Machine learning system, control device, and machine learning method for optimizing filter coefficients

Patent metadata

Field	Value
Publication number	US-11226599-B2
Application number	US-201916707113-A
Country	US
Kind code	B2
Filing date	Dec 9, 2019
Priority date	Jan 24, 2019
Publication date	Jan 18, 2022
Grant date	Jan 18, 2022

Abstract

A machine learning system that optimizes coefficients of at least one filter provided in a servo control device that controls a motor includes: an initial setting unit that sets initial values of the coefficients of the filter so as to attenuate at least one specific frequency component; a frequency characteristic calculation unit that calculates at least one of an input/output gain and an input/output phase delay of the servo control device based on an input signal and an output signal of the servo control device, of which the frequencies change; and a filter characteristic removing unit that removes filter characteristics of an initial filter in which the initial values are set to the filter from at least one of the input/output gain and the input/output phase delay obtained based on the input signal and the output signal obtained using the initial filter at the start of machine learning.

First claim

What is claimed is:

1. A machine learning system that performs machine learning of optimizing coefficients of at least one filter provided in a servo control device that controls a motor, the system comprising:
an initial setting unit that sets initial values of the coefficients of the filter so as to attenuate at least one specific frequency component;
a frequency characteristic calculation unit that calculates at least one of an input/output gain and an input/output phase delay of the servo control device on the basis of an input signal and an output signal of the servo control device, of which the frequencies change; and
a filter characteristic removing unit that removes filter characteristics of an initial filter in which the initial values are set to the filter by the initial setting unit from at least one of the input/output gain and the input/output phase delay obtained on the basis of the input signal and the output signal obtained using the initial filter at the start of machine learning, wherein
the machine learning system starts machine learning of the coefficients of the filter in which the filter characteristics of the initial filter are removed by the filter characteristic removing unit so that at least the input/output gain decreases or the phase delay decreases.

2. The machine learning system according to claim 1 , wherein
the input signal of which the frequency changes is a sinusoidal wave of which the frequency changes, and
the sinusoidal wave is generated by a frequency generation unit, and the frequency generation unit is provided inside or outside the servo control device.

3. The machine learning system according to claim 1 , further comprising a machine learning unit including:
a state information acquisition unit that acquires state information including the input/output gain and the input/output phase delay of the servo control device output from the frequency characteristic calculation unit and the coefficients of the initial filter or the coefficients of the filter after the start of the machine learning;
an action information output unit that outputs action information including adjustment information of the coefficients of the initial filter or the coefficients of the filter after the start of the machine learning, included in the state information;
a reward output unit that outputs a reward value of reinforcement learning based on the input/output gain and the input/output phase delay output from the state information acquisition unit; and
a value function updating unit that updates an action value function on the basis of the reward value output by the reward output unit, the state information, and the action information.

4. The machine learning system according to claim 3 , wherein
the frequency characteristic calculation unit outputs the input/output gain and the input/output phase delay, and
the reward output unit calculates a reward based on the input/output phase delay when the input/output gain of the servo control device is equal to or smaller than the input/output gain of an input/output gain standard model calculated from characteristics of the servo control device.

5. The machine learning system according to claim 4 , wherein
the input/output gain of the standard model is a constant value at a predetermined frequency or higher.

6. The machine learning system according to claim 4 , wherein
the reward output unit calculates the reward so that the input/output phase delay decreases.

7. The machine learning system according to claim 3 , further comprising:
an optimization action information output unit that outputs adjustment information of the coefficients on the basis of a value function updated by the value function updating unit.

8. A control device comprising:
the machine learning system according to claim 1 ; and
a servo control device that controls a motor and has at least one filter for attenuating a specific frequency component.

9. A machine learning method of a machine learning system that performs machine learning of optimizing coefficients of at least one filter for attenuating at least one specific frequency component, provided in a servo control device that controls a motor, the method comprising:
setting initial values of the coefficients of the filter so as to attenuate at least one specific frequency component;
calculating at least one of an input/output gain and an input/output phase delay of the servo control device on the basis of an input signal and an output signal of the servo control device, of which the frequencies change;
removing filter characteristics of an initial filter in which the initial values are set to the filter from at least one of the input/output gain and the input/output phase delay obtained on the basis of the input signal of the servo control device and the output signal of the servo control device obtained using the initial filter at the start of machine learning; and
starting machine learning of the coefficients of the filter in which the filter characteristics are removed so that at least the input/output gain decreases or the phase delay decreases.

Assignees

• Fanuc Corp

Inventors

• Maeda Kazuomi

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/006

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

• G06N3/088 Primary

  Non-supervised learning, e.g. competitive learning · CPC title

• G05B13/042

  in which a parameter or coefficient is automatically adjusted to optimise the performance · CPC title

• G05B13/0265 Primary

  the criterion being a learning criterion · CPC title