19A as well as Fig.19B 1 is a conceptual diagram for explaining how to determine the effect of maintenance work in the update process. Fig.19A An example is shown in which it is determined that the maintenance work is effective. Fig.19B An example is shown in which it is determined that the maintenance work is ineffective. Fig.19A The shape data set and maintenance operation examples shown are similar to Fig. 12A In addition, Fig.19B The shape data set and maintenance operation examples shown are similar to Fig. 12B Same as shown.

[0281] exist Fig.19A as well as Fig.19B In the example shown, the pre-maintenance consistency C is calculated using the shape data set of the five wound bodies 204 produced before maintenance and the equipment status diagnosis model M. before =0.90.

[0282] Moreover, in Fig.19AIn the example shown, the post-maintenance consistency C is calculated using the shape data set of the five wound bodies 204 produced after maintenance and the equipment status diagnosis model M. after =0.20. On the other hand, Fig.19B In the example shown, the post-maintenance consistency C is calculated using the shape data set of the five wound bodies 204 produced after maintenance and the equipment status diagnosis model M. after =0.90.

[0283] Therefore, in Fig.19A In the example shown, the consistency before maintenance C before Consistency with maintenance C after The difference is 0.70. On the other hand, Fig.19B In the example shown, the consistency before maintenance C before Consistency with maintenance C after The difference becomes 0. Therefore, for example, the threshold value Th for determining the presence or absence of the maintenance effect D When the value is 0.30, Fig.19A In the example shown, the maintenance work is judged to be effective. Fig.19B In the example shown, it is determined that the maintenance work is ineffective.

[0284] (Processing of the Equipment State Diagnosis Model Generator 124)

[0285] Next, the processing performed by the device state diagnosis model generation unit 124 in the update processing ( Fig.10 The processing of steps S25 and S26 is described below. Fig. 20 This is a flowchart for explaining the process executed by the device state diagnosis model generation unit 124 in the update process.

[0286] In step S81, the equipment state diagnosis model generation unit 124 reads the maintenance actual result data MD of the maintenance work determined to be effective by the maintenance effect determination unit 123. new .

[0287] In step S82, the equipment state diagnosis model generation unit 124 reads the pre-maintenance production performance data list PL from the production performance database 111. before Here, the pre-maintenance production actual performance data list PL read by the equipment state diagnosis model generation unit 124 is before The production actual result data list PL before maintenance read in the process of the maintenance effect determination unit 123 before Same (refer to Fig.11 Step S31).

[0288] In step S83, the equipment state diagnosis model generation unit 124 uses the read maintenance actual performance data MD and production actual performance data list PL before The actual production results data PD contained in the model is used to learn and generate a new equipment status diagnosis model M new .

[0289] In step S84, the device state diagnosis model generation unit 124 generates a new device state diagnosis model M as a result of the learning. new The device state diagnosis model M is added to the device state diagnosis model database 112 and updated.

[0290] Thus, in the updating process, the device state diagnosis model M generated in the learning process is used to generate a new device state diagnosis model M. new , and use the new equipment status diagnosis model M new The device state diagnosis model M registered in the device state diagnosis model database 112 is updated. In this way, the new device state diagnosis model M is used based on effective maintenance work. new The equipment state diagnosis model M of the equipment state diagnosis model database 112 is updated, so that the diagnosis accuracy of the equipment state of the production device 200 in the equipment state diagnosis unit 121 is gradually improved.

[0291] The present disclosure relates to a method for displaying information used for maintenance of a production device, based on first data indicating the position of a first end face corresponding to time information indicating the moment when a given number or more of winding bodies are respectively read, and second data indicating the position of a second end face corresponding to time information indicating the moment when a given number or more of winding bodies are respectively read. When it is judged that the continuity of the position of the first end face indicated by the first data intersects with the continuity of the position of the second end face indicated by the second data, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the core or the sensor is output to a display device.

[0292] In addition, the method for displaying information for maintenance of a production device involved in the present disclosure inputs, in a learning model created by the method for generating a learning model involved in the present disclosure, first data indicating the position of a first end face corresponding to time information indicating the moment when a given number of winding bodies or more are respectively read, and second data indicating the position of a second end face corresponding to time information indicating the moment when a given number of winding bodies or more are respectively read, which are obtained from a sensor. Furthermore, when information indicating that the position of the first end face indicated by the first data is judged to be continuous and the position of the second end face indicated by the second data is continuously crossed is output from the learning model, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core, or the sensor is output to the display device.

[0293] <Functions and Effects of the Maintenance Display Device 100 of the First Embodiment>

