Analysis Of Buffeting Response Of Long-span Bridge Based On Field Monitoring By Big Data Technology

In recent years, with the rapid development of economy and transportation, dueto the geographic restriction, long-span bridges have become the developing trendof bridge engineering.On the one hand, the increasement of the span of bridges leadsto the decreasement of stiffness and damping of the bridge structure; on the otherhand, with the cross section of long-span bridge deck increasingly streamlining, itswind sensitivity becomes higher. Wind-induced vibration problem of long-spanbridges becomes increasingly serious, and buffeting is the most common vibration.So far, research methods of wind-induced vibration of bridge mainly includetheoretical analysis, wind tunnel experiment and field measurement and numericalsimulation. However, each of the methods has some disadvantages, andconsequently the study results are in disparity with the actual wind-inducedvibration. The measured data of wind field and vibration characteristics isparticularly important as the data makes it possible to analyze buffeting response oflong-span bridge under a full scale condition. Unfortunately, we have not yet takenfull advantage of the data. In order to fully mine the valuable data mearsured in thefield and find useful knowledge of the bridge wind enineering, it is necessary to usedata mining techniques.This paper firstly introduces the process of the study on bridge buffeting andapplication study on the data mining techniques in engineering, and secondlyanalyzes the wind field characteristics and the acceleration response of a long-spansuspension bridge deck based on measured data. Finally with support vectormachine (SVM) method the data-driven prediction model of buffeting accelerationresponse is built, and it’s used to analyze the buffeting response factors.This paper applies the machine learning to the study on the bridge windengineering based on measured data. The study builds data-driven prediction modelsof buffeting response without any assumption in the buffeting theory, and findssome new knowledge of buffeting in the factors analysis of buffeting repsonse usingthe generated models. This paper provides a new effective method to analyzebuffeting response of a long-span bridge.