US-12625478-B2 - Industrial controller having AI-enabled multicore architecture

US12625478B2US 12625478 B2US12625478 B2US 12625478B2US-12625478-B2

Abstract

An industrial controller supports deterministic execution of control programs (e.g., ladder logic, function block diagrams, structured text, or other such control code) and is also capable of executing non-deterministic execution cycles, including mathematical optimization algorithms—in which a systematic search in a solution space is performed to identify a desired solution—or machine learning algorithms, either of which can be used by the controller to dynamically update the deterministic control program or code based on current or predicted states of the automation system being controlled by the controller.

Inventors

Bijan Sayyarrodsari
Dan Li

Assignees

ROCKWELL AUTOMATION TECHNOLOGIES, INC.

Dates

Publication Date: 20260512
Application Date: 20230601

Claims (20)

1 . An industrial controller, comprising: a memory that stores executable components; and a processor, operatively coupled to the memory, that executes the executable components, the executable components comprising: a program execution component configured to execute an industrial control program that causes the industrial controller to monitor and control an industrial automation system in accordance with a current control strategy; and an artificial intelligence (AI) engine component configured to execute a first non-deterministic search algorithm that performs a global search for a first replacement control strategy determined, based on analysis of a current operating condition of the industrial automation system, to improve one or more control performance metrics relative to the current control strategy, in response to determining that a change in the current operating condition renders the current control strategy infeasible and that the first replacement control strategy output by the first non-deterministic search algorithm is feasible for the current operating condition, replace the current control strategy with the first replacement control strategy, and in response to determining that the change in the current operating condition renders the current control strategy infeasible and that the first replacement control strategy is not feasible for the current operating condition, replace the current control strategy with a second replacement control strategy that is output by a second non-deterministic search algorithm that performs a feasible path search for the second replacement control strategy.
2 . The industrial controller of claim 1 , wherein the industrial control program executes on a first processor core of the industrial controller and at least one of the first non-deterministic search algorithm or the second non-deterministic search algorithm executes on a second processor core of the industrial controller that is separate from the first processor core.
3 . The industrial controller of claim 1 , wherein the AI engine component is configured to train, using data collected from the industrial automation system, a machine learning model used by the first non-deterministic search algorithm or the second non-deterministic search algorithm to analyze the current operating condition of the industrial automation system and determine the first replacement control strategy or the second replacement control strategy.
4 . The industrial controller of claim 3 , wherein the AI engine component is configured to add contextual information to the data collected from the industrial automation system based on first-principles information stored on the industrial controller to yield contextualized data, and to train the machine learning model using the contextualized data.
5 . The industrial controller of claim 4 , wherein the first-principles information is stored on an in-memory database of the industrial controller.
6 . The industrial controller of claim 4 , wherein the first-principles information comprises at least one of information regarding how a material or ingredient is processed during a phase of a manufacturing process carried out by the industrial automation system, a maximum or minimum bound on a properties of the manufacturing process, a causal relationship between events within the manufacturing process, a mathematical relationship between measured operating metrics of the manufacturing process, identification of data items known to be relevant to the one or more control performance metrics, or mathematical relationships between the data items and the one or more control performance metrics.
7 . The industrial controller of claim 3 , wherein the machine learning model uses parameterized symbolic learning.
8 . The industrial controller of claim 1 , wherein the industrial control program and at least one of the first non-deterministic search algorithm or the second non-deterministic search algorithm share results of their execution via an in-memory database of the industrial controller.
9 . The industrial controller of claim 1 , wherein the one or more control performance metrics comprise at least one of product quality, product throughput, energy consumption by the industrial automation system, emissions from the industrial automation system, or risk of downtime to a machine of the industrial automation system.
10 . The industrial controller of claim 1 , wherein the AI engine component is further configured to modify the industrial control program in accordance with the first replacement control strategy or the second replacement control strategy.
11 . A method, comprising: executing, by an industrial controller comprising a processor, an industrial control program that causes the industrial controller to monitor and control an industrial automation system in accordance with a current control strategy; executing, by the industrial controller, a first non-deterministic search algorithm that generates, based on analysis of a current operating condition of the industrial automation system, a first replacement control strategy predicted to improve one or more control performance metrics relative to the current control strategy, wherein the first non-deterministic search algorithm performs a global search for the first replacement control strategy; in response to determining that a change in the current operating condition renders the current control strategy infeasible and that the first replacement control strategy generated by the first non-deterministic search algorithm is feasible for the current operating condition, replacing, by the industrial controller, the current control strategy with the first replacement control strategy; and in response to determining that the change in the current operating condition renders the current control strategy infeasible and that the first replacement control strategy is not feasible for the current operating condition, replacing, by the industrial controller, the current control strategy with a second replacement control strategy that is generated by a second non-deterministic search algorithm that performs a feasible path search for the second replacement control strategy.
12 . The method of claim 11 , wherein the executing of the industrial control program and executing of the first non-deterministic search algorithm are performed on separate processor cores of the industrial controller.
13 . The method of claim 11 , further comprising training, by the industrial controller using data collected from the industrial automation system, a machine learning model used by the first non-deterministic search algorithm and the second non-deterministic search algorithm to analyze the current operating condition of the industrial automation system and determine the first replacement control strategy and the second replacement control strategy.
14 . The method of claim 13 , wherein the training comprises: adding, by the industrial controller, contextual information to the data collected from the industrial automation system based on first-principles information stored on the industrial controller to yield contextualized data; and training, by the industrial controller, the machine learning model using the contextualized data.
15 . The method of claim 13 , wherein the first-principles information is stored on an in-memory database of the industrial controller.
16 . The method of claim 11 , wherein the replacing of the current control strategy with the first replacement control strategy comprises modifying the industrial control program in accordance with the first replacement control strategy, and the replacing of the current control strategy with the second replacement control strategy comprises modifying the industrial control program in accordance with the second replacement control strategy.
17 . A non-transitory computer-readable medium having stored thereon instructions that, in response to execution, cause an industrial controller comprising a processor to perform operations, the operations comprising: executing an industrial control program that causes the industrial controller to monitor and control an industrial automation system in accordance with a first current control strategy; executing a first non-deterministic search algorithm that determines, based on analysis of a current state of the industrial automation system, a first replacement control strategy estimated to improve one or more control performance metrics relative to the current control strategy, wherein the first non-deterministic search algorithm determines the first replacement control strategy using a global search; in response to determining that a change in the current operating condition renders the current control strategy infeasible and that the first replacement control strategy generated by the first non-deterministic search algorithm is feasible for the current operating condition, replacing the current control strategy with the first replacement control strategy; and in response to determining that the change in the current operating condition renders the current control strategy infeasible and that the first replacement control strategy is not feasible for the current operating condition, replacing the current control strategy with a second replacement control strategy that is determined by a second non-deterministic search algorithm, wherein the second non-deterministic search algorithm determines the second replacement control strategy using a feasible path search.
18 . The non-transitory computer-readable medium of claim 17 , wherein the replacing of the current control strategy with the first replacement control strategy comprises modifying the industrial control program in accordance with the first replacement control strategy, and the replacing of the current control strategy with the second replacement control strategy comprises modifying the industrial control program in accordance with the second replacement control strategy.
19 . The non-transitory computer-readable medium of claim 17 , wherein the executing of the industrial control program and executing of the first non-deterministic search algorithm are performed on separate processor cores of the industrial controller.
20 . The non-transitory computer-readable medium of claim 17 , the operations further comprising training, using data collected from the industrial automation system, a machine learning model used by the first non-deterministic search algorithm and the second non-deterministic search algorithm to analyze the current operating condition of the industrial automation system and determine the first replacement control strategy and the second replacement control strategy.

