<p>Due to the frequent occurrence of geological disasters worldwide, which pose a serious threat to infrastructure such as pipelines, accurately assessing the stress state of pipelines has become an urgent issue to be addressed. To improve the real-time and prediction accuracy of underground pipeline stress monitoring in the context of geohazards, the study constructs a dynamic closed-loop pipeline stress state reconstruction and inversion framework based on digital twin technology. The mechanical state of the physical pipeline is mapped in real time by the digital twin, the numerical simulation and multi-source monitoring data are integrated, and the parameters of the twin model are dynamically optimized by combining the optimization algorithms of Particle Swarm Optimization (PSO) and Support Vector Machine (SVM), so as to realize the real-time prediction of the pipeline stress state and the dynamic updating of the disaster scenario. The experiment showed that the numerical simulation model could accurately simulate the stress-strain response of pipelines under geological hazards, which was basically consistent with the actual monitoring data. The accuracy of the inversion model was 95.14%, which was an average improvement of 11.06% compared to other models. The calculation time was 5.62&#xa0;s, which was an average reduction of 18.95%. Under the three geological disasters of earthquakes, mudslides, and landslides, the root mean square error of the model’s predictions was below 3 GPa, and the accuracy remained above 94%. The results indicate that the research model has high prediction accuracy and efficiency and can effectively handle the problem of predicting pipeline stress states under different geological disasters, providing a reliable method for evaluating pipeline stress states in geological disasters.</p>