Research On Processing And Analysis Of Bridge Health Monitoring Data

For decades, along with the rapid development of C hina, there are more and more bridges are built. At the same time the bridge collapse incident also happened sometimes, therefore, it have been attached more and more attention to the health of bridges. In order to know the problem of the bridge, bridge health monitoring system(BHM in short), which can evaluate the state of bridge through sensors installed on bridge, has been built on more and more bridges. For BHM, whether the data from sensors is correct is key factor. So it is necessary to rate the quality of data, detect and get rid of error data, insert substitute for error data. Although BHM have made great progress, including data gathering and data save technology, but research on rating of data quality is not enough, and current error data detecting methods are not fit for BHM. At the same time, there is lots of information hidden in the data of BHM need to be mined.In this paper, aims to the data reliability assessment, abnormal data and data interpolation techniques, per-processing and analysis of monitoring data has been studied. A method based on grey correlation analysis has been presented for data reliability assessment, which can locate the time of the unreliable data happened and provide the basis for the bridge structure early warning. Error data is detected based on it’s statistical characteristics. In order to insert suitable substitute for error data, a combined scheme of the prediction based on grey correlation analysis and the model predict for time series ARMA model is presented. Considering on the developing trend of monitoring data could reflect the changing model to some extent, the overall trend of monitoring data have been observed and analyzed. And then a best fitting model has been established based on ARMA model, which could predict short-term monitoring data. Experiment of this model shows it is better than single ARMA model prediction. The main innovation of this paper is to apply the theory of gray correlation on forecasting data and eliminating prediction error of the time-series model. All methods presented in this paper, which are tested by experiments, are applied to many bridge monitoring systems.