Machine-tool-state determination system and machine-tool-state determination method

Patent metadata

Field	Value
Publication number	US-10990085-B2
Application number	US-201916447611-A
Country	US
Kind code	B2
Filing date	Jun 20, 2019
Priority date	Jul 18, 2018
Publication date	Apr 27, 2021
Grant date	Apr 27, 2021

Abstract

A machine-tool-state determination system configured to determine a state associated with a machine tool including a rotation mechanism for processing a member, the system including: a sensor configured to acquire a state value from the machine tool; and an analysis device, in which the analysis device: performs spectral analysis with time series data of the state value, to extract a rotational frequency of the rotation mechanism and a harmonic wave to the rotational frequency; calculates a ratio of an amplitude of the rotational frequency to an amplitude of the harmonic wave; generates feature-amount data including the state value and the ratio as feature amounts; performs clustering with the feature-amount data; and determines a state associated with the machine tool, based on a result of the clustering.

First claim

What is claimed is:

1. A machine-tool-state determination system configured to determine a state associated with a machine tool including a rotation mechanism for processing a member, the machine-tool-state determination system comprising: a sensor configured to acquire a state value from the machine tool; a cutting mechanism that cuts the member in cooperation with an operation of the rotation mechanism and has a plurality of blades for cutting the member; and an analysis device including a processor and a memory connected to the processor, wherein the analysis device: performs spectral analysis with time series data of the state value, to extract a rotational frequency of the rotation mechanism and a harmonic wave to the rotational frequency; selects, from the harmonic wave, such a harmonic wave as a difference between the selected harmonic wave and a frequency obtained by multiplying the rotational frequency by the number of blades included in the cutting mechanism is smallest; calculates a ratio of an amplitude of the rotational frequency and an amplitude of the selected harmonic wave; generates feature-amount data including the state value and the ratio as feature amounts; performs principal component analysis with the feature-amount data to select a plurality of principal components; converts the feature-amount data in a feature amount space including axes corresponding to the feature amounts, into data in a feature amount space including axes corresponding to the selected principal components; performs clustering with the converted feature-amount data; determines a state associated with the machine tool, based on a result of the clustering; retains, as learning information, a result of clustering with learning data indicating a normal state and learning data indicating an abnormal state; and compares a result of clustering on feature-amount data newly generated, with the learning information, to detect an abnormality of the machine tool, and wherein the axes corresponding to the plurality of principal components include an axis defined by a linear combination of the axes corresponding to the feature amounts, with the axis corresponding to the ratio being large in weight factor.

2. The machine-tool-state determination system according to claim 1 , wherein the analysis device:
calculates an indicator for evaluating process accuracy of the member, based on a result of the clustering; and
evaluates the process accuracy of the member, based on the indicator.

3. The machine-tool-state determination system according to claim 1 , wherein the analysis device:
calculates, based on a result of the clustering, an evaluation value for determining whether a parameter for controlling the machine tool is suitable; and
determines, based on the evaluation value, whether the parameter is suitable.

4. The machine-tool-state determination system according to claim 1 ,
wherein the sensor is installed directly at the rotation mechanism of the machine tool or at a constituent different from the rotation mechanism of the machine tool, or is installed in non-contact with the machine tool, and
the machine-tool-state determination system includes, as the sensor, at least any of a force sensor, a strain sensor, a displacement sensor, a velocimeter, an accelerometer, an angular velocimeter, an acoustic sensor, an ultrasonic sensor, a microphone, a temperature sensor, a laser sensor, and a camera.

5. The machine-tool-state determination system according to claim 1 ,
wherein the machine tool is housed in a housing, and
part of the machine tool or part of the housing includes an acoustic absorption material.

6. A machine-tool-state determination method to be executed by a system that manages a machine tool having a rotation mechanism for processing a member and a cutting mechanism that has a plurality of blades that cuts the member in cooperation with an operation of the rotation mechanism, the system including a sensor configured to acquire a state value from the machine tool, and an analysis device including a processor and a memory connected to the processor, the machine-tool-state determination method comprising: performing spectral analysis with time series data of the state value, to extract a rotational frequency of the rotation mechanism and a harmonic wave to the rotational frequency, with the analysis device; selecting, from the harmonic wave, such a harmonic wave as a difference between the selected harmonic wave and a frequency obtained by multiplying the rotational frequency by the number of blades included in the cutting mechanism is smallest, with the analysis device; calculating a ratio of an amplitude of the rotational frequency to an amplitude of the selected harmonic wave, with the analysis device; generating feature-amount data including the state value and the ratio as feature amounts, with the analysis device; performing principal component analysis with the feature-amount data to select a plurality of principal components, with the analysis device; converting the feature-amount data in a feature amount space including axes corresponding to the feature amounts, into data in a feature amount space including axes corresponding to the selected principal components, with the analysis device; performing clustering with the converted feature-amount data, with the analysis device; determining a state associated with the machine tool, based on a result of the clustering, with the analysis device; retaining, as learning information, a result of clustering with learning data indicating a normal state and learning data indicating an abnormal state; and comparing a result of clustering on feature-amount data newly generated, with the learning information, to detect an abnormality of the machine tool, with the analysis device, and wherein the axes corresponding to the plurality of principal components include an axis defined by a linear combination of the axes, with the axis corresponding to the ratio being large in weight factor.

Assignees

• Hitachi Ltd

Inventors

• Otsu Kenji
• Watanabe Keiji
• Matsumoto Hisanori

Classifications

• G05B19/406 Primary

  characterised by monitoring or safety (G05B19/19 takes precedence) · CPC title

• G05B19/4184 Primary

  characterised by fault tolerance, reliability of production system · CPC title

• G06F18/23

  Clustering techniques · CPC title

• G06F2218/08

  Feature extraction · CPC title

• G06V2201/06

  Recognition of objects for industrial automation · CPC title