Description

TECHNICAL FIELD The subject matter disclosed herein relates generally to industrial automation systems, and, in particular, industrial controllers that monitor and control industrial automation systems or processes. BACKGROUND ART Since their adoption as critical components of industrial automation systems, industrial controllers have been designed to execute pre-developed deterministic control code (e.g., ladder logic, structured text, function block diagrams, etc.) whose programmatic structures and parameters remain fixed unless rewritten by a developer, and which generate predictable control outputs for a given set of inputs representing the current states of the controlled industrial machine or process. During runtime, the industrial controller executes this control code deterministically at a high frequency execution cycle to ensure that appropriate control commands are timely issued by the controller in response to changes to the states of the automation system. BRIEF DESCRIPTION The following presents a simplified summary in order to provide a basic understanding of some aspects described herein. This summary is not an extensive overview nor is it intended to identify key/critical elements or to delineate the scope of the various aspects described herein. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is presented later. In one or more embodiments, an industrial controller is provided, comprising a program execution component configured to execute an industrial control program that causes the industrial controller to monitor and control an industrial automation system in accordance with a first control strategy; and an artificial intelligence (AI) engine component configured to execute a non-deterministic search algorithm that determines, based on analysis of a current operating condition of the automation system, a second control strategy that improves one or more control performance metrics relative to the first control strategy, and to modify the industrial control program in accordance with the second control strategy. Also, one or more embodiments provide a method, comprising executing, by an industrial controller comprising a processor, an industrial control program that causes the industrial controller to monitor and control an industrial automation system in accordance with a first control strategy; executing, by the industrial controller, a non-deterministic search algorithm that generates, based on analysis of a current operating condition of the automation system, a second control strategy predicted to improve one or more control performance metrics relative to the first control strategy; and in response to generation of the second control strategy predicted to improve the one or more control performance metrics, modifying, by the industrial controller, the industrial control program in accordance with the second control strategy. Also, according to one or more embodiments, a non-transitory computer-readable medium is provided having stored thereon instructions that, in response to execution, cause an industrial controller comprising a processor to perform operations, the operations comprising executing an industrial control program that causes the industrial controller to monitor and control an industrial automation system in accordance with a first control strategy; executing a non-deterministic search algorithm that determines, based on analysis of a current state of the automation system, a second control strategy estimated to improve one or more control performance metrics relative to the first control strategy; and in response to determination of the second control strategy predicted to improve the one or more control performance metrics, updating the industrial control program in accordance with the second control strategy. To the accomplishment of the foregoing and related ends, certain illustrative aspects are described herein in connection with the following description and the annexed drawings. These aspects are indicative of various ways which can be practiced, all of which are intended to be covered herein. Other advantages and novel features may become apparent from the following detailed description when considered in conjunction with the drawings. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of an example industrial control environment. FIG. 2 is a block diagram of an example industrial controller that supports non-deterministic execution cycles. FIG. 3 is a diagram illustrating operation of the AI-enabled industrial controller during runtime. FIG. 4 is a diagram illustrating a general approach that can be carried out by embodiments of an AI engine component of the AI-enabled industrial controller for determining a suitable control strategy for controlling an industrial automation system or process. FIG. 5 is a diagram illustrating an example control model evaluation and replacement approac