ML-Based Control Device Replacement for Unknown Automation Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Updating or replacing control devices in automated systems is challenging when the exact functionality of the existing control device is unknown, leading to time-consuming manual reprogramming and compatibility issues with newer versions.
Innovation Solution
A method using a second control device with a computer-implemented mapping algorithm, such as a neural network or machine learning algorithm, to adapt and generate output data similar to the original control device, allowing for autonomous adaptation and replacement without knowing the exact functionality of the first control device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual reprogramming is performed to replace the control device, then the functionality can be updated, but the time consumption and complexity increase significantly
Solution Approach 1:
The patent creates a digital twin (virtual model) of the control device that replicates its functionality. This virtual model can be updated and trained independently without affecting the physical control device, enabling rapid functionality updates without manual reprogramming of the actual controller.
Solution Approach 2:
The system performs preliminary training of the virtual model using historical data before deployment. The mapping algorithm is pre-adapted to replicate the control device's behavior, so when updates are needed, the virtual model can be retrained offline and the updates applied without time-consuming on-site reprogramming.
2Adaptability or versatility
If the control device is replaced with a newer version, then future improvements can be implemented, but compatibility issues arise when the programming language is not compatible
Solution Approach 1:
The virtual model acts as an intermediary layer between the control device and the newer programming environment. It translates and adapts the control logic, allowing compatibility between different programming languages and versions without direct integration challenges.
Solution Approach 2:
The patent replaces the traditional mechanical/procedural programming approach with a data-driven machine learning model. Instead of manually coding compatibility layers, the system uses trained neural networks or other ML algorithms to automatically handle the translation and adaptation of control logic across different platforms.
3Ease of manufacture
If the exact functionality of the control device is unknown, then replacement becomes difficult, but detailed analysis of the control device increases time and resource requirements
Solution Approach 1:
The system enables the control device to effectively reprogram itself through the virtual model. By collecting and analyzing the control device's own operational data, the virtual model learns to replicate its functionality autonomously without requiring external experts to perform detailed functional analysis.
Solution Approach 2:
The system implements feedback loops where the virtual model continuously monitors the control device's input-output behavior and adjusts its mapping algorithm accordingly. This automated feedback mechanism eliminates the need for manual functionality analysis and enables automatic adaptation to the control device's actual behavior.
Data Source
AI summary
A method for operating an automated system, the system comprising:a controlled device for performing an action as a function of received control data;a first control device for receiving system data and generating control data for controlling the controlled device as a function of the received system data; anda second control device for receiving input data and generating output data as a function of the input data according to a computer-implemented mapping algorithm;wherein the method comprises:adapting the computer-implemented mapping algorithm such that the second control device, upon receiving the system data as input data generates output data that is similar to the control data generated by the first control device with a predetermined similarity degree, wherein the computer-implemented mapping algorithm includes a neural network algorithm and/or a machine learning algorithm.