Granted Patent B2

US 12,443,177 · App. 17/608,473 · Granted Oct 14, 2025

Inventors: Andreas Gienger (Kirchheim unter Teck, DE); Oliver Sawodny (Stuttgart, DE); Simon Alt (Ditzingen, DE); Martin Weickgenannt (Bietigheim-Bissingen, DE); Markus Hummel (Urbach, DE); Tobias Schlotterer (Hechingen, DE); Jens Berner (Möglingen, DE); Hauke Bensch (Lübeck, DE)

Assignee: Dürr Systems AG

G05B23/0254G05B2223/02

Abstract

Claims (80)

1. A method for anomaly and/or fault recognition in an industrial process, the method comprising:

identifying a process structure of the industrial process comprising determining a structure graph mapping relationships of components, in the industrial process;

wherein the components comprise sensors and/or actuators;

wherein the structure graph comprises a plurality of cliques describing physical relationships between the components, and wherein relationships between nodes of each are described by a probability density function of a plurality of probability density functions;

generating an anomaly and/or fault model of the industrial process, wherein the anomaly and/or fault model comprises the structure graph;

the anomaly and/or fault model comprising information on an occurrence probability of process values of the industrial process;

inputting the process values of the industrial process into the anomaly and/or fault model during operation of the industrial process,

using the probability density functions to recognize an anomaly and/or fault situation by determining the occurrence probability of the inputted process values of the industrial process and by checking the occurrence probability against a limit value; and

applying at least one corrective measure based on the recognized anomaly and/or fault situation and/or continuing operation of the industrial process.

2. The method according to claim 1 , wherein

the anomaly and/or fault model includes structural data containing information of the process structure of the industrial process, and/or

the anomaly and/or fault model includes parameterization data containing information on relationships between process values of the industrial process.

3. The method according to claim 1 , wherein, for generating the anomaly and/or fault model, one or more of the following steps is carried out:

determination of causalities in the identified process structure of the industrial process; or

structure parameterization of the relationships in the identified process structure of the industrial process.

4. The method according to claim 1 , wherein determination of the structure graph is performed using one or more of the following:

a machine learning method;

expert knowledge;

known circuit diagrams and/or procedure diagrams; or

designations in a numbering system of the industrial process.

5. The method according to claim 3 , wherein the industrial process is activated by test signals for structure identification for determining the structure graph.

6. The method according to claim 3 , wherein the determining of causalities in the identified process structure of the industrial process is performed using one or more of the following:

system input signals and system output signals that are generated on activation of the industrial process by test signals;

expert knowledge;

known circuit diagrams and/or procedure diagrams; or

designations in a numbering system of the industrial process.

7. The method according to claim 3 , wherein, for structure parameterization of the relationships in the identified process structure of the industrial process, one or more of the following is used:

models for determining probability density functions, wherein the models comprise Gaussian mixture models;

known physical relationships between process values; or

physical characteristic diagrams of functional elements of the industrial process.

8. The method according to claim 7 , wherein data from regular operation of the industrial process and/or data obtained by activation of the industrial process by test signals are used for structure parameterization using models for determining probability density functions, wherein the models comprise Gaussian mixture models.

9. The method according to claim 8 , wherein the data that are used for structure parameterization using models for determining probability density functions, wherein the models comprise Gaussian mixture models.

10. The method according to claim 1 , wherein during generation of the anomaly and/or fault model the limit value for the occurrence probability of the process value is established, and wherein an anomaly is recognized if the occurrence probability falls below the limit value.

11. The method according to claim 1 , wherein a fault cause of the recognized anomaly and/or a recognized fault situation is identified.

12. The method according to claim 1 , wherein the industrial process comprises one or more of the following treatment stations of a painting plant:

pre-treatment station;

station for cathodic dip coating;

drying stations;

industrial supply air plant; or

painting robot.

13. An anomaly and/or fault recognition system for recognizing an anomaly and/or fault, which is configured to perform the method for anomaly and/or fault recognition in an industrial process according to claim 1 , wherein the industrial process comprises a painting plant.

