# US 2021/0280895 A1 · Learned model generation method and apparatus

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0280895 A1
- Document ID: US20210280895A1
- Status: Published application
- Kind: A1
- Application number: 17/180,911
- Filed: 2021-02-22
- Published: 2021-09-09
- Earliest claimed: 2020-03-09
- Assignees: Panasonic Intellectual Property Management Co., Ltd.
- Applicants: Panasonic Intellectual Property Management Co., Ltd.
- Inventors: SHINYA HATAKENAKA
- Web page: https://patentgrep.com/patent/US20210280895A1

## CPC

- H01M 10/0409: {for cells with wound electrodes}
- B65H 26/00: Warning or safety devices, e.g. automatic fault detectors, stop-motions, for web-advancing mechanisms (safety devices in general F16P; investigating chemical or physical properties of materials in general G01N; indicating devices in general G08B)
- G05B 23/0283: {Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] (preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06)}
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- H01G 4/32: Wound capacitors
- G06N 3/08: Learning methods
- Y02E 60/10: Energy storage using batteries
- Y02P 70/50: Manufacturing or production processes characterised by the final manufactured product
- H01M: PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY
- B65H: HANDLING THIN OR FILAMENTARY MATERIAL, e.g. SHEETS, WEBS, CABLES
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H01G: CAPACITORS; CAPACITORS, RECTIFIERS, DETECTORS, SWITCHING DEVICES, LIGHT-SENSITIVE OR TEMPERATURE-SENSITIVE DEVICES OF THE ELECTROLYTIC TYPE (selection of specified materials as dielectric H01B3/00; capacitors having potential barriers H10D1/62, H10K10/10)
- Y02E: REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02P: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS

## Abstract

First group data indicating a position of a first end surface of a first sheet material and second group data indicating a position of a second end surface of a second sheet material are acquired from an inspection machine, and information indicating whether or not a winding body has a defect, and which constituent of a sheet material supplier is a cause of the defect in a case where the winding body has the defect, on the basis of a positional relationship among continuous positions of the first end surface indicated by the first group data, continuous positions of the second end surface indicated by the second group data, and reference lines, is output to a display for maintenance.

## Claims

### Claim 1

A learned model generation method of generating a learned model for maintenance of a winding apparatus including
a sheet material supplier having a first supply mechanism that supplies a first electrode sheet, a second supply mechanism that supplies a second electrode sheet, a first bonding roller that is provided on a first electrode sheet side, and a second bonding roller that is provided on a second electrode sheet side, and is paired with the first bonding roller to bond the first electrode sheet and the second electrode sheet to each other,
a first winding core,
a drive mechanism that moves the first winding core to a predetermined winding position, and winds the first electrode sheet and the second electrode sheet in an overlapping manner on the first winding core, and
a sensor that reads a first end surface of the first electrode sheet and a second end surface of the second electrode sheet along a radial direction of a first winding body in which the first electrode sheet and the second electrode sheet are wound in an overlapping manner by a plurality of turns on the first winding core, the learned model generation method comprising:
acquiring, from the sensor, first group data indicating a position of the first end surface read along the radial direction of the first winding body, and second group data indicating a position of the second end surface read along the radial direction of the first winding body;
determining whether or not the first winding body has a defect and whether or not a cause of the defect is the sheet material supplier when the first winding body has the defect on the basis of a positional relationship among continuous positions of the first end surface indicated by the first group data, continuous positions of the second end surface indicated by the second group data, and reference lines;
determining which of the first supply mechanism, the second supply mechanism, the first bonding roller, and the second bonding roller is the cause of the defect in a case where it is determined that the cause of the defect is the sheet material supplier;
outputting information for a notification of the cause of the defect to a display apparatus; and
determining whether or not the first group data and the second group data read before the cause of the defect is maintained are used for generating the learned model on the basis of a first occurrence degree of the defect of the first winding body before the cause of the defect is maintained and a second occurrence degree of the defect of the first winding body after the cause of the defect is maintained, and generating the learned model by using the first group data and the second group data read before the cause of the defect is maintained in a case where it is determined that the first group data and the second group data are used.

### Claim 2 (depends on 1)

The learned model generation method of 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 the battery.

### Claim 3 (depends on 1)

The learned model generation method of 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 the battery.

### Claim 4 (depends on 1)

The learned model generation method of claim 1, wherein
in a case of generating the learned model,
the first occurrence degree is a first defect ratio indicating a defect ratio of the first winding body before the cause of the defect is maintained, and the second occurrence degree is a second defect ratio indicating a defect ratio of the first winding body after the cause of the defect is maintained, and
in a case where a first difference between the first defect ratio and the second defect ratio is less than a predetermined value, the first group data and the second group data read before the cause of the defect is maintained are not used for generating the learned model, and, in a case where the first difference is greater than or equal to the predetermined value, the learned model is generated by using the first group data and the second group data read before the cause of the defect is maintained.

### Claim 5 (depends on 1)

The learned model generation method of claim 1, wherein
in a case where it is determined whether or not the first winding body has the defect and whether or not the cause of the defect is the sheet material supplier when the first winding body has the defect, when continuous positions of the first end surface of a winding body after replacement wound after replacement work for the first electrode sheet is performed in the first supply mechanism are not within a predetermined range, the first winding body is determined as being defective, and pre-replacement group data indicating a position of the first end surface of a winding body before the replacement wound before the replacement work is performed is acquired, and
in a case where which constituent of the sheet material supplier is the cause of the defect is determined, when continuous positions of the first end surface of the winding body before the replacement indicated by the pre-replacement group data are within the predetermined range, it is determined that the cause of the defect is the first supply mechanism.

