Machine-Learned PLC Replacement for Legacy Control Migration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Legacy programmable logic controllers (PLCs) pose challenges due to the lack of spare parts, proprietary protocols, and the loss of engineering skills, making it difficult and costly to replace them without incurring significant hardware and software engineering costs, and resulting in production downtime and vendor lock-in.
Innovation Solution
The implementation of machine learning techniques, such as deep neural networks and deep reinforcement learning, to train a replacement PLC (ML-PLC) to learn and imitate the behavior of the legacy PLC, allowing it to adapt and improve the automation environment without requiring extensive programming or dedicated resources, and enabling virtualization for efficient replacement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional PLC replacement methods are used, then hardware and software engineering costs are incurred, but replacement complexity and time increase
Solution Approach 1:
The patent creates a digital twin of the legacy PLC by training a neural network to replicate its behavior. The system observes input-output pairs from the legacy PLC and trains a model to reproduce its control logic, effectively copying its functionality without physical replacement. This eliminates the need for complex hardware/software engineering while maintaining operational reliability.
Solution Approach 2:
The patent replaces the mechanical/physical PLC system with a software-based neural network model. Instead of physically replacing the PLC hardware and reprogramming it, the system uses machine learning to create a virtual replacement that replicates the original PLC's behavior through trained neural networks, substituting physical replacement with computational modeling.
2Adaptability or versatility
If legacy PLCs are kept in operation, then vendor lock-in and lack of spare parts occur, but replacement causes production downtime
Solution Approach 1:
The patent performs preliminary training of the neural network model using historical data from the legacy PLC before actual replacement is needed. By pre-training the model with accumulated input-output pairs, the system prepares the replacement in advance, allowing for seamless substitution without production downtime. The training process can occur during normal operation using recorded data.
Solution Approach 2:
The system creates a functional copy of the legacy PLC's control logic through neural network training, enabling the replacement to immediately assume the same role without requiring physical compatibility or vendor-specific hardware. This digital copying approach breaks vendor lock-in while maintaining operational continuity.
3Measurement precision
If machine learning training data is collected from legacy PLC, then replacement accuracy improves, but data collection time increases
Solution Approach 1:
The patent implements continuous data collection during normal PLC operation, utilizing the existing operational data stream without interrupting the control process. The system continuously records input-output pairs from the legacy PLC, transforming otherwise idle data into training material. This approach eliminates dedicated data collection time while maintaining continuous operational accuracy improvement.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Over the past several decades, rapid advances in semiconductors, automation, and control systems have resulted in the adoption of programmable logic controllers (PLCs) in an immense variety of environments. Machine learning techniques help train replacement PLCs when a legacy PLC must be replaced, e.g., due to aging or failure. The techniques facilitate the efficient adoption and correct operation of replacement PLCs in the industrial environment.