Machine learning device, control system, control device, and machine learning method

Patent metadata

Field	Value
Publication number	US-11029651-B2
Application number	US-201816101968-A
Country	US
Kind code	B2
Filing date	Aug 13, 2018
Priority date	Aug 29, 2017
Publication date	Jun 8, 2021
Grant date	Jun 8, 2021

Abstract

A machine learning device includes: a state information acquisition unit configured to cause the control device to execute a tapping program to acquire from the control device, state information including a torque command value with respect to the spindle motor, a drive state including deceleration, a ratio of a movement distance in acceleration and a movement distance in deceleration; an action information output unit configured to output action information including adjustment information of the ratio of the movement distance in acceleration and the movement distance in deceleration, to the control device; a reward output unit configured to output a reward value in reinforcement learning based on a torque command value in deceleration, and a target torque command value in deceleration; and a value function update unit configured to update an action value function based on the reward value, the state information, and the action information.

First claim

What is claimed is:

1. A machine learning device that performs machine learning with respect to a control device of a machine tool, the control device controlling synchronous operation of a spindle motor and a feed axis motor,
the machine learning device comprising:
a first memory configured to store a first program; and
a first processor configured to execute the first program and control the machine learning device to:
cause the control device to execute a tapping program to acquire from the control device, state information including at least a torque command value with respect to the spindle motor, a drive state including deceleration of the spindle motor, and a ratio of a movement distance in acceleration and a movement distance in deceleration of the spindle motor;
output action information, including adjustment information of the ratio of the movement distance in acceleration and the movement distance in deceleration of the spindle motor included in the state information, to the control device;
output a reward value in reinforcement learning based on a torque command value in deceleration and a target torque command value in deceleration; and
update an action value function based on the reward value, the state information, and the action information.

2. The machine learning device according to claim 1 , wherein the first processor is further configured to execute the first program and control the machine learning device to obtain a reward by a following Formula 1 by using a torque command value Tm in deceleration, a target torque command value Tt in deceleration, and a coefficient a.

a
×

(

1
-



Tt
-
Tm



Tt

)

.

[

Formula
⁢

⁢
1

]

3. The machine learning device according to claim 1 , wherein the first processor is further configured to execute the first program and control the machine learning device to generate and output the ratio of the movement distance in acceleration and the movement distance in deceleration of the spindle motor, based on an updated value function.

4. A control system comprising the machine learning device according to claim 1 , and the control device of the machine tool, the control device controlling synchronous operation of the spindle motor and the feed axis motor.

5. A control device, comprising the machine learning device according to claim 1 , the control device controlling synchronous operation of the spindle motor and the feed axis motor.

6. The control system according to claim 4 , wherein the control device comprises:
a second memory configured to store a second program; and
a second processor configured to execute the second program and control the control device to:
create a spindle command and a feed axis command based on the tapping program;
control rotation operation of the spindle motor according to the spindle command;
detect a rotation position of the spindle motor;
control feed operation of a feed axis based on the rotation position according to the feed axis command;
acquire a total rotation amount and a maximum rotation speed of the spindle motor from a start position to a target position, from the tapping program, the total rotation amount and the maximum rotation speed of the spindle motor being used as the spindle command;
cause the spindle motor to accelerate from the start position, by speed control with the maximum rotation speed of the spindle motor as a target value;
detect a rotation amount of the spindle motor based on the rotation position during the acceleration;
detect a remained rotation amount of the spindle motor from a current position to the target position based on the total rotation amount and the rotation position;
detect current speed of the spindle motor based on the rotation position; and
cause the spindle motor to decelerate and reach the target position by position control based on the ratio of the movement distance in acceleration and the movement distance in deceleration of the spindle motor, the remained rotation amount, and the current speed, after the acceleration.

7. The control device according to claim 5 , wherein the control device comprises:
a second memory configured to store a second program; and
a second processor configured to execute the second program and control the control device to:
create a spindle command and a feed axis command based on the tapping program;
control rotation operation of the spindle motor according to the spindle command;
detect a rotation position of the spindle motor;
control feed operation of a feed axis based on the rotation position, according to the feed axis command;
acquire a total rotation amount and a maximum rotation speed of the spindle motor from a start position to a target position, from the tapping program, the total rotation amount and the maximum rotation speed of the spindle motor being used as the spindle command;
cause the spindle motor to accelerate from the start position by speed control with the maximum rotation speed of the spindle motor as a target value;
detect a rotation amount of the spindle motor based on the rotation position during the acceleration;
detect a remained rotation amount of the spindle motor from a current position to the target position based on the total rotation amount and the rotation position;
detect a current speed of the spindle motor based on the rotation position; and
cause the spindle motor to decelerate and reach the target position by position control based on the ratio of the movement distance in acceleration and the movement distance in deceleration of the spindle motor, the remained rotation amount, and the current speed, after the acceleration.

8. A machine learning method of a machine learning device that performs machine learning with respect to a control device of a machine tool, the control device controlling synchronous operation of a spindle motor and a feed axis motor,
the machine learning method comprising:
acquiring from the control device, state information including at least a torque command value with respect to the spindle motor, a drive state including deceleration of the spindle motor, and a ratio of a movement distance in acceleration and a movement distance in deceleration of the spindle motor, by causing the control device to execute a tapping program;
outputting action information, including adjustment information of the ratio of the movement distance in acceleration and the movement distance in deceleration of the spindle motor included in the state information, to the control device;
outputting a reward value in reinforceme

Assignees

• Fanuc Corp

Inventors

• Tajima Daisuke
• Morita Yuuki

Classifications

• G06N3/098

  Distributed learning, e.g. federated learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• G06N3/006

  based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO] · CPC title

• G05B2219/45216

  Tapping · CPC title