Study On Evaluation And Prediction Of Suspension Bridge State Based On Detection Information Fusion

Intelligent and preventive maintenance has become the development direction of bridge maintenance.To improve the timeliness,accuracy,and foresight of maintenance decision-making of large suspension bridges,it is necessary to investigate the state evaluation and prediction methods of suspension bridges.With the development of detecting technology,there is more and more detection information on the bridge.But the evaluation method of the bridge is still mainly for the periodic evaluation of appearance detecting information.The evaluation results can not provide timely and dynamic guidance for daily maintenance decisions,various detecting information can not be fully utilized,and the judgment of bridge condition is not comprehensive and accurate.In addition,the use of the previous assessment results is still limited to simple comparative analysis and does not meet the needs of preventive maintenance.Therefore,this study relies on the National Key R&D Program of China "Research and demonstration application of key technical standards for remote state evaluation of smart bridge(2020YFF0217801)",the comprehensive utilization of detection information was taken as the starting point,and the dynamic evaluation of the short-term state of the suspension bridge was realized by giving full play to the advantages of monitoring big data.The comprehensive evaluation of the long-term state of the suspension bridge was realized through the establishment of a multi-source heterogeneous information evaluation system and the exploration of a fuzzy fusion algorithm.Based on the mechanical response prediction,the integrated state prediction of suspension bridges was achieved by integrating historical assessment information.The main conclusions are as follows:(1)According to the composition analysis of the detection information of suspension bridges,the evaluation indexes of suspension bridges were divided into four categories: apparent damage index,material degradation index,mechanical response index,environmental impact index,and the element composition of each index was given.The 3σ-grubbs abnormal data processing method and signal noise reduction and temperature effect separation method were put forward based on MEEMD,which lays a foundation for follow-up research.In the example,the 3σ-grubbs method can eliminate the error effectually,and the variance of the data was reduced by 49.0% after processing.The improved MEEMD can effectively suppress the endpoint effect and reduce the computational effort caused by white noise.(2)Aiming at the problem that the daily maintenance decision-making lacks the state basis of the bridge structure and important components,a short-term state dynamic evaluation method of the suspension bridge was proposed based on the fusion of mechanical response data.The dynamic comprehensive evaluation theory was introduced into the bridge evaluation,and the bridge dynamic evaluation system was established.Through the improved batches estimated and adaptive weighted fusion method considering sensor accuracy,the eigenvalue extraction of the measured point data was realized.The scoring formula of measuring points based on efficacy coefficient was given,and the scoring was modified by the grey correlation method.The health monitoring information,design information,bridge completion information,and special detection information were integrated into the scoring of measuring points.The dynamic evaluation of the short-term condition of a bridge is first realised by calculating the measurement point weights based on the design reserve,objectively weighting the parameters and subjectively and objectively weighting the components based on real-time data.Case studies show that the data fusion method proposed in this paper provides better data fusion results than similar methods and minimises data loss while ensuring fusion accuracy,and that the fusion results are closer to the high accuracy sensors.(3)Aiming at the problems of insufficient comprehensiveness and accuracy of periodic maintenance decisions,a comprehensive evaluation method combining quantitative scoring and fuzzy fusion was proposed based on the fusion of apparent damage data,material degradation data,mechanical response data,and environmental impact data,which realized the fuzzy integrated assessment of the long-term condition of suspension bridges.A comprehensive evaluation system based on the index module was established.The weights of the evaluation system were obtained by interval calculation,group judgment,and set value analysis of the expert questionnaire results,and the time-value variable weight correction method for component weights was given.The fully numerical evaluation criteria and the fuzzy,fusion,and anti-fuzzy calculation methods of the evaluation process were given.The evaluation example shows that the improved evidence combination method proposed in this paper has a better fuzzy fusion effect than the fuzzy operator and similar methods.Compared with the current industry standard evaluation methods,the evaluation results of this method are more consistent with the actual maintenance decisions,and the judgment of bridge status is more in line with the objective reality.(4)The mechanical response prediction of the bridge were realized,which laid a foundation for the comprehensive state prediction of the bridge.The state of target traffic flow was defined,and based on WIM information,traffic flow characteristics were extracted,and a random traffic flow simulation was implemented using a sampling method,and the SARIMA model was used to realize the late prediction and past prediction of the targeted traffic flow state.The finite element loading method of random traffic flow based on the time history function was proposed.Through finite element calculation and result feature extraction,the mechanical parameter prediction scoring were obtained.Stress prediction for the cable tower of the Cuntan Yangtze River Bridge shows that the finite element loading method in this paper improves the calculation accuracy compared with similar methods,and the deviation between the predicted stress score and the actual stress score is 8.4%,which is a good prediction effect.(5)To meet the needs of preventive maintenance,based on the mechanical response prediction,a comprehensive condition prediction method for suspension bridges based on historical assessment information was proposed.The discrete dynamic Bayesian network model suitable for comprehensive state prediction of suspension bridges was established according to the characteristics of the maintenance decision.Moment estimation hyperparameters were used for Bayesian estimation learning,DNB parameter updating under the condition of small sample information was realized.The updating process of the network model for comprehensive state prediction of suspension bridges based on historical evaluation information and mechanical response prediction results was presented.The example verification shows that,compared with the actual evaluation results,except for individual components,the average value of the probability deviation of the component’s state is 14.7%,and the probability deviation of the bridge’s state is 10.2%.The prediction deviations are within acceptable limits.