Abstract: Roboticized intelligent reconstruction of cosmetic production lines faces severe challenges such as high risk, long cycle, performance fluctuations, and complex decision-making. To address these challenges, this study proposes and validates a digital twin-driven systematic progressive reconstruction framework integrated with a comprehensive risk control strategy. This framework constructs comprehensive system representation and dynamic process models (using "physical-logic" dual-dimension representation and time-attribute Petri net based methods), implements progressive reconstruction centered on iterative optimization, virtual verification, and dynamic adjustment, and systematically integrates a multi-level risk control mechanism based on Failure Mode and Effects Analysis (FMEA) to achieve quantitative risk assessment and proactive intervention. Semi-physical simulation validation (N = 20 independent trials) on a cosmetic production line case demonstrates that compared to traditional one-time reconstruction, the proposed Digital Twin-driven Progressive Strategy (DTPS) significantly reduces the reconstruction risk index (from 0.70 ± 0.05 to 0.45 ± 0.03), system fluctuation (from 38.3% ± 3.7% to 25.1% ± 2.2%), and total reconstruction time (from 72.5 ± 2.8 hours shortened to 50.3 ± 1.9 hours). Simultaneously, it improves the post-reconstruction production line balance rate to 87.00% (original 72.00%), reduces labor demand to 39 people (original 47 people), increases capacity to 64 bottles/min (original 57 bottles/min), and achieves an investment recovery period of approximately 30-32 months. The research indicates that the framework proposed in this paper provides a more comprehensive, risk-controllable solution for roboticized system reconstruction, effectively balancing multiple objectives.