US-12625488-B2 - Data extraction in industrial automation systems

US12625488B2US 12625488 B2US12625488 B2US 12625488B2US-12625488-B2

Abstract

A system and method for providing a configuration for data extraction from an automation system includes a signal selection agent configured to: receive a user selection of at least one module of the automation system; generate, for display to the user, a user interface identifying one or more selectable signals associated with the selected module and displaying one or more guidance elements comprising data mined from data sources pertaining to the automation system for guiding the user in the selection of relevant signals; receive a user selection of one or more of the selectable signals; and automatically generate the configuration for data extraction on the basis of the selected signals.

Inventors

Reuben Borrison
Julius Rueckert
Matthias Berning
Roland Braun

Assignees

ABB SCHWEIZ AG

Dates

Publication Date: 20260512
Application Date: 20230406
Priority Date: 20201007

Claims (13)

1 . A system for providing a configuration for data extraction from an automation system, the system comprising: at least one processor that executes instructions that are stored in a memory; a signal selection agent executed by at least one processor performs: receive a user selection of the at least one module of the automation system; generate, for display to the user, a user interface displaying one or more selectable signals associated with the selected module and displaying one or more tools enabling the user to select relevant signals associated with the selected module based on data mined from one or more data sources to the automation system; receive a user selection of one or more of the selectable relevant signals; automatically generate the configuration for data extraction on the basis of the selected relevant signals by translating the selected relevant signals to the configuration for data extraction, wherein the configuration uses concrete names of the at least one module and the selected relevant signals to provide a path in an address space exposed by the automation system, wherein the path is used to read the selected relevant signals associated with the selected module; wherein the signal selection agent comprises a domain knowledge mapping model comprising a machine learning model trained to map domain concepts based on the data mined from the one or more data sources to the selectable signals of a given module of the automation system, wherein the domain concepts provide classification data of the at least one module of the automation system.
2 . The system of claim 1 , wherein the domain knowledge mapping model is trained to map the selectable signals of the given module to the domain concepts described in knowledge graph comprising the data mined from the data sources.
3 . The system of claim 1 , wherein the signal selection agent comprises a documentation importer configured to mine the data from the data sources using one or more of (i) a natural language processing technique, (ii) a natural language understanding technique.
4 . The system of claim 1 , wherein the tools displayed by the user interface comprises a graphical interaction tool comprising one or more interactive graphical representations of parts of the automation system configured to enable one or more of (i) the user selection of the said module of the automation system, (ii) the user selection of the one or more selectable signals.
5 . The system of claim 1 , wherein the tools displayed by user interface comprises a textual interaction tool configured to provide one or more of (i) a search function, (ii) a filter function.
6 . The system of claim 1 , wherein the tools displayed by user interface comprises a template-based signal selection tool configured to display one or more further selectable signals associated with modules of the same type as the user-selected module.
7 . The system of claim 1 , wherein the tools displayed by user interface comprises a task-oriented signal selection tool configured to display one or more further selectable signals associated with a task specified by the user.
8 . The system of claim 1 , wherein the signal selection agent is configured to generate the user interface to display one or more further selectable signals associated with modules that are connected to the selected module in the automation system or which form part of a material flow with the user-selected module.
9 . The system of claim 1 , wherein the system is configured to provide to the machine learning model labeled training data specifying, as features, a plurality of signals exposed by the automation system and, as the target, labeled data mined from the data sources.
10 . The system of claim 9 , further configured to mine the data from the data sources optionally using one or more of (i) a natural language processing technique, (ii) a natural language understanding technique.
11 . The system of claim 9 , further configured to incorporate the data mined from the data sources into a knowledge graph.
12 . A method of providing a configuration for data extraction from at least one module of an automation system, the method comprising: receiving a user selection of at least one module of the automation system; generating, for display to the user, a user interface displaying one or more selectable signals associated with the selected module and displaying one or more enabling the user to select relevant signals associated with the selected at least one module based on data mined from data sources pertaining to the automation system; receiving a user selection of one or more of the selectable signals; and automatically generating the configuration for data extraction on the basis of the selected relevant signals by translating the selected relevant signals to the configuration for data extraction, wherein the configuration uses concrete names of the at least one module and the selected relevant signals to provide a path in an address space exposed by the automation system, wherein the path is used to read the selected relevant signals associated with the selected module; further comprising using a domain knowledge mapping model comprising a machine learning model trained to map domain concepts based on the data mined from the one or more data sources to the selectable signals of a given module of the automation system, wherein the domain concepts provide classification data of the at least one module of the automation system.
