Research On Fault Prediction And Early Warning Of Urban Rail Transit Bridge Equipment

At present,the traditional operation and maintenance mode is generally used in urban rail transit in my country,and the shortcomings of its maintenance lag have been exposed.Therefore,relevant scholars at home and abroad have gradually shifted their research focus to intelligent operation and maintenance and pushed the development of my country’s urban rail transit towards intelligence.Taking the urban rail transit bridge system as an example,this paper studies the intelligent fault prediction and early warning method of bridge equipment based on the bridge data.Firstly,a bridge data regression prediction model is established based on the Random Forest(RF)algorithm to locate and identify faults early.Then,considering the overall structure of the bridge,the information of different types of sensors is fused by determining the warning level indicators.At last,the accurate prediction of rail transit bridge system failures and comprehensive early warning of the overall structure is achieved.Firstly,taking the structural displacement of urban rail transit bridges as an example,this paper preprocesses the bridge data,then the Pearson coefficient method is used to analyze the structural displacement data trends of the various factors’ influences,such as structure temperature,bearing displacement,dynamic strain,longitudinal vibration.Finally,based on the actual operation data of an urban rail transit arch bridge,a failure prediction model based on the random forest is constructed,and the prediction performance is compared with Auto Regression Moving Average(ARMA),Principle Component Regression(PCR)and Support Vector Regression(SVR)models.The research results show that the mean square error of the random forest model combined with the Pearson coefficient is 0.0851,and the coefficient of determination is 0.9821.At,last,the advantages of random forest in bridge failure prediction are analyzed based on the combination of prediction performance and training time.Then,considering the interaction between multiple monitoring indicators of urban rail transit bridges,a comprehensive early warning method for urban rail transit bridges is designed based on Dempster/Shafer(D-S)evidence theory.Firstly,based on bridge sensor monitoring data,a homogenous sensor fusion based on Euclidean distance is adopted,and it implemented correction of abnormal data.After that,this paper sets the warning level for different types of sensors,and finally realizes the comprehensive early warning of the overall operation state of the bridge through D-S evidence theory information fusion.Furthermore,this paper studies the principles of multi-sensor information fusion technology and analyzes the shortcomings of the traditional D-S evidence theory algorithm.Starting from the conflict of evidence sources,the evidence weighting conflict function is introduced.The Artificial Bee Colony Algorithm(ABC)is used to improve the performance of D-S evidence theory algorithm.It first finds the optimal weight of evidence through artificial bee colony algorithm,and assigns weights to each sensor and then fuse them,which reduces the conflict in the fusion process.The simulation results show that the use of the improved D-S evidence theory algorithm for sensor information fusion can effectively solve the problem of conflicts between different equipment,improve the reliability of the overall status early warning of the bridge,and also improve the efficiency of urban rail transit equipment operation and maintenance.and quality.