Study On Graded Condition Assessment Method Of Long-span Continuous Rigid Frame Bridge Using Deflection Data

The long-span continuous rigid frame bridge has a series of advantages such as strong spanning ability,reasonable internal force distribution,good economic performance and beautiful appearance,and has become an important bridge structure form for crossing deep canyons with complex terrain.However,under the combined effect of multiple factors such as prestress loss,shrinkage and creep,and material aging,the deterioration of the health of the long-span continuous rigid frame bridge in service will be inevitable,which will seriously affect the normal use and carrying capacity of the bridge.Thus,fully diging the monitoring information,accurately assessing the health condition of the bridge,and predicting the unfavorable working status of the bridge in time is of great research significance and application value for enhancing the management level of the bridge,ensuring the safety of the bridge operation and the smooth flow of the road network.Based on the deflection monitoring data of long-span continuous rigid frame bridges,and on the basis of full investigation of relevant research results at home and abroad,this thesis focuses on the main line of long-span continuous rigid frame bridge deflection monitoring data processing-load identification-risk warning-safety assessment.Through theoretical analysis,numerical simulation,algorithm development,and real bridge verification,an efficient separation algorithm for deflection monitoring data was proposed,a prediction method for long-term deflection is proposed,a bridge random excitation recognition technology was established,a multi-level early warning mechanism based on separated deflection was established,and a practical safety assessment method for early warning using hierarchical information fusion is constructed.The main research contents and conclusions are as follows:(1)A method for separating bridge deflection monitoring data based on variational modal decomposition(VMD)and K-L divergence(KLD)is established.Firstly,VMD was used to decompose the monitoring signals of bridge deflection,and the kernel density of each component was estimated.Secondly,KLD was used to calculate the divergence between each component and the source signal,and then the false components were eliminated accurately to obtain each deflection component.Finally,Pearson correlation coefficient was used to evaluate the decomposed deflection components to further judge the authenticity of separated ingredients,thereby obtaining the temperature effect,live load effect and dead load effect of bridge deflection.In order to verify the effectiveness of the procedure,the empirical mode decomposition algorithm(EMD)is used for comparision.The results indicate that the proposed method effectively overcomes the inherent modal aliasing problems of the traditional EMD algorithm.In the numerical example,the separation effect of deflection day,annual temperature difference and longterm effects are increased by 7.52%,6.39% and 8.41%,respectively.(2)A bridge long-term deflection prediction method based on polynomial distributed lag and particle filter is proposed.Firstly,a prediction model is constructed to describe the long-term deflection development trend of the bridge.Secondly,combined with the particle filter algorithm,the prediction model is transformed into the state space,and the parameter estimation model based on the particle filter is constructed.Finally,the parameters of the bridge long-term deflection prediction model are identified according to the particle filter algorithm,and the long-term deflection is predicted.The results show that the proposed deflection prediction method can accurately predict the long-term deflection of bridges and update the model parameters quickly.(3)A bridge random excitation recognition technology based on power spectrum density transmissibility and adaptive multiplicative regularization is proposed.First,the concept of power spectrum density transmissibility is introduced to construct the transmission path between different responses;second,based on the separated dynamic deflection response power spectrum,the random excitation is located by minimizing the prediction error.Subsequently,an improved adaptive multiplicative regularization strategy was established,and the optimization function of each order of pseudo excitation synchronous recognition was constructed,and the GIRLS algorithm was used to solve it.Finally,the random excitation is reconstructed based on the identified pseudo excitation.The research results show that the proposed method can accurately locate the target random excitation under the interference of 20 d B white noise,and has good anti-noise ability.At the same time,compared with the traditional Tikhonov regularization method,the proposed method greatly improves the efficiency while maintaining similar reconstruction accuracy.(4)A hierarchical assessment method of bridge safety status based on multi-source information fusion is constructed.On the basis of obtaining the long-term effect of bridge deflection,live load effect and external random excitation,to accurately evaluate the health state of the bridge,firstly,the bridge separation deflection threshold is determined.The four-level early warning mechanism of color,yellow,orange,and red clarifies the specific connotation of the early warning threshold at each level;and then the calculation method of the deflection grading early warning threshold is proposed,which mainly involves the use of extreme statistical analysis methods to calculate the blue early warning threshold,and orange and red warning threshold calculation method of live load effect considering the randomness of structural parameters under random excitation,and the calculation method of yellow,orange and red warning thresholds for long-term effects of deflection based on statistical analysis of historical data;finally,the evidence theory is introduced into the multi-source information fusion of deflection early warning,the identification objectives and evaluation indicators are clarified,and the basic reliability distribution function from the interval number Euclidean distance is construced.The weighting coefficient is updated,and the updated basic reliability function is merged using Desmpster synthesis rules to form a comprehensive evaluation result.(5)The application of condition assessment method based on deflection monitoring of long-span continuous rigid frame bridges has been carried out.First,the project overview of the supporting project is introduced,and the situation of deflection monitoring is explained;then,the historical monitoring deflection data is separated based on the proposed VMD-KLD separation method,and the deflection live load effect and long-term effect are obtained;The proposed early-warning threshold calculation method establishes a hierarchical early-warning system of the above two components;finally,evidence theory is used to integrate the hierarchical early-warning information of monitoring deflection in a certain period of time to realize the comprehensive evalution of the bridge condition.The results demonstate that the proposed method obtains the deflection components better,establishes a rational hierarchical mechanism,makes the system being useful for early warning and forecast,and effectively combines multi-level early warning information to achieve both qualitative and quantitative evaluation.The evaluation conclusion indicates that the relying project is operating in good condition.