13 . The method of claim 12 , further comprising training the machine learning model of the domain knowledge mapping model by providing to the machine learning model labeled training data, specifying, as features, a plurality of signals exposed by the automation system and, as the target, labeled data mined from the data sources.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This patent application claims priority to International Patent Application No. PCT/EP2021/077408, filed on Oct. 5, 2021, and to European Patent Application No. 20200604.5, filed on Oct. 7, 2020, each of which is incorporated herein in its entirety by reference. FIELD OF THE DISCLOSURE The present disclosure relates to data extraction in industrial automation systems. BACKGROUND OF THE INVENTION Referring to FIG. 1, non-automation experts (e.g., data scientists) are commonly faced with the task of extracting control system data from a distributed control system (DCS) 100 providing plant automation. Although the data scientist is likely to have limited knowledge of the system 100, extracting the data is a mandatory step in enabling the data scientist to process the data using available software tools. To this end, the system 100 may provide a standardized interface 102 (e.g., OPC UA) which exposes data in the form of a browsable address space. However, manually identifying the relevant signals to export by browsing the address space is challenging, time-consuming or even infeasible for the data scientist due to the large size of the address space and the requisite knowledge of the system 100. Automation experts familiar with the details of the system may not always be available. Exporting all accessible data is infeasible due to the high load placed on the DCS, with most data being irrelevant for the analysis. BRIEF SUMMARY OF THE INVENTION There is therefore a need to facilitate data extraction in industrial automation systems with minimal involvement of the automation expert. The claimed subject-matter may provide for mapping of domain concepts and knowledge as well as documentation to signals, thus enabling signal identification based on engineering data, and supporting the data scientist in configuring data extraction from the DCS without the need to understand engineering details of the DCS. The claimed subject-matter may drastically reduce the time required to configure the data extraction and/or render the data extraction feasible in the first place. The claimed subject-matter may therefore represent an important enabler for easy “plug-and-play” configuration and provisioning of data collection mechanisms. The claimed guidance elements assist the user in performing a technical task in the form of searching for and retrieving relevant signals within the large address space exposed by the interface of the automation system by means of a guided human-machine interaction process. By “non-automation expert” is meant herein a person who is not an automation expert or process expert. The non-automation expert may also be referred to as an automation non-expert, a layperson in the field of automation, or simply as a user or operator. The term refers in particular to data scientists. The term “domain concept” refers to a generic automation entity such as a pump, sensor, field instrument, meter e.g., flow meter, motor, drive, controller, communications interface, operator workstation, or server. The entity may be a module or module type, in the case of a modular plant, or any other device, component or plant equipment forming part of an industrial automation system. The domain concept is generic in that it is not tied to any particular instance of the corresponding entity. A domain concept may thus be viewed as data providing one or more of a classification of an entity, its functions, typical use, and relation to other entities. By “domain” is meant herein the domain of industrial automation systems. The term “domain concept” may be replaced herein by “entity concept”. The term “domain knowledge” may thus be understood as a collection of domain concepts. By “properties” is meant functional and non-functional features/characteristics of a domain concept. A pump, for example, has a number of static features, such as a maximum load it can handle, its electric specification regarding power consumption, etc. In addition, it also has dynamic or operational features, such as the current speed of the pump, its temperature, etc. Some of the features provide information on the operational state of the equipment (e.g., running/off), whereas some might be relevant to triggering an action (e.g. turning the equipment on/off. Some features/characteristics may be mapped to read-only signals and others to read/write signals. The static features/characteristics may furthermore include information mined from external artifacts (as described further below). This could be information regarding the use of the equipment, links to documentation, etc. In the context of the present disclosure, the terms “module”, “device”, and “equipment” may be used interchangeably. The terms “signal” and “variable” may also be used interchangeably. In a further aspect, there is provided a documentation importer configured to mine data from one or more data sources pertaining to an automation system and to const