14. An industrial control system comprising:

a treatment station; and

an anomaly and/or fault recognition system according to claim 13 .

15. The method according to claim 1 , wherein, as a result of allocating a fault cause for a recognized anomaly and/or fault situation, affected ones of the cliques are expanded by adding at least one node and a probability density function of the recognized anomaly and/or fault situation is integrated into a relationship of the added at least one node.

16. A method for anomaly and/or fault recognition in an industrial process, the method comprising:

generating a structure graph by determining relationships between components of the industrial process;

wherein the components comprise sensors and/or actuators;

wherein the structure graph comprises a plurality of cliques describing physical relationships between the components, and wherein relationships between nodes of each are described by a probability density function of a plurality of probability density functions;

parametrizing the structure graph using a Gaussian mixture model (GMM) to determine the probability density function,

generating, an anomaly and/or fault model of the industrial process that includes information on an occurrence probability of process values of the industrial process;

wherein the anomaly and/or fault model comprises the parameterized structure graph;

inputting the process values of the industrial process into the anomaly and/or fault model during operation of the industrial process;

using the probability density functions to recognize an anomaly and/or fault situation by determining the occurrence probability of the inputted process values of the industrial process and by checking the occurrence probability against a limit value; and

applying at least one corrective measure based on the recognized anomaly and/or fault situation and/or continuing operation of the industrial process.

17. The method according to claim 16 , wherein

the anomaly and/or fault model includes structural data containing information on a process structure in the industrial process; and/or

the anomaly and/or fault model includes data containing information on relationships between process values of the industrial process.

18. The method according to claim 16 , wherein, for generating the anomaly and/or fault model, one or more of the following steps is carried out: structure identification for determining a process structure of the industrial process; determination of causalities in the determined process structure of the industrial process; or structure parameterization of the relationships in the determined process structure of the industrial process.

19. The method according to claim 18 , wherein, in a context of structure identification for determining a process structure of the industrial process, the structure graph maps relationships in the industrial process.

20. The method according to claim 19 , wherein determination of the process structure is performed using one or more of the following:

a machine learning method;

expert knowledge;

known circuit diagrams and/or procedure diagrams; or designations in a numbering system of the industrial process.

21. The method according to claim 18 , wherein the industrial process is activated by test signals for structure identification for determining the structure graph.

22. The method according to claim 18 , wherein the determining of causalities in the determined process structure of the industrial process is performed using one or more of the following:

system input signals and system output signals that are generated on activation of the industrial process by test signals;

expert knowledge;

known circuit diagrams and/or procedure diagrams; or designations in a numbering system of the industrial process.

23. The method according to claim 18 , wherein, for structure parameterization of the relationships in the determined process structure of the industrial process, one or more of the following is used: models for determining probability density functions, wherein the models comprise: Gaussian mixture models; known physical relationships between process values; or physical characteristic diagrams of functional elements of the industrial process.

24. The method according to claim 23 , wherein data from regular operation of the industrial process and/or data obtained by activation of the industrial process by test signals are used for structure parameterization using models for determining probability density functions, wherein the models comprise Gaussian mixture models.

25. The method according to claim 24 , wherein the data that are used for structure parameterization using models for determining probability density functions, wherein the models comprise Gaussian mixture models, are pre-processed before the structure parameterization.

26. The method according to claim 16 , wherein during generation of the anomaly and/or fault model the limit value is established, and wherein an anomaly is recognized if the occurrence probability falls below the limit value.

27. The method according to claim 16 , wherein a fault cause of a recognized anomaly and/or a recognized fault situation is identified.

28. The method according to claim 16 , wherein the industrial process includes or is formed by one or more of the following treatment stations of a painting plant:

pre-treatment station;

station for cathodic dip coating;

drying stations;

industrial supply air plant; or

a painting robot.

Priority Claims (2)

Continuity (1)

References Cited (210)