Machine learning device, CNC device and machine learning method for detecting indication of occurrence of chatter in tool for machine tool

• US 10,496,055 B2
• Filed: 12/13/2017
• Issued: 12/03/2019
• Est. Priority Date: 12/14/2016
• Status: Active Grant

• Current Assignee: Fanuc Corporation
• Original Assignee: Fanuc Corporation
• Inventors: Matsushima, Takahiko, Miyazaki, Yoshito
• Primary Examiner(s): Bahta, Kidest
• Application Number:
  US15/840,211
• Publication Number:
  US 20180164757A1
• Time in Patent Office: 720 Days
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G05B 13/042 in which a parameter or coefficient is automatically adjusted to optimise the performance
• CPC Class Codes: G05B 19/406 characterised by monitoring or safety G05B19/19 takes precedence
• CPC Class Codes: G05B 2219/33296 ANN for diagnostic, monitoring
• CPC Class Codes: G05B 2219/37433 Detected by acoustic emission, microphone
• CPC Class Codes: G05B 2219/37434 Measuring vibration of machine or workpiece or tool
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/088 Non-supervised learning, e.g. competitive learning

Abstract

A machine learning device for detecting an indication of an occurrence of chatter in a tool for a machine tool, includes a state observation unit which observes at least one state variable of a vibration of the machine tool itself, a vibration of a building in which the machine tool is installed, an audible sound, an acoustic emission and a motor control current value of the machine tool, in addition to a vibration of the tool; and a learning unit which generates a learning model based on the state variable observed by the state observation unit.

Claims

1. A machine learning device for detecting an indication of an occurrence of chatter in a tool for a machine tool, comprising:

a state observation unit which observes state variables including a vibration of the machine tool itself, at least one of a vibration of a building in which the machine tool is installed, an audible sound, and an acoustic emission, and a motor control current value of the machine tool; and

a learning unit which generates a learning model based on the state variables observed by the state observation unit, wherein

the learning unit

generates the learning model by performing unsupervised learning based on the state variables during normal operation in which no chatter occurs in a specific machining block, and

generates and outputs a normal score during the normal operation in which no chatter occurs in the specific machining block, and an abnormal score when there is an indication of the occurrence of chatter in the machining block; and

the machine learning device further comprises;

an output utilization unit which determines whether a score based on the state variables of the machining block corresponds to the normal score or the abnormal score, in order to detect an indication of the occurrence of chatter in the tool for the machine tool.

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

3. The machine learning device according to claim 1, wherein

the machine tool includes:

a vibration sensor which detects the vibration of the machine tool itself and provided in a holder or bit of the tool; and

at least one of an audible sound sensor which detects the audible sound, and an acoustic emission sensor which detects the acoustic emission.

4. The machine learning device according to claim 1, wherein the machine tool includes a vibration sensor which detects the vibration of the building in which the machine tool is installed.

5. The machine learning device according to claim 1, wherein the machine tool includes a current sensor which detects a motor control current value of the machine tool.

6. The machine learning device according to claim 5, wherein the current sensor is provided in a motor amplifier for driving a motor of the machine tool.

7. A CNC device, comprising:

a learning circuit which constitutes a machine learning device for detecting an indication of an occurrence of chatter in a tool for a machine tool, wherein the machine learning device includes

a state observation unit which observes state variables including a vibration of the machine tool itself, at least one of a vibration of a building in which the machine tool is installed, an audible sound, and an acoustic emission, and a motor control current value of the machine tool; and

a learning unit which generates a learning model based on the state variables observed by the state observation unit, wherein the learning unit

generates the learning model by performing unsupervised learning based on the state variables during normal operation in which no chatter occurs in a specific machining block, and

generates and outputs a normal score during the normal operation in which no chatter occurs in the specific machining block, and an abnormal score when there is an indication of the occurrence of chatter in the machining block; and

an output utilization unit which determines whether a score based on the state variables of the machining block corresponds to the normal score or the abnormal score, in order to detect an indication of the occurrence of chatter in the tool for the machine tool, and

wherein the CNC device controls the machine tool.

8. The CNC device according to claim 7, further comprising:

a determination circuit which compares a score outputted from the learning circuit with a certain determination reference value to make a determination; and

a CPU which outputs a stop signal to the machine tool based on a determination result from the determination circuit.

9. The CNC device according to claim 8, wherein the CPU outputs a warning signal to a host management system based on the determination result from the determination circuit.

10. A machine learning method for detecting an indication of an occurrence of chatter in a tool for a machine tool, comprising:

observing state variables including a vibration of the machine tool itself, at least one of a vibration of a building in which the machine tool is installed, an audible sound and an acoustic emission, and a motor control current value of the machine tool; and

generating a learning model by unsupervised learning based on the observed state variables, wherein

the generating the learning model includes

generating the learning model by performing unsupervised learning based on the state variables during normal operation in which no chatter occurs in a specific machining block, and

generating and outputting a normal score during the normal operation in which no chatter occurs in the specific machining block, and an abnormal score when there is an indication of the occurrence of chatter in the machining block; and

the machine learning method further comprises;

determining whether a score based on the state variables of the machining block corresponds to the normal score or the abnormal score, in order to detect an indication of the occurrence of chatter in the tool for the machine tool.

11. The machine learning device according to claim 1, wherein the output utilization unit is configured to compare the score based on the state variables of the machining block with a plurality of the normal scores generated by the learning unit to determine whether the score is abnormal.