Artificial Intelligence Systems at the Edge Supporting Programmable Logic Controller Control Based on Link Learning and Methods to Support Programmable Logic Controller Control Based on Link Learning

Abstract

The invention relates to an edge AI system supporting the control of a programmable logic controller based on linkage learning, and a method for supporting the control of a programmable logic controller based on linkage learning, particularly suitable for upgrading existing manufacturing plants. The system under the invention is equipped with an additional sensor cluster or set of sensors, e.g., vibration, temperature, current, sound, and pressure sensors, installed non-invasively into the existing actuator or group of existing actuators of the plant. The edge AI device incorporates an AI model to receive input data from the sensor cluster or set of sensors and produce output data containing control support information for the corresponding programmable logic controller.Through an industrial multi-protocol communication module acting as an intermediary, control support information for safety verification can be written to a separate memory area of ​​the programmable logic controller according to the advisor architecture. The safety advisor module verifies the control support information before it is transferred to the programmable logic controller. The linkage learning module allows multiple plants to jointly improve the artificial intelligence model without sharing raw production data. In addition, the system integrates a data drift detection module to distinguish anomalies from data drift, and an enhancement agent module to select adaptive actions under safety constraints. As a result, the system is capable of supporting real-time fault detection, maintaining model performance during operation, and contributing to reducing unplanned downtime.

Images