Method for determining state information relating to a belt grinder by means of a machine learning system

Patent metadata

Field	Value
Publication number	US-12397392-B2
Application number	US-202017595794-A
Country	US
Kind code	B2
Filing date	May 11, 2020
Priority date	May 27, 2019
Publication date	Aug 26, 2025
Grant date	Aug 26, 2025

Abstract

A method determines state information relating to a belt grinder. The belt grinder has at least one abrasive belt for grinding a workpiece. The method includes providing measurement data relating to the belt grinder, and determining the state information from the measurement data using a machine learning system. The machine learning system is configured to determine the state information based on the provided measurement data.

First claim

The invention claimed is:

1. A method for determining state information relating to a belt grinder, the belt grinder having at least one abrasive belt for grinding a workpiece, the method comprising:
generating measurement data relating to the belt grinder; and
determining the state information from the generated measurement data using a trained machine learning system, wherein the trained machine learning system includes a plurality of steps and configured to determine the state information based on the generated measurement data, wherein the plurality of steps comprising:
providing training data comprising training input data and training output data, wherein the training input data comprise measurement data relating to (i) a belt grinder for a plurality of pieces of state information, and (ii) at least two belt grinders of different types, at least two belt grinders of the same type with a different use, or two belt grinders of the same type with the same use, and wherein the training output data comprise in each case at least one assigned piece of the state information relating to the belt grinder;
training the machine learning system, wherein parameters of the machine learning system are adapted such that the machine learning system determines respectively assigned training output data depending on the adapted parameters and depending on the provided training input data; and
adding the trained machine learning system to a computer device of the belt grinder,
wherein the trained machine learning system is configured to determine the plurality of pieces of state information by:
receiving the measurement data; and
determining the state information from the received measurement data, and

wherein the belt grinder has at least one abrasive belt for grinding a workpiece.

2. The method as claimed in claim 1 , wherein the measurement data are generated using at least one sound sensor.

3. The method as claimed in claim 2 , wherein:
the measurement data are generated using at least one further sensor, and the at least one further sensor is selected from a list of sensors comprising:
sensors for current consumption,
air temperature sensors,
humidity sensors,
distance sensors,
range sensors,
imaging sensors,
temperature sensors,
IR sensors,
thermal imaging sensors,
thickness-measuring sensors,
torque sensors,
dust quantity measuring sensors,
inertial sensors,
acceleration sensors,
path length sensors,
location sensors,
touch-sensitive sensors, and
reflectance sensors.

4. The method as claimed in claim 1 , wherein the measurement data are retrieved from the belt grinder selectively.

5. The method as claimed in claim 1 , wherein the trained machine learning system comprises a neural network.

6. The method as claimed in claim 1 , wherein the trained machine learning system is configured to determine the state information at least relating to one of the following properties:
a property that characterizes the workpiece to be processed,
a property that characterizes manufacturing defects on the workpiece,
a property that characterizes an operating mode or operating parameter of the belt grinder,
a property that characterizes incorrect settings of the belt grinder,
a property that characterizes a load distribution of the belt grinder,
a property that characterizes a degree of wear or a wearing of the belt grinder,
a property that characterizes an abrasive belt used in the belt grinder,
a property that characterizes clogging and/or blunting of the abrasive belt, and
a property that characterizes a defect of the abrasive belt.

7. The method as claimed in claim 1 , wherein the belt grinder is controlled at least partly based on the determined state information and/or a piece of information is output by an output device at least partly based on the determined state information.

8. The method as claimed in claim 1 , further comprising:
filtering voice components from the measurement data before determining the state information.

9. A method for training a machine learning system, comprising:
providing training data comprising training input data and training output data, wherein the training input data comprise measurement data relating to (i) a belt grinder for a plurality of pieces of state information, and (ii) at least two belt grinders of different types, at least two belt grinders of the same type with a different use, or two belt grinders of the same type with the same use, and wherein the training output data comprise in each case at least one assigned piece of the state information relating to the belt grinder;
training the machine learning system, wherein parameters of the machine learning system are adapted such that the machine learning system determines respectively assigned training output data depending on the adapted parameters and depending on the provided training input data; and
adding the trained machine learning system to a computer device of the belt grinder,
wherein the trained machine learning system is configured to determine the plurality of pieces of state information by:
receiving the measurement data; and
determining the state information from the received measurement data, and wherein the belt grinder has at least one abrasive belt for grinding a workpiece.

10. The method as claimed in claim 9 , further comprising:
receiving further measurement data relating to the belt grinder, wherein at least one piece of the state information relating to the belt grinder is assigned to the further measurement data; and
further training the machine learning system using the received further measurement data.

11. The method as claimed in claim 10 , wherein the training input data comprise the measurement data and the further measurement data for a plurality of pieces of the state information and the training output data comprise in each case at least one assigned piece of the state information.

12. The method as claimed in claim 9 , wherein the training output data are selected from a list of pieces of the state information relating to at least the following properties:
a property that characterizes the workpiece to be processed,
a property that characterizes manufacturing defects on the workpiece,
a property that characterizes an operating mode or operating parameter of the belt grinder,
a property that characterizes incorrect settings of the belt grinder,
a property that characterizes a load distribution of the belt grinder,
a property that characterizes a degree of wear or a wearing of the belt grinder,
a property that characterizes an abrasive belt used in the belt grinder,
a property that characterizes clogging and/or blunting of the abrasive belt, and
a property that characterizes a defect of the abrasive belt,

or combinations thereof.

13. The method as claimed in claim 9 , wherein a non-transitory computer-readable storage medium is configured to store a computer program, which when executed on a computer device causes the computer device to carry out the method.

14. A belt grinder comprising:
an abrasive belt configured to grind a workpiece;
at least one sound sensor configured to generate measurement data; and
a trained machine learning system as claimed in claim 9 , the trained machine learning system configured to receive the measurement data and to determine a piece of state information relating to the belt grinder based on the received measurement data.

15. The belt grinder as claimed in claim 14 , further comprising:
a grinding shoe,
wherein the at least one sound sensor is arranged on or in the grinding

Assignees

• Bosch Gmbh Robert

Inventors

• Pleuger Wolfgang
• Klein Thorsten
• Kipfer Thomas
• Fischer Johannes
• Amann Matthias
• Kuppke Helena

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G05B2219/32335

  Use of ann, neural network · CPC title

• G05B19/182

  characterised by the machine tool function, e.g. thread cutting, cam making, tool direction control (G05B19/21 - G05B19/40 take precedence) · CPC title

• B24B49/16

  taking regard of the load · CPC title