[0294] As described above, the maintenance display device 100 includes the notification determination unit 122, the maintenance effect determination unit 123, and the equipment state diagnosis model generation unit 124 as an example of a model generation unit. The notification determination unit 122 acquires the first data indicating the position of the first end face and the second data indicating the position of the second end face read along the radial direction of the winding body 204 from the inspection machine 207 as a sensor. Furthermore, the notification determination unit 122 determines whether the winding body 204 is defective based on whether the continuity of the position of the first end face indicated by the first data intersects the continuity of the position of the second end face indicated by the second data. When the winding body 204 is defective, the notification determination unit 122 outputs information indicating that the cause of the defect is at least one of the first supply reel 50, the second supply reel 51, the first laminating roller 205A, the second laminating roller 205B, the winding core 206, the inspection machine 207, the cutting unit 209, the pressing unit 210, the joint welding unit 211, the tape pasting unit 212, and the roller 213 to the display unit 132 for maintenance. The maintenance effect determination unit 123 extracts the first data and the second data of each of the winding bodies of a given number or more generated before maintenance and read at different times, and calculates the first defect rate Nf of the winding bodies of a given number or more read before maintenance based on the extracted first data and second data. before , and extracting the first data and the second data of each of the winding bodies of a given number or more generated and read at different times after maintenance, and calculating the second defect rate Nf of the winding bodies of a given number or more read after maintenance based on the extracted first data and the second data after The equipment state diagnosis model generation unit 124 calculates the first defect rate Nf before and the second defect rate Nf afterFurthermore, when the device state diagnosis model generation unit 124 determines that the first difference is less than a given value, the first data and the second data of each of the given number of winding bodies 204 read before the first supply reel 50 is maintained are not used to generate or update the learned model. On the other hand, when it is determined that the first difference is greater than a given value, the learned model (device state diagnosis model M) is generated or updated using the first data and the second data of each of the given number of winding bodies 204 read before the first supply reel 50 is maintained.

[0295] On the other hand, the equipment state diagnosis model generation unit 124 calculates a first probability (degree of consistency) C that the defect of the wound body 204 is improved by inputting the first data and the second data of each of the predetermined number or more of wound bodies 204 before the first bonding roller 205A is maintained into the learned model. before Furthermore, the equipment state diagnosis model generation unit 124 calculates a second probability (degree of consistency) C that the defect of the wound body 204 is improved by inputting the first data and the second data after the maintenance of the first bonding roller 205A into the learned model. after . Then, the equipment status diagnosis model generation unit 124 calculates the second difference between the first probability and the second probability. When it is judged that the second difference is less than a given value, the first data and the second data of each of the given number of winding bodies 204 read before the first bonding roller 205A is maintained are not used to create or update the learned model. On the other hand, when it is judged that the second difference is greater than a given value, the first data and the second data of each of the given number of winding bodies 204 read before the first bonding roller 205A is maintained are used to create or update the learned model.

[0296] As described above, the maintenance display device 100 involved in the first embodiment can perform: learning processing, generating an equipment status diagnosis model M for diagnosing the equipment status of the production device 200 through learning; identification processing, using the equipment status diagnosis model M to identify whether an abnormality or a sign of an abnormality has occurred in the production device 200, and notifying the situation when an abnormality or a sign of an abnormality has occurred; and update processing, updating the equipment status diagnosis model M based on actual maintenance results data MD corresponding to the maintenance work performed based on the notification.

[0297] In more detail, in the learning process, the maintenance display device 100 determines whether the maintenance work is effective based on the newly registered maintenance actual results data MD and the production actual results data PD of the winding body 204 produced before and after the maintenance work and establishes a corresponding shape data group with the time information (the production time of the winding body 204 or the reading time of the shape data), and generates the equipment status diagnosis model M using the maintenance actual results data MD and the shape data group corresponding to the maintenance work determined to be effective.

[0298] In addition, during the identification process, the maintenance display device 100 calculates the consistency C between the shape data group of the winding body 204 produced after the maintenance work and the equipment status diagnosis model M for each maintenance team, and based on the size of the consistency C, determines whether to issue an alarm and notify the content of the maintenance work, or only notify the content of the maintenance work, or not to make the notification itself.

[0299] Furthermore, during the update process, the maintenance display device 100 determines whether the maintenance operation is effective based on the newly registered maintenance actual result data MD and the shape data group included in the production actual result data PD of the winding body 204 produced before and after the maintenance operation, and generates a new equipment status diagnosis model M using the maintenance actual result data MD and the shape data group corresponding to the maintenance operation determined to be effective. new , and use the new equipment status diagnosis model M new To update the equipment status diagnosis model M.

[0300] According to such a structure, the state of the production device 200 can be appropriately diagnosed by using the learned model (equipment state diagnosis model M) generated based on effective maintenance operations (with reduced defective rate) among the maintenance operations actually performed. In addition, since the learned model is updated at any time, the accuracy of the diagnosis can be improved. Furthermore, in the case where it is diagnosed that an abnormality has occurred in the production device 200, an alarm can be issued to allow the user to take emergency measures, and in the case where it is diagnosed that a sign of abnormality has occurred, the content of the maintenance operation that can be foreseen to be improved is notified to the user, so that the maintenance operation can be performed when the defective product generation rate in the production device 200 is low.

[0301] Furthermore, according to the maintenance display device 100 according to the first embodiment, the equipment state diagnosis model M is created using the shape data group in which the shape data of each of the given number or more of the winding bodies 204 are arranged in time series, and it is determined whether the maintenance is effective. Therefore, for example, even if the winding body 204 that is defective is produced due to the deterioration of each structure of the winding unit 201 over time, since the processing is performed based on the result of learning using many shape data groups, it is possible to accurately determine which structure of the winding unit 201 is the cause, and whether the cause is caused by the deterioration of the structure over time or is temporary.

[0302] The maintenance display device involved in this embodiment has a notification unit, a maintenance effect determination unit, and an equipment status diagnosis model generation unit. The notification unit notifies the content of the maintenance operation based on the equipment status diagnosis model that establishes an association between the content of the maintenance operation and the actual production result data before the maintenance operation and is registered in the database at each maintenance operation performed in the past, and the newly input actual production result data. The maintenance effect determination unit determines whether the maintenance operation is effective based on the actual production result data before the time point when the maintenance operation is performed and the actual production result data after the time point when the maintenance operation is performed. The equipment status diagnosis model generation unit generates a new equipment status diagnosis model based on the actual production result data before the time point when the maintenance operation is determined to be effective and the content of the maintenance operation determined to be effective.

