Machine learning method and machine learning device for learning fault conditions, and fault prediction device and fault prediction system including the machine learning device

• US 10,317,853 B2
• Filed: 07/27/2016
• Issued: 06/11/2019
• Est. Priority Date: 07/31/2015
• Status: Active Grant

• Current Assignee: Fanuc Corporation, Preferred Networks, Inc.
• Original Assignee: Fanuc Corporation, Preferred Networks, Inc.
• Inventors: Inagaki, Shougo, Nakagawa, Hiroshi, Okanohara, Daisuke, Okuta, Ryosuke, Matsumoto, Eiichi, Kawaai, Keigo
• Primary Examiner(s): Tran, Vincent H
• Application Number:
  US15/220,925
• Publication Number:
  US 20170031329A1
• Time in Patent Office: 1,049 Days
• CPC Class Codes: B25J 9/163 learning, adaptive, model based, rule based expert control
• CPC Class Codes: B25J 9/1674 characterised by safety, monitoring, diagnostic
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G05B 15/02 electric
• CPC Class Codes: G05B 19/4063 Monitoring general control system G05B19/4062 takes precedence
• CPC Class Codes: G05B 19/4065 Monitoring tool breakage, life or condition
• CPC Class Codes: G05B 2219/31359 Object oriented model for fault, quality control
• CPC Class Codes: G05B 2219/33321 Observation learning
• CPC Class Codes: G05B 2219/34477 Fault prediction, analyzing signal trends
• CPC Class Codes: G05B 2219/37214 Detect failed machine component, machine performance degradation
• CPC Class Codes: G05B 2219/50185 Monitoring, detect failures, control of efficiency of machine, tool life
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 5/048 Fuzzy inferencing
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
• CPC Class Codes: Y10S 901/47 Optical

Abstract

A fault prediction system includes a machine learning device that learns conditions associated with a fault of an industrial machine. The machine learning device includes a state observation unit that, while the industrial machine is in operation or at rest, observes a state variable including, e.g., data output from a sensor, internal data of control software, or computational data obtained based on these data, a determination data obtaining unit that obtains determination data used to determine whether a fault has occurred in the industrial machine or the degree of fault, and a learning unit that learns the conditions associated with the fault of the industrial machine in accordance with a training data set generated based on a combination of the state variable and the determination data.

Claims

1. A machine learning device configured to learn a condition associated with a fault of an industrial machine, the machine learning device comprising:

a memory; and

at least one processor coupled to the memory, the at least one processor being configured to;

while the industrial machine is in operation or at rest, obtain a state variable including at least one of

output data from a sensor configured to detect a state of one of the industrial machine and a surrounding environment,

internal data of control software controlling the industrial machine, and

computational data obtained based on one of the output data and the internal data;

obtain determination data indicating a fault of the industrial machine, and

learn the condition associated with the fault of the industrial machine in accordance with a training data set including a combination of the state variable and the determination data, wherein

the fault of the industrial machine is predicted in accordance with the learned condition, and

the at least one processor is further configured to update the condition by weighting each of the determination data comprised in the training data set, in accordance with a length of time from when each of the determination data is obtained until the fault actually occurs.

2. The machine learning device according to claim 1, wherein the at least one processor is further configured to learn the condition in accordance with the training data set generated for each of a plurality of industrial machines.

3. The machine learning device according to claim 1, wherein the at least one processor is further configured to learn a normal state only in a predetermined period and thereafter detect occurrence of the fault of the industrial machine.

4. A fault prediction device, comprising the machine learning device according to claim 1 and configured to predict the fault of the industrial machine, wherein

the at least one processor is further configured to output fault information indicating one of whether the fault has occurred in the industrial machine and a degree of the fault, in response to input of a current state variable of the state variable, based on a result of learning by the at least one processor in accordance with the training data set.

5. The fault prediction device according to claim 4, wherein the at least one processor is further configured to learn a condition again in accordance with an additional training data set generated based on a combination of the current state variable and the determination data.

6. The fault prediction device according to claim 4, wherein

the machine learning device is connected to the industrial machine via a network, and

the at least one processor is configured to obtain the current state variable via the network.

7. The fault prediction device according to claim 6, wherein the machine learning device is located on a cloud server.

8. The fault prediction device according to claim 4, wherein the machine learning device is mounted in a controller that controls the industrial machine.

9. The fault prediction device according to claim 4, wherein the industrial machine comprises a plurality of industrial machines, and the learning result obtained by the machine learning device is shared by the plurality of industrial machines.

10. A fault prediction system, comprising:

the fault prediction device according to claim 4; and

the sensor configured to detect the state of said one of the industrial machine and the surrounding environment, and output the output data, wherein the at least one processor is further configured to notify an operator of the fault information.

11. The fault prediction system according to claim 10, wherein a time at which the at least one processor notifies the operator of the fault information satisfies at least one of

precedence to a time defined by a first predetermined period preceding a time at which a fault occurs, and

subsequence to a time defined by a second predetermined period preceding the time at which a fault occurs.

12. A machine learning method of learning a condition associated with a fault of an industrial machine, the method comprising:

while the industrial machine is in operation or at rest, obtaining a state variable including at least one of

output data from a sensor configured to detect a state of one of the industrial machine and a surrounding environment,

internal data of control software controlling the industrial machine, and

computational data obtained based on one of the output data and the internal data;

obtaining determination data indicating a fault of the industrial machine; and

learning the condition associated with the fault of the industrial machine in accordance with a training data set including a combination of the state variable and the determination data, wherein

the fault of the industrial machine is predicted in accordance with the learned condition, and

the learning the condition associated with the fault of the industrial machine further comprises;

updating the condition by weighting each of the determination data comprised in the training data set, in accordance with a length of time from when each of the determination data is obtained until the fault actually occurs.

13. The machine learning method according to claim 12, wherein the learning the condition associated with the fault of the industrial machine further comprises learning the condition in accordance with the training data set generated for each of a plurality of industrial machines.

14. The machine learning method according to claim 12, wherein the learning the condition associated with the fault of the industrial machine further comprises learning a normal state only in a predetermined period and thereafter detecting occurrence of the fault of the industrial machine.

15. A fault prediction method, comprising:

the machine learning method according to claim 12; and

predicting the fault of the industrial machine, wherein

the predicting the fault of the industrial machine comprises outputting fault information indicating one of whether the fault has occurred in the industrial machine and a degree of the fault, in response to input of a current state variable of the state variable, based on a result of learning by the learning the condition associated with the fault of the industrial machine.

16. The fault prediction method according to claim 15, wherein the learning the condition associated with the fault of the industrial machine further comprises learning a condition again in accordance with an additional training data set generated based on a combination of the current state variable and the determination data.

17. The fault prediction method according to claim 15, wherein

the machine learning device is connected to the industrial machine via a network, and

the obtaining the state variable comprises obtaining the current state variable via the network.