Research On Decision Support Theory And Method Of Intelligent Bridge Detection And Maintenance In Highway Traffic Network

With the rapid development of China’s economic and infrastructure construction,the transportation tasks of the highway transportation network are becoming increasingly arduous.China has numerous bridges and the construction speed is very fast.To effectively meet the high societal demands for the transportation and traffic capacities of existing bridges,it is necessary to develop scientific bridge detection and maintenance strategies led by information technology to improve the service performance of the bridges and reduce capital investment.In this dissertation,an array of measured data collection,intelligent algorithms,and relative methods are used to systematically study the theories and methods of intelligent detection and maintenance decision-making in the life cycle of highway traffic network bridges.First,the cumulative work of highway bridge detection and maintenance in multiple regions of China was studied.Then,the long-term highway bridge detection and maintenance data were collected and organized,followed by the formulation of a bridge detection and maintenance database.The prediction formula for the technical condition of a bridge before its maintenance was established using the measured data.Based on the measured data of the technical condition of the repaired bridge,the calculation method of the maintenance gain coefficient considering its service time and the number of repairs were proposed.According to the improved inverse Gaussian degradation,a quantitative method for the deterioration of the technical condition of the repaired bridge was proposed,considering the influence of different maintenance periods and deterioration degrees on its deterioration rate.Finally,Bayesian updating was used to correct the prediction results based on the newly obtained data.On this basis,a dynamic prediction method of bridge maintenance based on its technical conditions was proposed.Through the prediction performance test,the accuracy of the prediction method was verified to meet the actual engineering needs.Additionally,the impact of different maintenance schemes on the change law of the bridge’s technical conditions was calculated using the prediction method,which validated the practicality of the proposed method in real-life bridge maintenance.Second,based on the actual project owner cost data for bridge detection and maintenance database,the random forest method was used to analyze the importance of the influencing factors,and exploratory data analysis was used to verify.The classified data were screened for outliers using the Bayesian modified confidence test.A bridge owner cost prediction model was established by combining the chaos genetic algorithm and multilayer neural network and using the Bayesian optimization algorithm to determine the bridge’s structural topology,and the prediction accuracy was tested.The results are as follows: the adopted intelligent algorithm and established bridge operation and maintenance owner cost prediction model exhibited high prediction accuracy,which could provide a reliable reference for bridge detection and maintenance owner cost analysis.Third,by incorporating the influence function of bridge maintenance on the highway capacity into the traffic distribution model of the highway traffic network and combining it with the dynamic bridge technical condition prediction method,a network traffic flow redistribution model considering the influence of bridge maintenance was established.Based on the calculation of vehicle travel time delay and travel distance extension in a traffic network during bridge maintenance,a social cost evaluation method was proposed.Through the case analysis of bridge maintenance in an actual traffic network,the social costs resulting from single-and multiplebridge maintenance were calculated.The results showed that the proposed social cost evaluation method can effectively serve the social cost calculations caused by bridge maintenance.Fourth,combined with engineering practice,the decision variables and rules were established on the basis of reliability theory,and the improved gray wolf algorithm was used to adaptively optimize the decision vector to determine the vector with the lowest decision cycle cost.The actual bridge was considered an optimization example,and the minimum cost for the owner was assumed as the optimization goal.The calculation results verified an optimal solution of the decision vector and the effectiveness of the adaptive optimization model.Furthermore,by combining the decision-maker tomographic analysis method with the multiattribute utility theory,the bridge maintenance strategy that considers the owner and social costs was optimized and compared with the maintenance strategy,only considering the owner cost.Finally,a spatiotemporal correlation analysis method was proposed for bridge maintenance under normal service conditions,comprehensively considering the bridge performance degradation degree,performance degradation rate,and network traffic flow changes.Based on this,maintenance strategy optimization of bridge groups in a regional traffic network was performed.The calculated spatiotemporal correlation coefficient matrix and length of maintenance bridges can effectively assist in planning maintenance strategies for bridge groups in regional traffic networks.Finally,to realize an effective connection between the research results and actual bridge detection and maintenance work,the Visual Studio development tool was used to systematically combine the dynamic prediction method of the bridge’s technical condition,the cost prediction model of inspection and maintenance owners,the social cost assessment method of bridge maintenance,and the dynamic maintenance strategy model of the bridge.By combining machine learning,probability theory,and mathematical statistics with computer software programming technology,an intelligent inspection and maintenance decision support platform was constructed for highway traffic network bridges.It realizes the data-driven quantitative prediction of the bridge’s technical condition and cost,along with the precise decision-making of bridge groups in highway traffic networks.The actual road traffic network in the main urban area of a provincial capital in China and 19 main bridges in the network were used for the case analysis and discussion to comprehensively demonstrate the computing functions and operation methods of the platform.