[0303] The maintenance display device involved in this embodiment also has: an equipment status diagnosis unit that generates an equipment status diagnosis index, which is the degree of consistency between the newly registered actual production results data and the actual production results data before the maintenance operation included in the equipment status diagnosis model. In addition, the notification unit notifies the content of the maintenance operation based on the equipment status diagnosis index.

[0304] In the maintenance display device involved in this embodiment, the equipment status diagnosis model generation unit uses actual production results data before the time point of the maintenance work determined to be effective, and actual maintenance results data related to the maintenance work, to generate an equipment status diagnosis model through machine learning.

[0305] The maintenance display device according to the present embodiment performs maintenance work not based on the content of the maintenance work notified by the notification unit, and when maintenance actual result data related to the maintenance work is newly input, the defect rate of the production actual result data with inspection results being defective during the period from the time point when the maintenance work of the newly input maintenance actual result data is performed to the given time before is calculated based on the data related to the inspection results of the products of the production equipment contained in the production actual result data. Furthermore, the defect rate of the production actual result data with inspection results being defective during the period from the time point when the maintenance work of the newly input maintenance actual result data is performed to the given time after is calculated. Then, the maintenance effect determination unit calculates the difference between the defect rate before the maintenance work and the defect rate after the maintenance work, and determines whether the maintenance work is effective based on the size of the difference.

[0306] (Second embodiment)

[0307] Hereinafter, a second embodiment of the present invention will be described. Fig.21 The diagram illustrates a configuration of a maintenance display device 100A according to the second embodiment. In the maintenance display device 100A according to the second embodiment, the processing performed by a maintenance effect determination unit 123A included in a control unit 120A of a server 10A is different from that of the maintenance effect determination unit 123 according to the first embodiment described above.

[0308] The following describes the differences from the first embodiment. The same configurations as those in the first embodiment are described by assigning the same reference numerals as those in the first embodiment, and configurations different from those in the first embodiment are described by assigning reference numerals "A".

[0309] In the first embodiment, it is not assumed that the user of the maintenance display device 100 performs maintenance work other than the content notified by the maintenance display device 100. However, in actual operation of the production device 200, appropriate and necessary maintenance work (maintenance work other than the maintenance content notified by the maintenance display device 100) may be performed at any time by on-site judgment or the like. In the second embodiment, a maintenance display device 100A that can also cope with the case where maintenance work other than the maintenance content notified by the maintenance display device 100A is performed will be described.

[0310] Fig. 22 This is a flowchart for explaining the processing executed by the maintenance effect determination unit 123A in the second embodiment.

[0311] exist Fig. 22 In step S91, the maintenance effect determination unit 123A determines whether maintenance achievement data MD is newly registered in the maintenance achievement database 113 of the storage unit 110. newIf it is determined that no new maintenance actual performance data MD has been registered new If the maintenance result determination unit 123A determines that the new maintenance result data MD is registered, the maintenance result determination unit 123A executes step S91 repeatedly. new (step S91: Yes), the maintenance effect determination unit 123A advances the process to step S92.

[0312] In step S92, the maintenance effect determination unit 123A determines the maintenance result based on the newly registered maintenance actual result data MD. new The maintenance date and time data included in the maintenance result data MD newly registered is determined new Whether a given time has passed since the corresponding maintenance work. The given time is similar to the given time described in the first embodiment, and is, for example, the time required to produce a certain number of wound bodies 204 in the target production device 200 after the maintenance work is performed.

[0313] If it is determined that a given time has passed since the maintenance work (step S92: Yes), the maintenance effect determination unit 123A advances the process to step S93. If it is determined that a given time has not passed since the maintenance work (step S92: No), the maintenance effect determination unit 123A repeatedly executes the process of step S92.

[0314] In step S93, the maintenance effect determination unit 123A determines whether there is any maintenance achievement data MD corresponding to the newly registered maintenance achievement data MD. new A corresponding maintenance plan ID is established. As described in the first embodiment, the notification determination unit 122 notifies not only the maintenance work content but also the maintenance team of the maintenance content and establishes a corresponding maintenance plan ID. The operator performs the maintenance work indicated by the notified maintenance plan ID. The operator establishes a correspondence between the performed maintenance work and the notified maintenance plan ID and inputs the maintenance actual result data MD. Thus, the maintenance actual result data MD and the maintenance plan ID that became the opportunity for maintenance are established in correspondence. In this step S93, the newly registered maintenance actual result data MD is determined in this way. new Whether maintenance is performed in response to a notification from the maintenance display device 100A.

[0315] In step S93, there is maintenance actual performance data MD newly registered new If the corresponding maintenance plan ID is established, it is judged to be consistent with the maintenance actual performance data MD new The corresponding maintenance work is triggered by the notification of the maintenance content by the maintenance display device 100A. newIf the corresponding maintenance plan ID is established, it is judged to be consistent with the maintenance actual performance data MD new The corresponding maintenance work is not triggered by the notification of the maintenance content by the maintenance display device 100A.

[0316] In step S93, it is determined that the maintenance performance data MD is newly registered. new If the maintenance plan ID is included in the maintenance result data MD (step S93: Yes), the maintenance effect determination unit 123A advances the process to step S94. On the other hand, it is determined that the maintenance result data MD new When the maintenance plan ID is not included in (step S93 : No), the maintenance effect determination unit 123A advances the process to step S95 .