### Claim 6 (depends on 1)

The learned model generation method of claim 1, wherein
the winding apparatus further includes a second winding core that generates a second winding body by winding the first electrode sheet and the second electrode sheet in an overlapping manner at the predetermined winding position with the drive mechanism,
the sensor further reads a third end surface of the first electrode sheet and a fourth end surface of the second electrode sheet along to a radial direction of the second winding body in which the first electrode sheet and the second electrode sheet are wound in an overlapping manner by a plurality of turns on the second winding core,
in the learned model generation method,
the first group data indicating the position of the first end surface read along the radial direction of the first winding body, the second group data indicating the position of the second end surface read along the radial direction of the first winding body, third group data indicating a position of the third end surface read along the radial direction of the second winding body, and fourth group data indicating a position of the fourth end surface read along the radial direction of the second winding body are acquired from the sensor, and
in a case where continuous positions of the first end surface and continuous positions of the second end surface indicated by the first group data and the second group data are tilted with respect to the reference lines, and continuous positions of the third end surface and continuous positions of the fourth end surface indicated by the third group data and the fourth group data are tilted with respect to the reference lines, it is determined that the first winding body and the second winding body have defects, and a cause of the defects is the first bonding roller or the second bonding roller.

### Claim 7 (depends on 1)

The learned model generation method of claim 1, wherein
the sheet material supplier of the winding apparatus further includes a cylinder that applies a tension between the first winding core, and the first bonding roller and the second bonding roller when the first electrode sheet and the second electrode sheet are wound in an overlapping manner on the first winding core, and
in the learned model generation method, in a case where at least either of continuous positions of the first end surface of the winding body and continuous positions of the second end surface of the winding body meander with respect to the reference lines, it is determined that the first winding body has a defect and a cause of the defect is the cylinder.

### Claim 8

A learned model generation method of generating a learned model for maintenance of a winding apparatus including
a sheet material supplier having a first supply mechanism that supplies a first electrode sheet, a second supply mechanism that supplies a second electrode sheet, a first bonding roller that is provided on a first electrode sheet side, and a second bonding roller that is provided on a second electrode sheet side, and is paired with the first bonding roller to bond the first electrode sheet and the second electrode sheet to each other,
a first winding core,
a drive mechanism that moves the first winding core to a predetermined winding position, and winds the first electrode sheet and the second electrode sheet in an overlapping manner on the first winding core, and
a sensor that reads a first end surface of the first electrode sheet and a second end surface of the second electrode sheet along a radial direction of a first winding body in which the first electrode sheet and the second electrode sheet are wound in an overlapping manner by a plurality of turns on the first winding core, the learned model generation method comprising:
acquiring, from the sensor, first group data indicating a position of the first end surface read along the radial direction of the first winding body, and second group data indicating a position of the second end surface read along the radial direction of the first winding body;
determining whether or not the first winding body has a defect and whether or not a cause of the defect is the sheet material supplier when the first winding body has the defect on the basis of a positional relationship among continuous positions of the first end surface indicated by the first group data, continuous positions of the second end surface indicated by the second group data, and reference lines;
determining which constituent of the sheet material supplier is the cause of the defect in a case where it is determined that the cause of the defect is the sheet material supplier;
outputting information for a notification of the cause of the defect to a display apparatus; and
not using the first group data and the second group data read before the cause of the defect is maintained for updating the learned model in a case where it is determined that a second difference between a first probability that the defect of the first winding body is improved and a second probability that the defect of the first winding body is improved is less than a predetermined value, the first probability being obtained by inputting the first group data and the second group data before the cause of the defect is maintained to the learned model, and the second probability being obtained by inputting the first group data and the second group data after the cause of the defect is maintained to the learned model, and updating the learned model by using the first group data and the second group data before the cause of the defect is maintained in a case where it is determined that the second difference is greater than or equal to the predetermined value.

### Claim 9

An apparatus outputting information for displaying information regarding maintenance of a winding apparatus including
a sheet material supplier having a first supply mechanism that supplies a first electrode sheet, a second supply mechanism that supplies a second electrode sheet, a first bonding roller that is provided on a first electrode sheet side, and a second bonding roller that is provided on a second electrode sheet side, and is paired with the first bonding roller to bond the first electrode sheet and the second electrode sheet to each other,
a first winding core,
a drive mechanism that moves the first winding core to a predetermined winding position and winds the first electrode sheet and the second electrode sheet in an overlapping manner on the first winding core, and
a sensor that reads a first end surface of the first electrode sheet and a second end surface of the second electrode sheet along a radial direction of a first winding body in which the first electrode sheet and the second electrode sheet are wound in an overlapping manner by a plurality of turns on the first winding core, the apparatus comprising:
an acquirer that acquires, from the sensor, first group data indicating a position of the first end surface read along the radial direction of the first winding body, and second group data indicating a position of the second end surface read along the radial direction of the first winding body;
a notification determinator that determines whether or not the first winding body has a defect and whether or not a cause of the defect is the sheet material supplier when the first winding body has the defect on the basis of a positional relationship among continuous positions of the first end surface indicated by the first group data, continuous positions of the second end surface indicated by the second group data, and reference lines, determines which constituent of the sheet material supplier is the cause of the defect in a case where it is determined that the cause of the defect is the sheet material supplier, and outputs information for a notification of the cause of the defect to a display apparatus for maintenance; and
a model generator that determines whether or not the first group data and the second group data read before the cause of the defect is maintained are used for generating the learned model on the basis of a first occurrence degree of the defect of the first winding body before the cause of the defect is maintained and a second occurrence degree of the defect of the first winding body after the cause of the defect is maintained, and generates the learned model by using the first group data and the second group data read before the cause of the defect is maintained in a case where it is determined that the first group data and the second group data are used.
