Research On Performance Deterioration And Safety Of Bridge Components Using Dynamic Bayesian Networks

Bridges have an important position in the transportation networks.However,bridges will be subjected to complex natural environment factors during the long-term operation,and may also bear the increasing overload load,as well as,bridges themselves may have various problems of low design standards or poor construction quality,which will gradually degrade the performance of the bridge components,and the degradation failure of the bridge components will accelerate the overall degradation of the bridges structure and reduce the safety of the bridges.Therefore,it is of great significance to establish an effective model to analyze the deterioration and safety of bridge structure,and to improve the quality of bridge management and maintenance decision.Based on Bayesian Networks theory,bridge components performance deterioration and safety are studied.So the main contents of this paper are as follows:(1)The Bayesian theorem and Bayesian Network theory are introduced,and Bayesian Networks modeling,learning,inference and extension algorithm are realized by MATLAB software programming.The Bayesian Network model for the degradation of concrete bridge’s resistance performance is established,and the probability explanation for the degradation of resistance performance is obtained by model reasoning.(2)For the traditional static Bayesian Network,the method of time slice is extended to discrete,conditional Gaussian and linear Dynamic Bayesian network by adding the direct mechanism of time correlation between variables.The bridge deterioration process is discretized into time series annually,the discrete Dynamic Bayesian Network and the conditional Gaussian Dynamic Bayesian Network analysis models of the performance degradation process of bridge components are established respectively by integrating prior knowledge such as statistical data and expert opinion,and the "Forwards-Backwards" algorithm based on Hidden Markov Model is used for these models inference.By adding the evidence node in these models,the bridge detection information at any time are effectively utilized,and the idea of transforming the former static analysis into dynamic updating of detection information is realized.The results show that the dynamic Bayesian Networks have efficient modeling,reasoning and updating ability in bridge performance degradation process analysis.(3)In the process of bridge components performance degradation,safety also decrease with the passage of time,aiming at this problem,a linear Dynamic Bayesian Network analysis model for the performance degradation process of simply supported T beams is established based on the prior knowledge of concrete structure performance degradation mechanism,and a "Forwards-Backwards" algorithm based on Kalman filter model is used for model inference.Degradation prediction model and structural reliability model are combined by setting the reliability analysis node in the Dynamic Bayesian Network model.So the safety analysis is realized in predicting the degradation process of flexural bearing capacity of simply supported T beams,the parameters of the prior model and reliability inference results are updated dynamically based on the bridge detection information.(4)The Dynamic Bayesian Network model for analyzing the performance degradation and security of bridge components is extended into decision graph model,by introducing decision nodes and utility nodes.The uncertainty of decision-making process is reduced by updating probability inference through Dynamic Bayesian Network.The probabilistic inference and update results are used,security conditions and objective function to minimize expected total cost are setted in the decision graph model.Safety analysis and maintenance cost and expected failure cost analysis of simply supported T beam are carried out,and the maintenance decision with minimum expected total cost is obtained.The results show that the Dynamic Bayesian Network has robust expansion ability and can provide effective support for the maintenance decision of bridge component degradation process.