[0317] Step S94 is to newly register the maintenance actual performance data MD new The corresponding maintenance operation is processed when the notification of the maintenance content based on the maintenance display device 100A is used as an opportunity. Therefore, in step S94, the maintenance effect determination unit 123A transfers to the process of determining whether the maintenance operation triggered by the notification of the maintenance content based on the maintenance display device 100A is effective. In addition, the maintenance effect determination process of the maintenance triggered by the notification of the maintenance content based on the maintenance display device 100A is different from the maintenance effect determination process of the maintenance performed with reference to the first embodiment described above. Fig.18 The processing described is substantially the same, so the description is omitted.

[0318] On the other hand, step S95 is to maintain the actual performance data MD new The corresponding maintenance operation is not triggered by the notification of the maintenance content based on the maintenance display device 100A. Therefore, the maintenance effect determination unit 123A shifts to the process of determining whether the maintenance operation not triggered by the maintenance display device 100A is effective. In addition, the maintenance effect determination process for the maintenance not triggered by the notification of the maintenance content based on the maintenance display device 100A is different from the maintenance effect determination process described in the first embodiment. Fig.11 The processing described is substantially the same, so the description is omitted.

[0319] As described above, according to the maintenance display device 100A according to the second embodiment, even when a maintenance operation is performed that is not triggered by the notification of the maintenance content by the maintenance display device 100A, the maintenance result data MD can be appropriately registered. new In addition, using Fig. 22 The processing of the maintenance effect determination unit 123A described above can be executed in either the learning processing or the updating processing described above.

[0320] The maintenance display device involved in this embodiment generates a device state diagnosis index before the maintenance operation based on the production actual result data from the time point of the maintenance operation of the newly registered maintenance actual result data to the previous given time and the equipment state diagnosis model associated with the content of the maintenance operation in the notification that became the opportunity for the maintenance operation of the newly input maintenance actual result data. In addition, the device state diagnosis index after the maintenance operation is generated based on the production actual result data from the time point of the maintenance operation of the newly input maintenance actual result data to the next given time and the equipment state diagnosis model associated with the content of the maintenance operation in the notification that became the opportunity for the maintenance operation of the newly input maintenance actual result data. Then, the maintenance effect determination unit calculates the difference between the device state diagnosis index before the maintenance operation and the device state diagnosis index after the maintenance operation, and determines whether the maintenance operation is effective based on the size of the difference.

[0321] (Third embodiment)

[0322] Hereinafter, a third embodiment of the present invention will be described. Fig.23 1 is a diagram illustrating the structure of a maintenance display device 100B according to the third embodiment. The maintenance display device 100B according to the third embodiment is different from the maintenance display device 100 according to the first embodiment described above in that the storage unit 110B of the server 10B further includes an ineffective device state diagnosis model database 114, and the control unit 120B includes a notification determination unit 122B, a maintenance effect determination unit 123B, and a device state diagnosis model generation unit 124B.

[0323] In the first embodiment described above, the equipment state diagnosis model generation unit 124 generates a new equipment state diagnosis model M using the maintenance actual result data MD determined to be effective. new (Reference Fig.14 In the third embodiment, the equipment state diagnosis model generation unit 124B further generates a new equipment state diagnosis model M using the maintenance actual result data MD determined to be maintenance ineffective. new .

[0324] Fig.24 This is a flowchart for explaining the processing performed by the device state diagnosis model generation unit 124B in the third embodiment. Fig.24 The described processing can be executed in either the learning processing or the updating processing.

[0325] In step S101, the equipment state diagnosis model generation unit 124B reads out newly registered maintenance achievement data MD from the maintenance achievement database 113. newHere, the equipment state diagnosis model generation unit 124B reads the maintenance actual achievement data MD regardless of the determination result of the effect by the maintenance effect determination unit 123B. new .

[0326] In step S102, the equipment state diagnosis model generation unit 124B reads the production actual result data list PL before the maintenance work from the production actual result database 111. before .

[0327] In step S103, the equipment state diagnosis model generation unit 124B uses the read maintenance actual performance data MD new , and production actual results data list PL before The actual production results data PD contained in the equipment status diagnosis model M is learned new .

[0328] In step S104, the device state diagnosis model generation unit 124B generates the newly generated device state diagnosis model M new The model generated based on the maintenance achievement data MD determined to be ineffective is registered in the ineffective equipment state diagnosis model database 114. On the other hand, the equipment state diagnosis model generation unit 124B registers the newly generated equipment state diagnosis model M new Among them, the model generated based on the maintenance actual achievement data MD determined to be effective is registered in the equipment state diagnosis model database 112 .

[0329] In this way, the equipment state diagnosis model generation unit 124B generates an equipment state diagnosis model M using not only the maintenance achievement data MD of maintenance determined to be effective, but also the maintenance achievement data MD of maintenance determined to be ineffective.

[0330] The device state diagnosis model M generated in this way is used to perform the recognition process by the device state diagnosis unit 121 and the notification determination unit 122B. The process performed by the device state diagnosis unit 121 is similar to that in the first embodiment described above. Fig.15 The processing described is substantially the same, so the description is omitted.

[0331] Hereinafter, the processing executed by the notification determination unit 122B in the recognition processing of the third embodiment will be described. Fig.25 This is a flowchart for explaining the processing executed by the notification determination unit 122B in the third embodiment.

