Method and system to generate control logic for performing industrial processes

Patent metadata

Field	Value
Publication number	US-12153406-B2
Application number	US-201917419514-A
Country	US
Kind code	B2
Filing date	Dec 20, 2019
Priority date	Dec 31, 2018
Publication date	Nov 26, 2024
Grant date	Nov 26, 2024

Abstract

The invention relates to a method and system to generate control logic for performing industrial processes with a controller in a process plant. The method includes receiving a control narrative comprising one or more control requirements of the industrial process, and extracting a plurality of control entities and a plurality of set points, from the control narrative using one or more sets of predetermined regular expressions and one or more models. The method further includes identifying a set of inputs, outputs and control elements from the plurality of control entities using a domain dictionary, detecting a plurality of actions from the control narrative using an intent classifier, identifying a relationship between the set of inputs, outputs and control elements, the plurality of set points, and the plurality of actions, and generating based on the relationship identified the control logic for the controller to perform the process.

First claim

We claim:

1. A method for generating a control logic for performing an industrial process with a controller in a process plant, the method comprising:
receiving, with a system associated with the controller in the process plant, a control narrative associated with the industrial process from a user, wherein the control narrative comprises one or more control requirements required for controlling the industrial process;
extracting, by the system, a plurality of control entities and a plurality of set points from the control narrative using one or more sets of predetermined regular expressions and one or more models stored in a database associated with the system, wherein each set point of the plurality of set points is associated with a control entity of the plurality of control entities, wherein the plurality of control entities control one or more parameters of the industrial process;
identifying, by the system, at least one of type or category for each of the extracted plurality of control entities, the type being identified using a database of control entities that includes a plurality of types of control entities associated with a plurality of industrial processes, the category being identified using a domain dictionary that defines a nomenclature for control entities applicable to a particular industry, wherein the category of the control entity is a set of inputs, outputs, and control elements;
detecting, by the system, a plurality of actions from the control narrative using an intent classifier, wherein the intent classifier identifies actions using keywords present in the control narrative;
identifying, by the system, a relationship between the set of inputs, outputs, and control elements, the plurality of set points, and the plurality of actions using one of coreference resolution or dependency parsing;
generating, by the system, the control logic for the controller to perform the industrial process, wherein the control logic is based on the relationship identified between the set of inputs, outputs, and control elements, the plurality of set points, and the plurality of actions; and
providing the control logic to the controller to be executed by the controller to perform the industrial process.

2. The method of claim 1 , wherein extracting the plurality of control entities and the plurality of set points comprises:
parsing the control narrative to obtain a plurality of expressions;
extracting a control entity from the plurality of expressions based on a first set of predetermined regular expressions and a first model, wherein the first set of predetermined regular expressions comprises a mapping between a set of control entities to a set of regular expressions, wherein the first model comprises a mapping between a set of annotations associated with the set of control entities to the set of regular expressions, wherein an annotation of a control entity is based on a domain of the industrial process; and
extracting a set point from the plurality of expressions based on a second set of predetermined regular expressions and a second model, wherein the second set of predetermined regular expressions comprises a mapping between a set of set points to another set of regular expressions, wherein the second model comprises a mapping between a set of annotations associated with the set of set points to the another set of regular expressions, wherein an annotation of a set point is based on the domain of the industrial process.

3. The method of claim 1 , wherein detecting the plurality of actions comprises:
detecting an anomaly within the plurality of actions based on a set of domain defined rules associated with a domain of the industrial process; and
updating the relationship with one or more of a missing information and a correct information associated with the anomaly, wherein the missing information is detected using natural language processing built on the domain of the industrial process.

4. The method of claim 1 , wherein generating the control logic comprises:
generating one or more control logic blocks, wherein each of the one or more control logic blocks is mapped to a control library module existing in a library of modules based on a semantic score, wherein each semantic score is obtained by comparing semantics of each control logic block to semantics of each control library module; and
generating the control logic in a machine-readable format based on the control library module mapped to each of the one or more control logic blocks.

5. A system for generating a control logic for performing a process with a controller of a process plant, wherein the system is communicatively coupled to the controller, the system comprising:
an input interface to receive a control narrative associated with the industrial processes from a user, wherein the control narrative comprises one or more control requirements required for controlling the industrial process; and
a processor to (i) extract a plurality of control entities and a plurality of set points from the control narrative using one or more sets of predetermined regular expressions and one or more models stored in a database associated with the system, wherein each set point of the plurality of set points is associated with a control entity of the plurality of control entities, wherein the plurality of control entities control one or more parameters of the industrial process; (ii) detect a plurality of actions from the control narrative using an intent classifier, wherein the intent classifier identifies actions using keywords present in the control narrative; (iii) identify at least one of type or category for each of the extracted plurality of control entities, the type being identified using a database of control entities that includes a plurality of types of control entities associated with a plurality of industrial process es, the category being identified using a domain dictionary that defines a nomenclature for control entities applicable to a particular industry, wherein the category of the control entity is a set of inputs, outputs, and control elements; (iv) identify a relationship between the set of inputs, outputs, and control elements, the plurality of set points, and the plurality of actions using one of coreference resolution or dependency parsing; and (v) generate the control logic for the controller to perform the industrial process, wherein the control logic is based on the relationship identified between the set of inputs, outputs, and control elements, the plurality of set points, and the plurality of actions; and
an output interface to communicate the control logic to the controller over a communication network to be executed by the controller for performing the industrial process.

6. The system of claim 5 , wherein the processor is further configured to:
build a first model from a set of documents comprising a set of annotations associated with a set of control entities, wherein the first model comprises a mapping between the set of annotations to a set of regular expressions, wherein an annotation of a control entity is based on a domain of the industrial process; and
build a second model associated with a set of set points, wherein the second model comprises a mapping between a set of annotations associated with the set of set points to the another set of regular expressions, wherein an annotation of a set point is based on the domain of the industrial process.

Assignees

• Abb Schweiz Ag

Inventors

• Jetley Raoul
• Sharma Divyasheel
• Peedikayil Abdulla Puthan
• Schulz Dirk
• Ramu Vadthyavath

Classifications

• G05B2219/31449

  Monitor workflow, to optimize business, industrial processes · CPC title

• G05B2219/13023

  Convert natural language, graphic to coded states, input · CPC title

• G05B19/41865 Primary

  characterised by job scheduling, process planning, material flow · CPC title

• G05B19/0426 Primary

  Programming the control sequence · CPC title