Prediction of the bridge monitoring data based on support vector machine

According to the mass health monitoring data accumulated of bridge structure for a long time, this paper proposes a method for reconstitute the time series in phase space. Since the phase points are regressed by support vector machine (SVM), the relevant time series of past behavior patterns are established. Then, it could infer the future development trend and form the basis of the online security early warning of bridge structure. The strain and tilt monitoring data of Pian Yan-zi bridge in Chongqing are analyzed and compared with the prediction data of the auto regression moving average (ARMA). The results show that: ? as to the bridge monitoring data, the prediction accuracy of SVM is better than that of ARMA; ? With the increase of the number of the prediction steps, the prediction accuracy of ARMA drops dramatically. And ARMA is only applicable for short-term prediction while SVM is able to predict a longer period of time. ? SVM prediction requires a smaller size of samples for modeling but with higher prediction efficiency.