Machine learning apparatus, laser machining system and machine learning method

Patent metadata

Field	Value
Publication number	US-10664767-B2
Application number	US-201715460850-A
Country	US
Kind code	B2
Filing date	Mar 16, 2017
Priority date	Mar 17, 2016
Publication date	May 26, 2020
Grant date	May 26, 2020

Abstract

A machine learning apparatus that learns laser machining condition data of a laser machining system includes: a state amount observation unit that observes a state amount of the laser machining system; an operation result acquisition unit that acquires a machined result of the laser machining system; a learning unit that receives an output from the state amount observation unit and an output from the operation result acquisition unit, and learns the laser machining condition data in association with the state amount and the machined result of the laser machining system; and a decision-making unit that outputs laser machining condition data by referring to the laser machining condition data learned by the learning unit.

First claim

The invention claimed is:

1. A machine learning apparatus for learning laser processing condition data of a laser processing system,
the laser processing system comprising:
at least one laser apparatus that includes at least one laser oscillator;
at least one processing head that emits a laser light from the laser apparatus to a workpiece;
at least one output light detection unit that detects an amount of the laser light emitted from the processing head;
at least one reflected light detection unit that detects a reflected light emitted from the processing head and reflected on a surface or near the surface of the workpiece to return to the laser apparatus via an optical system in the processing head;
at least one processing result observation unit that observes, without human intervention, at least one of a processing state and a processing result of the workpiece at least during laser processing or after the laser processing; and
at least one driving apparatus that changes a relative positional relationship between the processing head and the workpiece,
the machine learning apparatus comprising:
a state amount observation unit that observes a state amount of the laser processing system;
an operation result acquisition unit that acquires a processing result of the laser processing system;
learning unit that receives an output from the state amount observation unit and an output from the operation result acquisition unit, and learns the laser processing condition data in association with the state amount of the laser processing system and the processing result; and
a decision-making unit that outputs laser processing condition data by referring to the laser processing condition data leaned by the learning unit; wherein
when during processing of the workpiece based on given laser processing condition data, an amount of the reflected light detected by the reflected light detection unit exceeds a second predetermined level set lower than a first predetermined level set lower than an alarm level indicating that at least one of the processing head, the laser apparatus, and a laser light propagation optical component between the processing head and the laser apparatus may be damaged by the reflected light,
the learning unit refers to the learned laser processing condition data to output laser processing condition data enabling a processing result close to a processing result of the given laser processing condition data to be predicted without the amount of the reflected light detected by the reflected light detection unit exceeding the second predetermined level; and
the learning unit includes:
a learning model for learning laser processing condition data varied from one laser processing content to another;
an error calculation unit that calculates a difference between a laser processing result including a processing speed obtained by the operation result acquisition unit or time expended for predetermined processing and a roughly ideal processing result including a processing speed set for each laser processing content or time expended for predetermined processing or a target processing result; and
a learning model update unit that updates the learning model according to the difference.

2. The machine learning apparatus according to claim 1 , wherein the state amount observed by the state amount observation unit includes at least one of the followings: light output characteristics of the laser apparatus indicating a relationship between a light output command for the laser apparatus and a light output actually emitted from the laser apparatus; a light output emitted from the laser apparatus; a ratio of a light output emitted from the processing head to the light output from the laser apparatus; a temperature of a portion thermally connected to the laser oscillator; temperatures of portions including a component changed in temperature due to laser oscillation in the laser apparatus; a temperature of the processing head; a temperature of the optical system that propagates the laser light from the laser apparatus to the processing head; a temperature of the driving apparatus; a temperature of a structural component that supports the processing head or the driving apparatus; a type, a temperature, and a flow rate of fluids for cooling the component increased in temperature due to the laser oscillation; a temperature and humidity of air in the laser apparatus; an environmental temperature and humidity around the laser apparatus; an actual current of a driving motor of the driving apparatus; an output from a position detection unit of the driving apparatus; and sizes including a thickness, material quality, specific heat, a density, heat conductivity, a temperature, and a surface state of the workpiece.

3. The machine learning apparatus according to claim 1 , wherein the laser processing condition data output from the decision-making unit includes at least one of the followings: a light output, a light output waveform, beam mode, and a laser wavelength of the laser light emitted from each laser apparatus; a focal distance, an F-value, and a transmittance of the optical system that emits the laser light; a relative positional relationship including a time change between a focus of the laser light emitted to the workpiece and a processed surface of the workpiece; a spot size, a power density, and a power density distribution of the laser light emitted to the workpiece on the processed surface of the workpiece; a relative positional relationship including a time change between the processing head and the workpiece; an angle formed between an optical axis of the laser light and the processed surface of the workpiece; a processing speed; and a type and a flow rate or supply pressure of assist gas.

4. A machine learning apparatus for learning laser processing condition data of a laser processing system,
the laser processing system comprising:
at least one laser apparatus that includes at least one laser oscillator;
at least one processing head that emits a laser light from the laser apparatus to a workpiece;
at least one output light detection unit that detects an amount of the laser light emitted from the processing head;
at least one reflected light detection unit that detects a reflected light emitted from the processing head and reflected on a surface or near the surface of the workpiece to return to the laser apparatus via an optical system in the processing head;
at least one processing result observation unit that observes, without human intervention, at least one of a processing state and a processing result of the workpiece at least during laser processing or after the laser processing; and
at least one driving apparatus that changes a relative positional relationship between the processing head and the workpiece,
the machine learning apparatus comprising:
a state amount observation unit that observes a state amount of the laser processing system;
an operation result acquisition unit that acquires a processing result of the laser processing system;
a learning unit that receives an output from the state amount observation unit and an output from the operation result acquisition unit, and learns the laser processing condition data in association with the state amount of the laser processing system and the processing result; and
a decision-making unit that outputs laser processing condition data by referring to the laser processing condition data leaned by the learning unit; wherein
when during processing of the workpiece based on given laser processing condition data, an amount of the reflected light detected by the reflected light detection unit exceeds a second predetermined level set lower than a first predetermined level set lower than an alarm level indicating that at least one of the processing head, the laser apparat

Assignees

• Fanuc Corp

Inventors

• Takigawa Hiroshi
• Ohyama Akinori

Classifications

• B23K26/21

  by welding · CPC title

• G05B2219/45041

  Laser cutting · CPC title

• B23K26/702

  Auxiliary equipment · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06N20/00 Primary

  Machine learning · CPC title