关键词: 基坑, 建筑倾斜, 时间序列预测, 最小二乘支持向量机, 粒子群优化算法

Abstract: Against the complexity of influential factors of building inclination deformation around foundation pit, and the small sample and nonlinear characteristics of monitoring data, a time series forecast method of building inclination around foundation pit based on Least Squares Support Vector Machine（LSSVM） optimized by Particle Swarm Optimization（PSO） is proposed. Firstly, phase space reconstruction is used to reconstruct the differential settlement time series of early working stage of the building around foundation pit, the samples of differential settlement forecast are composed, and then the samples are input to LSSVM to train. Secondly, the parameters of LSSVM are optimized by PSO to obtain the optimal forecasting model, the differential settlement of the later working stage is rolling predicted by this model and plugs the forecast results into formula to calculate the inclination deformation value. Finally, this method has been successfully applied in the surrounding building inclination forecast of a foundation pit engineering in Kunming, the result is satisfied.

Key words: foundation pit, building inclination, time series forecast, Least Squares Support Vector Machine（LSSVM）, Particle Swarm Optimization（PSO）