Abstract: Dock engineering in soft soil stratum faces challenges such as high uncertainty of soil parameters and large errors in traditional early warning due to complex geological conditions and long construction periods. To address these issues, this study proposes a dynamic early warning method for safety status based on multi-point time-series monitoring information fusion. First, a Bayesian network model is constructed, which utilizes the response relationships of network nodes and Monte Carlo Simulation (MCS) to calculate prior probabilities. Then, time-series monitoring data are introduced to update posterior probabilities through the Markov Chain Monte Carlo (MCMC) simulation, dynamically quantifying indicators such as the safety factor (F_s) and failure probability (P_f) to establish a direct correlation between monitoring information and safety status. Finally, based on the characteristics of soft soil and relevant specifications, dynamic early warning criteria incorporating 5 warning levels and 3 response states are established. Applying the method to a dock project in Shanghai, this method integrates multi-point time-series monitoring information for parameter inversion, effectively reducing the uncertainty of soil parameters (with the coefficient of variation decreased by 14%~30%). Meanwhile, it realizes time-sensitive safety evaluation, where the time-varying curve of the safety factor shows a regular decrease as deformation increases. Based on this, a hierarchical dynamic early warning standard is established, which determines that the dock is in Safety Level I during the construction period (F_s=1.945 > 1.50, P_f < 0.001%). Additionally, the feasibility of the early warning standard triggering alerts under hazardous conditions is verified through extreme working condition simulation. The proposed dynamic early warning system significantly improves the accuracy of parameter inversion and the accuracy and real-time performance of safety evaluation, effectively avoids the risk of false alarms from single-source data, enhances the precision and comprehensiveness of dynamic early-warning, and provides a scientific method for risk prevention and control in similar engineering projects.