Robot system for anomaly detection

Patent metadata

Field	Value
Publication number	US-12576530-B2
Application number	US-202218682581-A
Country	US
Kind code	B2
Filing date	Aug 10, 2022
Priority date	Aug 13, 2021
Publication date	Mar 17, 2026
Grant date	Mar 17, 2026

Abstract

The invention relates to a robot system for detection of an operation anomaly. The robot system comprises: an industrial robot; a robot controller configured to control operation of said industrial robot; a robot operation program which is executable by said robot controller to operate said industrial robot according to a robot operation cycle; respective program nodes integrated in said robot operation program, wherein each of said respective program nodes is associated with a separate operational element of said robot operation cycle; wherein said robot controller is configured to obtain reference data based on operation parameters associated with execution of said robot operation program; and an anomaly detection block which for at least one of said respective program nodes is configured to evaluate an operation anomaly of said robot operation parameters relative to a representation of said reference data.

First claim

The invention claimed is:

1 . A system comprising:
an industrial robot; a controller configured to control operation of the industrial robot; memory storing an operation program that is executable by the controller to operate the industrial robot according to a cycle, the cycle comprising operations to be performed by the industrial robot; wherein the operation program comprises program nodes, each of the program nodes being associated with a respective operation; wherein the controller is configured to obtain reference data corresponding to one or more parameters associated with execution of the operation program, wherein the controller is configured to obtain the reference data by processing one or more values of the one or more parameters obtained over previous cycles of the industrial robot; and an anomaly detector configured to evaluate, based on the reference data, one or more values of the one or more of the parameters obtained during the cycle to detect an anomaly associated with operation of the industrial robot.

2 . The system of claim 1 , wherein each of the program nodes is associated with a respective node condition set, each node condition set specifying a condition associated with a respective node and being based on the reference data.

3 . The system of claim 2 , wherein the anomaly detector is configured to compare a condition associated with a node to the one or more parameters to detect the anomaly.

4 . The system of claim 1 , wherein the controller is configured to obtain the reference data during a training session of the industrial robot.

5 . The system of claim 1 , wherein two or more sets of the reference data are obtained based on a single execution of the operation program; and
wherein at least one of the two or more sets of reference data is associated with abnormal values of one or more of the parameters.

6 . The system of claim 1 , wherein the anomaly detector is configured to implement a machine learning process to detect the anomaly, the machine learning process comprising one of: supervised machine learning, unsupervised machine learning, semi-supervised machine learning, or reinforcement learning.

7 . The system of claim 1 , wherein the controller is configured to display a user interface, the user interface showing a program node at which the anomaly occurred.

8 . The system of claim 1 , wherein the one or more parameters are based on internal operation parameters of the industrial robot.

9 . The system of claim 8 , wherein the internal operation parameters comprises a timestamp.

10 . The system of claim 8 , wherein the internal operation parameters comprise a time associated with a program node.

11 . The system of claim 1 , wherein at least one parameter of at least one of the program nodes is exempt from evaluation performed by the anomaly detector.

12 . The system of claim 1 , wherein at least some of the program nodes form a decision tree comprising branches; and
wherein different ones of the program nodes are located on different branches of the decision tree.

13 . The system of claim 1 , wherein the program nodes comprise one or more hybrid nodes, wherein each of the one or more hybrid nodes relates to at least two operations to be performed by the industrial robot, the at least two operations comprising: movement, tool activation, waiting, machine tending, or measurement.

14 . The system of claim 2 , wherein at least one node condition set comprises a first node condition set and a second node condition set;
wherein at least one node condition in the first node condition set is different from at least one node condition in the second node condition set; wherein the anomaly detector is configured to compare the at least one node condition in the first node condition set with one or more values of the one or more of the parameters wherein the anomaly detector is configured to compare at least one node condition in the second node condition set with one or more values of the one or more of the parameters; and wherein a first program node associated with the first node condition set and a second program node associated with the second node condition set are different.

15 . The system of claim 14 , wherein the at least one node condition in the first node condition set or in the second node condition set comprises a node-specific parameter threshold.

16 . The system of claim 14 , wherein the at least one node condition in the first node condition set or in the second node condition set is based on node-specific reference data; and
wherein the node-specific reference data is based, at least partly, on previous normal operation of the industrial robot.

17 . The system of claim 14 , wherein the at least one node condition in the first node condition set or in the second node condition set is based on node-specific anomaly data; and
wherein the node-specific anomaly data is based, at least partly, on previous abnormal operation of the industrial robot.

18 . The system of claim 14 , wherein the at least one node condition in the first node condition set or in the second node condition set is based on node-specific machine learning.

19 . The system of claim 15 , wherein the node-specific parameter threshold of a node that is a waiting state is based on a time threshold.

20 . A method for detecting an anomaly in operation of an industrial robot, the method comprising:
executing an operation program on a controller to control operation of the industrial robot according to a cycle, the cycle comprising operations to be performed by the industrial robot, the operation program comprising program nodes, each of the program nodes being associated with an operation; obtaining reference data based on parameters associated with execution of the operation program, wherein the reference data is obtained by processing values of the one or more parameters obtained over previous cycles of the industrial robot; and using an anomaly detector to evaluate, based on the reference data, one or more values of the one or more of the parameters obtained during the cycle to detect the anomaly.

21 . The method of claim 20 , wherein the industrial robot is part of a robot system comprising the controller, the anomaly detector, and memory storing the operation program.

22 . The method of claim 20 , further comprising:
obtaining node condition sets, wherein each of the node condition sets specifies at least one node condition associated with a respective program node; and the anomaly detector comparing, for each respective program node, at least one node condition with the one or more values of the one or more of the parameters; wherein the anomaly is detected based on the comparing.

23 . The method of claim 20 , wherein the anomaly is an exogenous operation anomaly.

24 . The method of claim 20 , further comprising:
training a neural network based on the reference data to obtain at least one of the node conditions.

25 . The method of claim 24 , wherein the neural network is at least a part of an autoencoder; and
wherein the autoencoder comprises an input layer, an output layer, and at least one hidden layer connecting the input layer and the output layer.

26 . The method of claim 22 , further comprising:
modifying at least one node condition based on at least one value of one of the parameters.

Assignees

• Universal Robots As

Inventors

• Graabæk Søren Goddiksen
• Fugl Andreas Rune
• Ancker Emil Vincent

Classifications

• B25J9/1661

  characterised by task planning, object-oriented languages · CPC title

• G06N20/00

  Machine learning · CPC title

• G05B23/024

  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 title

• G05B2219/42271

  Monitor parameters, conditions servo for maintenance, lubrication, repair purposes · CPC title

• B25J9/1674 Primary

  characterised by safety, monitoring, diagnostic · CPC title