[0332] In step S111, the notification determination unit 122B uses the consistency C generated by the device state diagnosis unit 121 to total the consistency C for each maintenance team and generate a total value A. In the third embodiment, information (flag) indicating whether the maintenance work is determined to be effective is associated with each maintenance team by the maintenance effect determination unit 123B.

[0333] In step S112 , the notification determination unit 122B generates a maintenance plan list ML which is a list of maintenance teams arranged in descending order of the total value A.

[0334] In step S113, the notification determination unit 122B determines whether the maintenance operation is determined to be effective for each maintenance group included in the maintenance plan list ML. As described above, in the third embodiment, the device state diagnosis unit 121 associates each maintenance group with a mark indicating whether the maintenance operation is effective, so the notification determination unit 122B refers to the mark to perform the processing of this step S113. The notification determination unit 122B advances the processing to step S114 for the maintenance group of the maintenance operation determined to be effective. On the other hand, the notification determination unit 122B advances the processing to step S117 for the maintenance group of the maintenance operation determined to be ineffective.

[0335] In step S114, the notification determination unit 122B determines whether the total value A is greater than a given sign threshold value Th for each maintenance team determined to be effective. f Even if there is only one total value A greater than the warning threshold Th f In the case of a maintenance team (step S114: Yes), the notification determination unit 122B also advances the process to step S115. f In the case of a maintenance team (step S114: No), the notification determination unit 122B ends the processing.

[0336] In step S115, the notification determination unit 122B determines whether there is a maintenance group whose total value A is greater than a given abnormality threshold value Th among the maintenance groups determined to be effective. a Maintenance team. When the total value A is greater than the abnormal threshold Th a If there is no maintenance team (step S115: Yes), the notification determination unit 122B advances the process to step S116. a In the case of a maintenance team (step S115: No), the notification determination unit 122B advances the process to step S118.

[0337] In step S116, the notification determination unit 122B notifies that the total value A is determined to be greater than the warning threshold value Th in step S114.f The maintenance content corresponding to the maintenance team is determined, and an alarm is issued to notify that an abnormality has occurred in the target production device 200.

[0338] In step S117, the notification determination unit 122B determines whether the total value A is greater than a given ineffectiveness threshold value Th for each maintenance team determined to be ineffective. ie No effect threshold Th ie It is the minimum value of the total value that is assumed to be notified of no effect. ie If there is no maintenance team (step S117: Yes), the notification determination unit 122B advances the process to step S118. ie In the case of a maintenance team (step S117: No), the notification determination unit 122B ends the processing.

[0339] In step S118, the notification determination unit 122B notifies that the total value A is determined to be greater than the warning threshold value Th in step S114. f At the same time, the notification determination unit 122B notifies the maintenance group that the total value A is greater than the ineffective threshold value Th in step S117. ie The maintenance content corresponds to the maintenance team.

[0340] With this structure, the maintenance display device 100B according to the third embodiment can notify not only the maintenance contents that are assumed to improve the production device 200, but also the contents of maintenance operations that were performed in the past but were ineffective. This can avoid the situation where the user repeatedly performs ineffective maintenance operations, thereby shortening the time required for maintenance and reducing the labor required for maintenance.

[0341] In the maintenance display device according to the present embodiment, the equipment state diagnosis model generation unit generates a new equipment state diagnosis model based on production actual result data before the time point of the maintenance operation determined to be ineffective and maintenance actual result data related to the maintenance operation. The notification unit notifies the content of the maintenance operation determined to be effective as an effective maintenance operation, and notifies the content of the maintenance operation associated with the equipment state diagnosis model generated based on the maintenance actual result data related to the maintenance operation determined to be ineffective as an ineffective maintenance operation.

[0342] (Variation)

[0343] Above, the embodiments of the present disclosure are described with reference to the accompanying drawings, but the present disclosure is not limited to this example. It goes without saying that a person skilled in the art can think of various variations or modifications within the scope of the technical solution, and they can certainly be understood as belonging to the technical scope of the present disclosure. In addition, the constituent elements in the above embodiments can also be arbitrarily combined within the scope of the disclosed subject matter.

[0344] In the above-mentioned embodiment, in the learning process, in the maintenance effect determination process performed by the maintenance effect determination unit 123, the maintenance effect is determined by determining whether the difference in the defect rate before and after the maintenance work is greater than a given threshold value (see Fig. 12A as well as Fig. 12B ).

[0345] However, the maintenance effect determination unit 123 may use other methods to determine whether the maintenance work is effective. Fig.26A as well as Fig.26B This is a diagram for explaining a modified example of the method of determining whether or not a maintenance work is effective, which is performed by the maintenance effect determination unit 123 in the learning process.

[0346] exist Fig.26A as well as Fig.26B In the example shown, the defect rate before maintenance is not referenced, but based on the defect rate after maintenance Nf after Whether it is greater than a given threshold (such as 20%) is used to determine whether it has an effect. Fig.26A In the example shown, Nf after =0, which is less than the given threshold of 20%, so it is judged to be effective. Fig.26B In the example shown, Nf after =100%, which is greater than the given threshold of 20%, so it is judged to be ineffective.

[0347] Likewise, even during the updating process, the maintenance effect determination unit 123 may determine whether the maintenance work is effective or not using a method different from that in the above-described embodiment.

