From Digital Twin to Digital Cognition: A Multilevel Model of Algorithmic Intelligence in Smart Manufacturing Systems

Abstract

This study extends the dominant digital twin paradigm by introducing the concept of digital cognition to explain the evolving nature of smart manufacturing systems. While prior research largely conceptualizes digital twins as passive representations for monitoring and simulation, we argue that advanced manufacturing systems increasingly exhibit algorithmic intelligence through the integration of predictive models, real-time control, and adaptive feedback mechanisms. Drawing on an in-depth case study of a fully automated precision grinding system, this paper develops a multilevel model that explains how algorithmic intelligence emerges from the interaction of three interdependent layers: (1) a perceptual infrastructure enabled by high-resolution sensor networks and real-time data acquisition; (2) an algorithmic reasoning layer driven by machine learning models such as LSTM-based prediction and thermal deformation compensation algorithms; and (3) an autonomous actuation layer realized through high-speed synchronized control architectures. The findings show that the emergence of digital cognition transforms manufacturing systems from reactive optimization tools into proactive and self-adaptive agents, significantly enhancing precision, efficiency, and operational reliability. This transformation also reconfigures control structures by shifting decision-making authority from human operators to algorithmically mediated systems, raising important implications for governance and human–machine interaction.