ML Control Schemes for Real-Time Industrial Process Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Industrial manufacturing environments face challenges in extracting enterprise-level insights from vast data sets and limited ability to adjust control programs in real-time due to the complexity and resource-intensive nature of operational analytics, making it difficult to leverage operational data effectively.

Innovation Solution

Integration of machine learning models into industrial control code to optimize parameters such as performance, yield, and energy conservation, allowing for automatic adjustments and improvements in industrial processes through pre-packaged machine learning models that can be easily integrated and managed within existing control systems.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If traditional operational analytics are used to extract enterprise-level insights from industrial data, then analysis depth can be achieved, but computing power requirements and time consumption become excessively high

Engineering Contradiction:
Improveenterprise-level insightsVSAvoidcomputing power
Core Design Contradiction:
Loss of informationVSUse of energy by moving object

Solution Approach 1:

The patent segments the control program into multiple versions, each optimized for different parameters (e.g., one version optimizes performance, another optimizes yield). This allows the system to switch between pre-segmented control strategies based on current operational goals, avoiding the need to perform comprehensive operational analytics for every decision while still achieving optimized outcomes in specific domains.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If control programs are manually edited to adjust parameters based on operational data, then customization and optimization are possible, but the process becomes extremely difficult and time-consuming

Engineering Contradiction:
Improvecontrol program adjustmentVSAvoidediting time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-compiling multiple versions of the control program before runtime, with each version optimized for specific parameters. During operation, the system can immediately switch between these pre-prepared versions without requiring time-consuming manual editing, thus achieving rapid adaptability while eliminating the time loss associated with real-time program modification.

Inventive Principle:
Principle #10Preliminary action

3Ease of operation

If control programs are made easily adjustable during runtime, then real-time optimization is possible, but the complexity of managing and switching between different control strategies increases

Engineering Contradiction:
Improvereal-time adjustmentVSAvoidcontrol program management
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent implements a universal control program structure that can execute multiple specialized versions through a single interface. The control program is designed to accept different parameter sets and switch between optimized versions based on operational context, providing ease of real-time adjustment while managing complexity through a unified, multi-functional architecture rather than requiring separate management systems for each control strategy.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4155841A1Machine learning models for asset optimization within industrial automation environments
Publication Date: 2023.03.29 ROCKWELL AUTOMATION TECH INC
  • EP4155841A1 patent drawingFigure 1
  • EP4155841A1 patent drawingFigure 2A~2D
  • EP4155841A1 patent drawingFigure 3

AI summary

Various embodiments of the present technology generally relate to solutions for integrating machine learning models into industrial automation environments. More specifically, embodiments of the present technology include systems and methods for implementing machine learning models within industrial control code to improve performance, increase productivity, and add capability to existing control programs. In an embodiment, a system comprises: a storage component configured to maintain a set of model control schemes for controlling an industrial process, a control component configured to control the industrial process with a control program running a model control scheme, wherein the model control scheme is configured to optimize a first parameter of the industrial process, and a model management component configured to change the model control scheme to optimize a second parameter of the industrial process that is distinct from the first parameter.