[0348] In the above-mentioned embodiment, in the updating process, in the maintenance effect determination process performed by the maintenance effect determination unit 123, the maintenance effect is determined based on whether the difference in the degree of consistency before and after the maintenance work is greater than a given threshold value (see Fig.19A as well as Fig.19B ).

[0349] Fig.27A as well as Fig.27B This is a diagram for explaining a modified example of the method of determining whether or not the maintenance work is effective, which is performed by the maintenance effect determination unit 123 in the update process.

[0350] exist Fig.27A as well as Fig.27B In the example shown, the consistency before maintenance is not referenced, but based on the consistency after maintenance C after Whether it is greater than a given threshold (such as 0.30) is used to determine whether it has an effect. Fig.27A In the example shown, C after =0.20, which is less than the given threshold of 0.30, so it is judged to be effective. Fig.27B In the example shown, C after =0.90, which is greater than the given threshold of 0.30, and is therefore judged to have no effect.

[0351] According to the present disclosure, it is possible to detect a sign of abnormality in a device.

[0352] Industrial Applicability

[0353] The present disclosure is useful in an apparatus for outputting information for displaying information related to maintenance of production equipment.

Claims

1. A method for generating a learned model, wherein the learned model is used for maintenance of a production device, wherein the production device comprises: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. In the method for generating the learned model, The first data indicating the position of the first end surface corresponding to the time information and the second data indicating the position of the second end surface corresponding to the time information are acquired from the sensor, wherein: The time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced. When it is determined based on the first data and the second data that the position of the first end face represented by the first data is continuous and the position of the second end face represented by the second data is continuous and intersects, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor is output to a display device for maintenance, Based on the time information, the first data and the second data of each of the given number or more winding bodies generated before the maintenance and having different reading times or production times of the winding bodies are extracted, and the first defective rate of the given number or more winding bodies read before the maintenance is calculated based on the extracted first data and the second data, Based on the time information, the first data and the second data of each of the given number or more winding bodies generated after the maintenance and having different reading times or production times of the winding bodies are extracted, and the second defective rate of the given number or more winding bodies read after the maintenance is calculated based on the extracted first data and the second data, When it is judged that the first difference between the first defective rate and the second defective rate is less than a given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are not used to generate the learned model. On the other hand, when it is judged that the first difference is greater than the given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are used to generate the learned model.

2. The method for generating a learning model according to claim 1, wherein: The first electrode sheet is a positive electrode sheet of a battery, and the second electrode sheet is a negative electrode sheet of a battery.

3. The method for generating a learning model according to claim 1, wherein: The first electrode sheet is a negative electrode sheet of a battery, and the second electrode sheet is a positive electrode sheet of a battery.

4. A method for generating a learned model, wherein the learned model is used for maintenance of a production device, wherein the production device comprises: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. In the method for generating the learned model, Acquire from the sensor first data indicating the position of the first end surface corresponding to time information and second data indicating the position of the second end surface read along the radial direction corresponding to the time information, wherein: The time information indicates the time when the winding bodies are read along the radial direction of each of the given number or more winding bodies or the time when the winding bodies are produced. When it is determined based on the first data and the second data that the position of the first end face represented by the first data is continuous and the position of the second end face represented by the second data is continuous and intersects, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor is output to a display device for maintenance, Based on the time information, the first data and the second data of each of the given number of winding bodies having different reading times or production times are extracted before and after the maintenance, respectively. When it is judged based on the extracted first data and the second data that the second defective rate of the given number of winding bodies read after the maintenance is greater than a given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are not used to generate the learned model. On the other hand, when it is judged that the second defective rate is less than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to generate the learned model.

5. The method for generating a learning model according to claim 4, wherein: The first electrode sheet is a positive electrode sheet of a battery, and the second electrode sheet is a negative electrode sheet of a battery.

6. The method for generating a learning model according to claim 4, wherein: The first electrode sheet is a negative electrode sheet of a battery, and the second electrode sheet is a positive electrode sheet of a battery.

7. A method for generating a learned model, wherein the learned model is used for maintenance of a production device, wherein the production device comprises: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. In the method for generating the learned model, The first data indicating the position of the first end surface corresponding to the time information and the second data indicating the position of the second end surface corresponding to the time information are acquired from the sensor, wherein: The time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced. In a case where it is judged based on the first data and the second data that the position of the first end face represented by the first data is continuous and the position of the second end face represented by the second data is continuously intersected, a learning model is generated for judging that the cause of the defect of the winding body is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor, Based on the judgment using the learned model, information indicating that the cause of the failure is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core, or the sensor is output to a display device for maintenance. When it is judged based on the time information that the second difference between the first probability and the second probability is less than a given value, the first data and the second data of each of the given number of windings read before the maintenance are not used to update the learned model. On the other hand, when it is judged that the second difference is greater than the given value, the first data and the second data of each of the given number of windings read before the maintenance are used to update the learned model, wherein the first probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings generated before the maintenance and different from the time of reading or the time of producing the winding into the learned model, and the second probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings generated after the maintenance and different from the time of reading or the time of producing the winding into the learned model.

8. The method for generating a learning model according to claim 7, wherein: The first electrode sheet is a positive electrode sheet of a battery, and the second electrode sheet is a negative electrode sheet of a battery.

9. The method for generating a learning model according to claim 7, wherein: The first electrode sheet is a negative electrode sheet of a battery, and the second electrode sheet is a positive electrode sheet of a battery.

