Local Supervisory Controller Redundancy via Machine Learning Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for handling loss of connectivity between advanced control processes on cloud platforms and local control infrastructure in industrial automation are costly and inefficient, as they require redundant network access points and secondary communication channels.

Innovation Solution

A system that includes a remote supervisory controller hosted on a computing platform and a local supervisory controller with a configured local process model, where the local controller provides intermediate set points to regulatory controllers during communication failures, using a configuration server to determine functional correlations and configure the local model for continued process regulation until connectivity is restored.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If redundant network access points or secondary communication channels are provided to handle loss of connectivity, then reliability is improved, but device complexity and cost increase

Engineering Contradiction:
Improveconnectivity reliabilityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent creates a simplified copy of the advanced process control application's functionality by training a machine learning model on historical process data and control actions. This local model copy enables the regulatory controller to generate appropriate control outputs without requiring actual connectivity to the advanced control application, thus maintaining reliability while avoiding the complexity of redundant communication infrastructure.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The regulatory controller is enhanced with self-service capability through the integrated machine learning model, allowing it to independently determine control outputs based on current process conditions when connectivity is lost. This eliminates the need for external redundant control systems or communication channels, reducing overall system complexity while maintaining autonomous operation.

Inventive Principle:
Principle #25Self-service

2Reliability

If redundant network access points or secondary communication channels are provided, then reliability is improved, but cost increases

Engineering Contradiction:
Improveconnectivity reliabilityVSAvoidhardware resources
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

Instead of deploying additional physical hardware such as redundant network access points or secondary communication channels, the patent creates a virtual copy of control functionality through a machine learning model that runs on the existing regulatory controller's processor, eliminating the need for extra hardware resources.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical/physical approach of adding redundant communication infrastructure with an information-processing approach using machine learning algorithms. The control functionality is substituted from physical redundant systems to computational models, reducing hardware requirements while maintaining reliability.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Manufacturing precision

If advanced process control applications are shifted to cloud platform, then manufacturing precision is improved, but loss of information occurs during connectivity failure

Engineering Contradiction:
Improvecontrol precisionVSAvoidcontrol information
Core Design Contradiction:
Manufacturing precisionVSLoss of information

Solution Approach 1:

The patent performs preliminary action by training the machine learning model offline using historical process data and control actions from the advanced process control application. This pre-trained model stores the knowledge and control logic in advance, enabling the regulatory controller to continue operating with high precision even when connectivity to the cloud platform is lost, preventing loss of control information.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The machine learning model acts as an intermediary between the cloud-based advanced process control application and the local regulatory controller. It receives and processes control information locally, maintaining the control loop's integrity and preventing information loss during connectivity failures while preserving the precision benefits of advanced control.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3161562B1A method for controlling a process plant using a redundant local supervisory controller
Publication Date: 2020.05.20 ABB (SCHWEIZ) AG
  • EP3161562B1 patent drawingFigure 1
  • EP3161562B1 patent drawingFigure 2

AI summary

The present invention discloses a method for controlling a process using a plurality of regulatory controllers (160) connected to a remote supervisory controller (120) over a communication link (155) and to a local supervisory controller (130) over a process bus (170). The remote supervisory controller (120) controls the regulatory controllers (160) over the communication link (155) with the use of a remote process model. During operation of the plant a local process model is configured in the local supervisory controller (130) with control data send on the communication link (155). Upon failure of the communication link (155) the local supervisory controller (130) takes over the control through the process bus (170), acting as a redundant or a backup controller. The method comprises: monitoring transmission of control data between the remote supervisory controller (120) and the plurality of controllers (160) over the communication link (155); determining functional correlations between measured process variables and control process variables, based on the control data and a predetermined correlation function; configuring local process model based on the determined functional correlations; and providing intermediate set points using the configured local process model, upon an intermediate failure of the communication link (155), where the communication link recovers from the intermediate failure within a predetermined time period.