基于不同状态模型Kalman滤波的建筑物沉降预测研究

Building settlement prediction based on Kalman filtering of different state models

建筑物形变预测预报对减少生命财产损失具有重要意义,基于形变监测数据构建模型是形变预测分析的重要方法。本文利用Kalman滤波,基于不同状态模型构建预测算法,并分析验证不同模型的预测精度,以获取最优预测模型,提高预测精度。本文选取了长安大学地学科技大厦的各期形变监测数据,并结合Kalman滤波算法,分别采用随机游走、常速度、常加速度3种不同状态的估计模型,建立建筑物沉降预测算法,并对不同算法进行实验验证和对比分析。结果表明,基于常加速度状态预测模型的预测结果精度优于常速度状态预测模型和随机游走预测模型Kalman滤波的预测结果,其精度分别提高了15%和21%。

The prediction of building deformation is of great importance to reduce the loss of human life and property.Therefore,building a model based on deformation monitoring data is an important method for deformation prediction analysis.This paper uses Kalman filtering and builds prediction algorithms based on different state models,verifies and analyzes the prediction accuracy of different models,and obtains the optimal prediction model to improve the prediction accuracy.It selects the deformation monitoring data of Chang'an University Geoscience and Technology Building in various phases and combines with Kalman filtering algorithm,uses three different state estimation models of random wandering,constant velocity and constant acceleration to build the building settlement prediction algorithm,and verfies and compares the different algorithms by experiments.The results show that the accuracy based on the constant acceleration state model is better than based on the constant velocity state model and based on the random walk model,with an improvement of 15%and 21%respectively.