Frameworks for Continuous Learning and Knowledge Updating in Dynamic AI Systems: Foundations, Methods, and Applications in AI-Enabled Intelligent Engineering Systems

Abstract

The dynamic environment in intelligent engineering system needs an AI framework that can learn and update its knowledge at any time, so that it can run normally even when the situation changes. Based on system engineering theory, network physical system (CPS) principle and information theory, this review discusses the framework to help adaptive AI combine physical processes with computational intelligence. An important theoretical basis says that you must adapt to feedback and measurement uncertainty to manage changing data. Methods include continuous learning strategy, concept change detection, intensive learning for dynamic optimization, gradually adaptive computer vision, and rapidly updated edge calculation. We show practical application cases in intelligent manufacturing and intelligent infrastructure, and find that it improves the operation intensity and the accuracy of prediction. The key analysis shows that there are great benefits, such as less downtime and more system autonomy, but there are also problems, such as disastrous forgetting, a lot of calculation, and the risk of instability in important cycle safety. An initial idea is to use a boot-based trigger mechanism in the CPS feedback architecture to better choose when to update and how to use resources. The conclusion of the review is that we need to create a standard and reliable adaptive framework so that the evolution of artificial intelligence can meet the needs of engineering systems throughout their life cycle.