AI Function Block Integration for IEC 61499 Machine Controllers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The integration of artificial intelligence into industrial control systems faces significant hurdles due to architectural variations and incompatibilities, requiring considerable effort and expertise, which prolongs production downtime and delays in constructing new machines.
Innovation Solution
A method that automates the integration of AI function blocks into industrial machines using an IEC 61499 runtime environment, allowing operators to select and link AI models with minimal clicks, without requiring knowledge of software libraries or communication configurations, and executes them on suitable computing devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If artificial intelligence is integrated into industrial control systems using standard methods, then AI functionality can be achieved, but the integration requires considerable effort and expertise due to architectural variations and incompatibilities
Solution Approach 1:
The patent creates a universal integration layer that standardizes AI model execution across different architectures. The runtime environment and containerization approach enable a single interface to support multiple AI frameworks (TensorFlow, PyTorch, ONNX, etc.), making the system universally compatible with various AI models without requiring separate integration processes for each architecture.
Solution Approach 2:
The patent introduces an intermediary layer consisting of the runtime environment and containerization infrastructure that mediates between the control system and diverse AI models. This intermediary handles architecture-specific details, allowing operators to integrate AI models without directly dealing with architectural variations and incompatibilities.
2Adaptability or versatility
If artificial intelligence is integrated into industrial control systems, then advanced control capabilities are achieved, but the integration process takes considerable time and expertise
Solution Approach 1:
The patent performs preliminary actions by pre-configuring runtime environments and containerization infrastructure before actual AI model deployment. The system pre-establishes execution contexts for different AI frameworks, so when an operator selects an AI model, the integration process is already prepared and can proceed rapidly without time-consuming setup steps.
Solution Approach 2:
The runtime environment and containerization system automatically handle integration tasks without requiring operator expertise. The system self-configures execution parameters, manages resource allocation, and handles compatibility adjustments automatically, reducing integration time from hours or days to minutes while eliminating the need for specialized AI integration knowledge.
3Ease of operation
If operators manually configure AI integration with full control over software libraries and communication interfaces, then precise control is achieved, but the process becomes complex and time-consuming
Solution Approach 1:
The patent extracts complex configuration details (software library selection, communication interface setup, execution parameter tuning) from the operator's workflow and relocates them to the runtime environment's automatic configuration system. Operators only need to select high-level AI model parameters, while the extracted detailed configuration is handled automatically by the system.
Solution Approach 2:
The patent uses containerization to create standardized, pre-configured execution templates for different AI frameworks. Instead of requiring operators to configure each AI model from scratch, the system copies proven configuration templates and adapts them automatically, reducing configuration complexity while maintaining precise control over execution parameters.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for integrating at least one AI function block, comprising artificial intelligence, into a controller for an industrial machine. In this method, an AI model is selected for execution within the AI function block, the AI function block is at least partially automatically linked to other function blocks of the industrial machine's control software, and the AI function block is executed at least partially automatically, wherein the execution of the AI function block is carried out using an IEC 61499 runtime environment.