Research On Kalman Filtering For Dynamic Deformation Monitoring

Dynamic deformation monitoring for large engineering structures can monitor their displacement status real-time, and corresponding maintenance measures can be adopted to avoid catastrophic accidents. Kalman filtering is a widely used dynamic data processing method; it can estimate state parameters form the noised observation data which is polluted with noise, and can make next forecast as well. In this paper, Kalman filtering method and its application in dynamic deformation monitoring are studied. The main contents and contributions are as follow:(1)The optimal algorithm of classical Kalman filtering is based on the precise function models and stochastic models; consequently it shows poor performance in some practical application. In this paper, several primary filtering method, such as classical Kalman filtering, robust Kalman filtering, Sage filtering and adaptively robust Kalman filtering are introduced and programmed with MATLAB language. Furthermore, the algorithms of equivalent weight matrixes and adaptive factors of adaptively robust Kalman filtering are researched and improved, and its application in the GPS observation data of a simulated vibrating table and a slop of certain high speed road have proved the availability. What is more, the strengths and weaknesses of each method of five are clear after comparative analysis, so as to guide practical application in dynamic deformation monitoring.(2)In the kinematic positioning or dynamic deformation monitoring, random walk model, constant velocity model or constant acceleration model is usually applied as the state model of Kalman filter. However, with this simple kinematic model to describe the complex movement of the deformed, the model errors are inevitably contained. In this paper, the three above-mentioned models are introduced. Besides, these three models are analyzed with a vibrating table simulation data. After comparative analysis, characteristic of theirs are shown, and a reference is provided for the choice of state model of the Kalman filter in the application of deformation monitoring. (3)In the high-precision positioning with GPS, when the environment of the surveying point keep little changed, the multipath effect has strong repeatability. With this characteristic establishing the error correction model is an effective method to weaken the multipath effect influence. But as the time interval goes on, its repeatability decreases, and the corresponding effectiveness of this method drastically reduces. Therefore, based on augmented parameters Kalman filtering this paper proposes a multipath effect system error estimation method with state matrix augmentation, taking the systematic errors as state parameters and establishing AR model of first class, meanwhile using multipath repeatability characteristics to update multipath error correction model. With this method, the problem of that as the time goes on the repeatability decrease and the corresponding effectiveness of fix multipath error correction model reduces has been resolved to a certain extent. Finally, an example with 16 days GPS observation data has proved that this method has a certain feasibility and effectiveness.(4)GPS-RTK offers direct three-dimensional coordinates measurements for structural health monitoring; accelerometer offers precise, high sampling rate acceleration observations in structural vibration monitoring. Aiming at making full use of the two kinds of sensors, a Kalman filtering method with the integration of GPS and accelerometer is researched, which is of multi-rate filtering and robust features. The experimental results show that:this method can effectively improve the accuracy, reliability and sampling rate of high building vibration monitoring.