| In order to ensure the operational and structural safety of large span suspension bridges,it is urgent and necessary to make assessment of its status and make warning of anomalies.Although the bridge maintenance and management data are characterized by large amount and diverse types,the condition assessment of bridges is still dominated by visual inspection data,causing low utilization rate of existing bridge maintenance and management data.For the bridge alarming system,the static alarming method has the limitaions of false alarm and omission,which will weaken owners' confidence in alarming system.Under the background of bridge big data,this paper makes full use of the existing bridge multi-source information,and establishes the condition evaluation model for long-span suspension bridges.An applicable evaluation method of sequence index is proposed based on the in-depth mining of time series data.Considering the multilaterality of the bridge operation environment,the influence of the traffic volume,accumulated damage and ambient temperature is taken into account to determine the alarming threshold.The main research contents of this paper are as follows:(1)Study on the condition evaluation model of long-span suspension bridges based on multi-source information.Based on the multi-source bridge maintenance and management information(i.e.,visual inspection data,nondestructive testing data,and long-term monitoring data),combining with the characteristics of suspension bridge structure and deficiency features,obeying the rules “comprehensiveness,simplicity,independence,objectivity and testability”,through expert questionnaires,expert meetings,literature investigation and field research,multi-source information-based condition evaluation model of large-span suspension bridges is established.The initial weight of the index system is determined by the group AHP method after the survey of 45 experts nationwide.Compared to the weight values stipulated in the code,it is found that the weights of the main tower and affiliated facilities in this paper seem more rational,which could be used as a reference for the revision of the related specifications.(2)Study on sensor fault diagnosis method based on multi-source information fusion.In view of the limitations of existing sensor fault diagnosis models in multi-fault diagnosis and the coupling problem of sensor fault and structural damage,a fault diagnosis method is presented based on the similarity of monitoring data from symmetric position sensors.The similarity of symmetrical position sensor monitoring data is validated via theoretical and practical data analysis.On the premise that the similarity of monitoring data from the symmetric position sensors meets the requriments,Dasarathy's information fusion model is adopted to realize the fault recognition with Euclidean distance as the similarity index.Sensors are detected pair by pair to solve the problem of multi-fault diagnosis.Aiming at the coupling problem of sensor fault and structural damage,the concept of integrated similarity of target region is proposed based on the evidential reasoning theory to deduce the reason of data abnomality.Once determining the suspicious sensor pair,regression fitting analysis is used to isolate and reconstruct the sensor fault according to the physical redundancy information.The effectiveness of the proposed sensor fault diagnosis method is verified by the monitoring data from multiple sensor types(e.g.,girder alignment data and girder stress data).(3)Study on evaluation method of sequence index based on data mining.Due to the stability of the dead load effects,the dead load effects are selected as the main body for evaluation of the sequence index.In view of the high cost of obtaining the dead load effects by closing the bridge,long-term monitoring data are combined with numerical analysis(calculating the proportional coefficient)to extract the dead load effect level under random vehicle loadings.The accuracy of the extraction method is verified using the monitoring data of girder alignment of a suspension bridge.Based on the extracted dead load effects,condisering the homogeneity and non-homogeneity of the sequence index,the grey relational degree method is used to evaluate the sequence index.Taking muiti-year data of a suspension bridge as an example,the dead load alignment sequence index of stiffness girder of a suspension is evaluated.(4)Study on bridge condition evaluation algorithm based on system engineering theory.Two alternative algorithms are proposed to resolve the problem of the balance between the indicators in the assessment process.Based on the factor-based variable weight theory,the definition of age-dependent variable weight is put forward.Considering the limitations of constant weight model and traditional variable weight model,an age-factor dual variable weight model is proposed.The effectiveness of the age-factor dual variable weight model is verified by four typical cases of a suspension bridge.Based on the local variable weight theory,a local variable weight model for bridge evaluation is presented.The values of the two kinds of parameters(i.e.,penalty level and variable weight range)in the local variable weight model are determined through a large number of trials.The effectiveness of the local variable weight model is validated through four typical cases.For the uncertainties in the evaluation process,a normal cloud model for bridge evaluation is established according to the cloud theory.Compared to the classical fuzzy membership model,the normal cloud model not only considers the fuzziness,but also the randomness.(5)Development of an intelligent evaluation system for large span suspension bridges.Considering the complexity of the algorithm,in order to further improve the operability of the evaluation method,the intelligent evaluation system of large span suspension bridges is designed and developed.The system is equipped with MYSQL database and B/S network mode.The system includes the functions in bridge information management,inspection and detection,evaluation,warning and prediction,statistical statements,and query and help.Three typical large span suspension bridges are taken as examples to verify the applicability of the developed evaluation software platform.(6)Study on dynamic alarming method based on time series data.On the basis of summarizing the existing alarming systems,the classic 2-level alrming system(i.e.,yellow alarm and red alarm)is selected according to the physical significance of each level.In order to resolve the problems of false alarm and omission in static alarming system,the concept of dynamic alarming is presented.Pareto extreme value theory is used to forecast the baseline of yellow alarming threshold corresponding to 95% guarantee rate within the design baseline period(100 years).Considering the influence of traffic growth and accumulated damage,it is suggested to update the threshold bechenmark value regularly with the latest monitoring data.Based on the field testing data,the finite element model needs to be updated regularly to determine the baseline of the red alarm threshold.Based on the reference value of alarming threshold,considering the impact of environmental temperature,the signal processing method is used to simulate the thermal effect to make the threshold line change with the ambient temperature.Taking the midspan deflection of stiffened girder of a suspension bridge as an alarming index,the effectiveness of the dynamic alarming method is verified. |
PDF Full Text Request