Abstract: To address the drawbacks of fixed-threshold early warning for tunnelling parameters in conventional shield construction, including frequent false and missing alarms, excessive reliance on empirical judgment for parameter optimization, and poor adaptability to dynamic stratum variations, this paper proposes a dynamic early warning and multi-objective robust optimization method for shield tunnelling parameters incorporating quantitative stratum variability. First, a Markov chain is adopted to invert stratum random fields and establish dynamic early-warning thresholds driven by stratum variation. Second, stratum uncertainty is embedded into the NSGA-Ⅱ multi-objective optimization framework as robust constraints to obtain optimal tunnelling parameters, thereby forming a closed-loop linkage mechanism of "parameter early warning – parameter optimization". Finally, a management system based on the B/S architecture is developed. Field engineering applications verify that compared with conventional approaches, the proposed method reduces the false alarm rate by 36.1%, achieves a parameter recommendation accuracy of 83% and a 7% improvement in tunnelling efficiency, with the system response time controlled within 3 seconds. The developed technique can satisfy the requirements of real-time decision-making at shield construction sites.