10. A method for generating a learned model, wherein the learned model is used for maintenance of a production device, wherein the production device comprises: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. In the method for generating the learned model, The first data indicating the position of the first end surface corresponding to the time information and the second data indicating the position of the second end surface corresponding to the time information are acquired from the sensor, wherein: The time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced. In a case where it is judged based on the first data and the second data that the position of the first end face represented by the first data is continuous and the position of the second end face represented by the second data is continuously intersected, a learning model is generated for judging that the cause of the defect of the winding body is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor, Based on the judgment using the learned model, information indicating that the cause of the failure is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core, or the sensor is output to a display device for maintenance. When it is judged that the second probability is greater than a given value based on the time information, the first data and the second data of each of the given number of winding bodies read before the maintenance are not used to update the learned model. On the other hand, when it is judged that the second probability is less than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to update the learned model, wherein the second probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding body into the learned model.

11. The method for generating a learning model according to claim 10, wherein: The first electrode sheet is a positive electrode sheet of a battery, and the second electrode sheet is a negative electrode sheet of a battery.

12. The method for generating a learning model according to claim 10, wherein: The first electrode sheet is a negative electrode sheet of a battery, and the second electrode sheet is a positive electrode sheet of a battery.

13. An information output device for displaying information related to maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of the winding bodies of the given number or more. The information output device acquires first data indicating the position of the first end surface corresponding to time information and second data indicating the position of the second end surface corresponding to the time information from the sensor, wherein: The time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced. The device for outputting information comprises: a notification judgment unit, based on the first data and the second data, judging whether the winding body is defective based on whether the continuity of the position of the first end face represented by the first data intersects with the continuity of the position of the second end face represented by the second data, and in the case of being defective, outputting information indicating that the cause of the defect is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor to a display device for maintenance; and A model generating unit extracts the first data and the second data of each of the given number of winding bodies generated before the maintenance and different from the time of reading or the time of producing the winding bodies based on the time information, and calculates the first defect rate of the given number of winding bodies read before the maintenance based on the extracted first data and the second data; extracts the first data and the second data of each of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding bodies based on the time information, and calculates the second defect rate of the given number of winding bodies read after the maintenance based on the extracted first data and the second data; and generates a learned model by using the first data and the second data of each of the given number of winding bodies read before the maintenance when the first difference between the first defect rate and the second defect rate is judged to be greater than the given value instead of using the first data and the second data of each of the given number of winding bodies read before the maintenance when the first difference is judged to be greater than the given value.

14. An information output device for displaying information related to maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. The information output device acquires first data indicating the position of the first end surface corresponding to time information and second data indicating the position of the second end surface corresponding to the time information from the sensor, wherein: The time information indicates the time when the winding body is read along the radial direction or the time when the winding body is produced. The device for outputting information comprises: a notification judgment unit, based on the first data and the second data, judging whether the winding body is defective based on whether the continuity of the position of the first end face represented by the first data intersects with the continuity of the position of the second end face represented by the second data, and in the case of being defective, outputting information indicating that the cause of the defect is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor to a display device for maintenance; and The model generating unit extracts the first data and the second data of each of the given number of winding bodies having different reading times or production times before and after the maintenance based on the time information, and generates a learned model by using the first data and the second data of each of the given number of winding bodies read before the maintenance when it is judged that the second defective rate of the given number of winding bodies read after the maintenance is greater than the given value based on the extracted first data and the second data, instead of using the first data and the second data of each of the given number of winding bodies read before the maintenance when it is judged that the second defective rate of the given number of winding bodies read after the maintenance is greater than the given value.

15. An information output device for displaying information related to maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. The information output device acquires first data indicating the position of the first end surface corresponding to time information and second data indicating the position of the second end surface corresponding to the time information from the sensor, wherein: The time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced. The device for outputting information comprises: A model generating unit determines whether the winding body is defective based on the first data and the second data and based on whether the continuity of the position of the first end face represented by the first data intersects with the continuity of the position of the second end face represented by the second data, and if it is defective, generates a learned model for judging that the defective cause is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor; and The notification judgment unit outputs information indicating that the cause of the failure is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor to a display device for maintenance based on the judgment using the learned model. When the model generating unit determines that the second difference between the first probability and the second probability is less than a given value based on the time information, the first data and the second data of each of the given number of winding bodies read before the maintenance are not used to update the learned model. On the other hand, when it is determined that the second difference is greater than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to update the learned model, wherein the first probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each of the given number of winding bodies generated before the maintenance and different from the time of reading or the time of producing the winding body into the learned model, and the second probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding body into the learned model.

16. An information output device for displaying information related to maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. The information output device acquires first data indicating the position of the first end surface corresponding to time information and second data indicating the position of the second end surface corresponding to the time information from the sensor, wherein: The time information indicates the time when the winding bodies of the given number or more are read or the time when the winding bodies are produced. The device for outputting information comprises: A model generating unit determines whether the winding body is defective based on the first data and the second data and based on whether the continuity of the position of the first end face represented by the first data intersects with the continuity of the position of the second end face represented by the second data, and if it is defective, generates a learned model for judging that the defective cause is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor; and The notification judgment unit outputs information indicating that the cause of the failure is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core or the sensor to a display device for maintenance based on the judgment using the learned model. When the model generating unit determines that the second probability is greater than a given value based on the time information, the first data and the second data of each of the given number of winding bodies read before the maintenance are not used to update the learned model. On the other hand, when it is determined that the second probability is less than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to update the learned model, wherein the second probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding body into the learned model.

