Abstract The cross-section deformation is one of the major factors affecting the service safety of the shield tunnel. However, the cross-section deformation prediction method with reasonable consideration on multi-source uncertainty is still lacking, and the influence of reinforcement corrosion is neglected. In this paper, the segment bending degradation model is built considering the corrosion effect, and then a cross-section deformation model is built using uniform rigidity ring method based on a Bayesian network .The monitoring data and Markov Chain Monte Carlo simulation method are used to update the model. The study shows that the model prediction after Bayesian updating is closer to the monitoring data, the model prediction in 2013 after updating is nearly the same as the monitoring data,and there has only 1.6% error between the model prediction in 2015 after updating and the monitoring data. In contrast, the prediction errors for 2013 and 2015 are 21.1% and 21.9% respectively. It is obvious that the model prediction of Bayesian network is revised after fusing the monitoring data, making it closer to the monitoring data and the model prediction accuracy is improved greatly.
Abstract：
The cross-section deformation is one of the major factors affecting the service safety of the shield tunnel. However, the cross-section deformation prediction method with reasonable consideration on multi-source uncertainty is still lacking, and the influence of reinforcement corrosion is neglected. In this paper, the segment bending degradation model is built considering the corrosion effect, and then a cross-section deformation model is built using uniform rigidity ring method based on a Bayesian network .The monitoring data and Markov Chain Monte Carlo simulation method are used to update the model. The study shows that the model prediction after Bayesian updating is closer to the monitoring data, the model prediction in 2013 after updating is nearly the same as the monitoring data,and there has only 1.6% error between the model prediction in 2015 after updating and the monitoring data. In contrast, the prediction errors for 2013 and 2015 are 21.1% and 21.9% respectively. It is obvious that the model prediction of Bayesian network is revised after fusing the monitoring data, making it closer to the monitoring data and the model prediction accuracy is improved greatly.
2022, Vol. 59 