Damage Identification Based On Big Data Of Response Statistics Collecting From Vehicle-Bridge Interaction

Structural health monitoring of bridges has become a hot issue in the field of civil engineering.Bridges will fall in damaged situation of varying degrees during operation due to various factors.Combined with vehicle-bridge coupling vibration caused by vehicles crossing bridge,it will pose a potential safety hazard to vehicle-bridge system.Vehicle-bridge coupling and damage identification are of great significance to the normal maintenance,safe operation and life evaluation of bridges.Bridge health monitoring system is an important way to make data monitoring,damage identification and condition assessment of bridge structure during operation.However,at present,there are still some shortcomings in data processing and data mining methods of massive health monitoring data,which makes it unable to give full use of its role.How to use monitoring data to effectively and timely assess the operation status of bridges is a key and difficult problem that scholars in the field of bridges pay attention to,and also a restriction point to play the role of massive monitoring data.Therefore,in this paper,random traffic flow is used as the excitation to extract the response signals of bridge under different damage conditions.Machine learning is used to identify structural damage.An intelligent damage identification method combining the statistical characteristics of bridge response is proposed.The contents of this paper are detailed below.1)Developing a program for the analysis of coupled vibration of bridge under random traffic flow.Based on the cellular automata-NaSch rule,combined with the measured traffic flow data,a single lane random traffic flow model is established to obtain the traffic flow data that are in line with the actual traffic flow.A vehicle-bridge coupling vibration analysis program based on Newmark-? method is developed.By combining the two methods,the random traffic-bridge coupling vibration analysis program is developed.2)Taking the bridge response under the coupling of random traffic flow as the input of damage identification,combining with various machine learning methods.A damage identification method combining the statistical characteristics of bridge response data is proposed.The response time series of bridges in vehicle-bridge coupled vibration under different damage conditions is obtained,and the influence of damage on the response of measurement points is discussed.According to the statistical characteristics of bridge measurement point response,a damage identification sample bank is established,and machine learning methods such as Random Forest,Support Vector Machine and Gradient Boosting Decision Tree are introduced for damage identification.