Machine learning device and robot system to learn processing order of laser processing robot and machine learning method thereof

Patent metadata

Field	Value
Publication number	US-10518357-B2
Application number	US-201715808921-A
Country	US
Kind code	B2
Filing date	Nov 10, 2017
Priority date	Nov 29, 2016
Publication date	Dec 31, 2019
Grant date	Dec 31, 2019

Abstract

A machine device for learning a processing order of a laser processing robot, includes a state observation unit that observes, as a state variable, one of a plasma light from a laser processing point of the laser processing robot and a processing sound from the laser processing point of the laser processing robot; a determination data obtaining unit that receives, as determination data, a cycle time in which the laser processing robot completes processing; and a learning unit that learns the processing order of the laser processing robot based on an output of the state observation unit and an output of the determination data obtaining unit.

First claim

What is claimed is:

1. A machine learning device for learning a processing order of a robot system which includes a plurality of laser processing robots and the plurality of laser processing robots complete all processing in the system, comprising:
a processor that
observes, as a state variable, one of a plasma light from a laser processing point of each of the laser processing robots and a processing sound from the laser processing point of each of the laser processing robots,
receives, as determination data, a system cycle time in which the plurality of laser processing robots complete all processing in the system, and
learns a processing order of the robot system including the plurality of laser processing robots based on the state variable and the determination data.

2. The machine learning device according to claim 1 , wherein
the processor further receives, as the determination data, one of a processing speed at which each of the laser processing robots performs a laser processing, a focal length of laser, a posture of a processing tool, and a flow rate of assist gas.

3. The machine learning device according to claim 1 wherein
the processor further decides an operation of the laser processing robot based on the processing order of the robot system including the plurality of laser processing robots that the processor has learned.

4. The machine learning device according to claim 1 , wherein
the processor:
calculates a reward based on the state variable and of the determination data, and
updates a value function that determines a value of a processing order of the robot system including the plurality of laser processing robots based on the state variable, the determination data, and the reward.

5. The machine learning device according to claim 4 , wherein
the processor sets a negative reward when the system cycle time is long and sets a positive reward when the system cycle time is short.

6. The machine learning device according to claim 4 , wherein
the processor further sets a negative reward when the plasma light from the laser processing point is far from an optimal plasma light value, and sets a positive reward when the plasma light from the laser processing point is close to the optimal plasma light value, or
the processor further sets a negative reward when the processing sound from the laser processing point is far from an optimal processing sound value, and sets a positive reward when the processing sound from the laser processing point is close to the optimal processing sound value.

7. The machine learning device according to claim 4 , wherein
the processor further sets a negative reward when the processing speed at which each of the laser processing robots performs laser processing is low, and sets a positive reward when the processing speed at which each of the laser processing robots performs laser processing is high.

8. The machine learning device according to claim 1 , further comprising a neural network.

9. The machine learning device according to claim 1 , wherein the machine learning device is located on a cloud server or a fog server.

10. A robot system, comprising:
a laser processing robot control device;
a plurality of laser processing robots controlled by the laser processing robot control device; and
a machine learning device for learning a processing order of the plurality of laser processing robots which complete all processing in the system,
the machine learning device comprising a processor that
observes, as a state variable, one of a plasma light from a laser processing point of each of the laser processing robots and a processing sound from the laser processing point of each of the laser processing robots,
receives, as determination data, a system cycle time in which the plurality of laser processing robots complete all processing in the system, and
learns a processing order of the robot system including the plurality of laser processing robots based on the state variable and the determination data.

11. The robot system according to claim 10 , each of the laser processing robots comprising at least one of:
an optical sensor that detects the plasma light from the laser processing point of the laser processing robot within a certain wavelength range, and
a sound sensor that detects the processing sound from the laser processing point of the laser processing robot within a certain frequency range.

12. A machine learning method for learning a processing order of a robot system which includes a plurality of laser processing robots and the plurality of laser processing robots complete all processing in the system, comprising:
observing, as a state variable, one of a plasma light from a laser processing point of each of the laser processing robots and a processing sound from the laser processing point of each of the laser processing robots;
receiving, as determination data, a system cycle time in which the plurality of laser processing robots complete all processing in the system; and
learning a processing order of the robot system including the plurality of laser processing robots based on the observed state variable and the received determination data.

Assignees

• Fanuc Corp

Inventors

• Furuya Yoshitake

Classifications

• B25J9/163 Primary

  learning, adaptive, model based, rule based expert control · CPC title

• B23K26/032

  using optical means · CPC title

• B23K26/702

  Auxiliary equipment · CPC title

• B25J11/005

  Manipulators for mechanical processing tasks · CPC title

• B23K26/0884 Primary

  in at least three axial directions, e.g. manipulators, robots · CPC title