Displacement prediction of baishuihe landslide based on time series and PSO-SVR model

作者：Zhang, Jun^*; Yin, Kunlong; Wang, Jiajia; Huang, Faming

来源：Chinese Journal of Rock Mechanics and Engineering, 2015, 34(2): 382-391.

DOI：10.13722/j.cnki.jrme.2015.02.017

• BP neural network model
• Displacement prediction
• Particle swarm optimization algorithm
• Slope engineering
• Support vector machine regression (SVR)
• Support vector regression (SVR)
• Three gorges reservoir area
• Time series prediction

The choices of landslide displacement prediction models and relevant parameters are two of the most important issues in the landslide displacement prediction. The existing models have many limitations and shortcomings in predicting the landslide displacements. In this study, the trend displacement and the periodic displacement of Baishuihe landslide in the Three Gorges Reservoir area were separated based on the method of time series analysis. The former was considered to be mainly controlled by the internal factors of landslide(the composition, the geological structure, the topography, etc. ), and was fitted and predicted with the polynomial function. The latter was caused by the external influence factors(the seasonal rainfall, the water level changes of reservoir, etc. ). Taking the rainfall of current month, the cumulative rainfall of anterior two months, the reservoir level, the fluctuation of reservoir level of current month, the fluctuation of reservoir level of anterior two months and the cumulative increment of total displacement in current year as the influencing factors of periodic displacement, a combination of the particle swarm optimization algorithm(PSO) optimizing the parameters of model and the support vector machine regression(SVR) method for the periodic displacement prediction was proposed. The predicted values of the two kinds of displacements were superimposed to obtain the cumulative displacement. The result showed that the landslide displacements based on the time series prediction and PSO-SVR model were better than those from the grid search SVM and the BP neural network models.

• 出版日期2015-2-15