<p>To enhance flatness control intelligence in cold-rolled copper strip production, this study proposes an integrated digital-twin system architecture incorporating a mechanistic model for work roll multi-zone cooling. A multi-source sensing network captures real-time physical characteristics of the rolling process, with structured data transmission and storage via the Internet and cloud platforms. A three-dimensional interactive interface enables dynamic process parameter optimisation and real-time visualisation of rolling conditions. A co-modelling framework—combining mechanistic analysis and data-driven intelligence—simulates mill dynamics, enabling a digital twin with real-time, bidirectional interaction. The heat-flux-density distribution of multi-zone work roll cooling is modelled using Gaussian functions. By integrating mechanistic analysis, numerical simulation, and influence-coefficient methods, both a mechanistic simulation model and a rapid-calculation matrix model for flatness control are developed and embedded in the digital twin. The system was validated on a 600 mm six-high cold-rolling mill for copper strip. Real-time interaction between the digital twin and physical mill enabled deployment of an influence-matrix-based multi-zone cooling control strategy. This reduced root-mean-square flatness error by approximately 0.8 I and local flatness deviations by about 2.0 I, significantly improving control of higher-order and local flatness defects. The findings provide an engineering-ready paradigm for intelligent rolling of non-ferrous metal strip.</p>