17. A program product, comprising a program executed by a computer, the program for generating a learned model for maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. The program causes the computer to execute the following process: The first data indicating the position of the first end surface corresponding to the time information and the second data indicating the position of the second end surface corresponding to the time information are acquired from the sensor, wherein: The time information indicates the time when the given number or more of the winding bodies are read or the time when the winding bodies are produced; When it is determined based on the first data and the second data that the position of the first end face represented by the first data is continuous and the position of the second end face represented by the second data is continuous and intersects, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor is output to a display device for maintenance; Based on the time information, extract the first data and the second data of each of the given number or more winding bodies generated before the maintenance and having different reading times or production times of the winding bodies, and calculate the first defective rate of the given number or more winding bodies read before the maintenance based on the extracted first data and the second data; Based on the time information, extract the first data and the second data of each of the given number or more winding bodies generated after the maintenance and having different reading times or production times of the winding bodies, and calculate the second defective rate of the given number or more winding bodies read after the maintenance based on the extracted first data and the second data; and When it is judged that the first difference between the first defective rate and the second defective rate is less than a given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are not used to generate the learned model. On the other hand, when it is judged that the first difference is greater than the given value, the first data and the second data of each of the winding bodies having a given number or more read before the maintenance are used to generate the learned model.

18. A program product, comprising a program executed by a computer, the program for generating a learned model for maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. The program causes the computer to execute the following process: The first data indicating the position of the first end surface corresponding to the time information and the second data indicating the position of the second end surface corresponding to the time information are acquired from the sensor, wherein: The time information indicates the time when the winding bodies are read along the radial direction of each of the given number or more winding bodies or the time when the winding bodies are produced; When it is determined based on the first data and the second data that the position of the first end face represented by the first data is continuous and the position of the second end face represented by the second data is continuous and intersects, information indicating that the winding body is defective and the cause of the defect is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor is output to a display device for maintenance; and Based on the time information, the first data and the second data of each of the given number of winding bodies having different reading times or production times are extracted before and after the maintenance, respectively. When it is judged based on the extracted first data and the second data that the second defective rate of the given number of winding bodies read after the maintenance is greater than a given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are not used to generate the learned model. On the other hand, when it is judged that the second defective rate is less than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to generate the learned model.

19. A program product, comprising a program executed by a computer, the program for generating a learned model for maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. The program causes the computer to execute the following process: The first data indicating the position of the first end surface corresponding to the time information and the second data indicating the position of the second end surface corresponding to the time information are acquired from the sensor, wherein: The time information indicates the time when the given number or more of the winding bodies are read or the time when the winding bodies are produced; In a case where it is judged based on the first data and the second data that the continuity of the position of the first end face represented by the first data and the continuous intersection of the position of the second end face represented by the second data, a learning model is generated to judge that the cause of the defect of the winding body is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor; Based on the judgment using the learned model, information indicating that the cause of the failure is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core, or the sensor is output to a display device for maintenance; and When it is judged based on the time information that the second difference between the first probability and the second probability is less than a given value, the first data and the second data of each of the given number of windings read before the maintenance are not used to update the learned model. On the other hand, when it is judged that the second difference is greater than the given value, the first data and the second data of each of the given number of windings read before the maintenance are used to update the learned model, wherein the first probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings generated before the maintenance and different from the time of reading or the time of producing the winding into the learned model, and the second probability is the probability that the defect of the winding is improved by inputting the first data and the second data of each of the given number of windings generated after the maintenance and different from the time of reading or the time of producing the winding into the learned model.

20. A program product, comprising a program executed by a computer, the program for generating a learned model for maintenance of a production device, the production device comprising: A first supply reel for supplying a first electrode sheet; a second supply reel for supplying a second electrode sheet; A first laminating roller, arranged on the first electrode sheet side; a second laminating roller, disposed on the second electrode sheet side, and forming a pair with the first laminating roller to laminate the first electrode sheet and the second electrode sheet; A winding core that winds the bonded first electrode sheet and the second electrode sheet to form a predetermined number or more of winding bodies; and The sensor reads the first end face of the first electrode sheet and the second end face of the second electrode sheet along the radial direction of each of the winding bodies of the given number or more. The program causes the computer to execute the following process: The first data indicating the position of the first end surface corresponding to the time information and the second data indicating the position of the second end surface corresponding to the time information are acquired from the sensor, wherein: The time information indicates the time when the given number or more of the winding bodies are read or the time when the winding bodies are produced; In a case where it is judged based on the first data and the second data that the continuity of the position of the first end face represented by the first data and the continuous intersection of the position of the second end face represented by the second data, a learning model is generated to judge that the cause of the defect of the winding body is any one of the first supply reel, the second supply reel, the first bonding roller, the second bonding roller, the winding core or the sensor; Based on the judgment using the learned model, information indicating that the cause of the failure is any one of the first supply reel, the second supply reel, the first laminating roller, the second laminating roller, the winding core, or the sensor is output to a display device for maintenance; and When it is judged that the second probability is greater than a given value based on the time information, the first data and the second data of each of the given number of winding bodies read before the maintenance are not used to update the learned model. On the other hand, when it is judged that the second probability is less than the given value, the first data and the second data of each of the given number of winding bodies read before the maintenance are used to update the learned model, wherein the second probability is the probability that the defect of the winding body is improved by inputting the first data and the second data of each of the given number of winding bodies generated after the maintenance and different from the time of reading or the time of producing the winding body into the learned model.

Images