基于因子分析的滑坡位移多模型预测综合评判  

Comprehensive Evaluation on Multiple Predictions of the Landslide Displacements Based on Component Analysis

作　　者：缪海波[1] 殷坤龙[1] 徐峰[1] 刘斌[1] 

机构地区：[1]中国地质大学工程学院,武汉430074

出　　处：《武汉理工大学学报》2010年第19期65-70,共6页Journal of Wuhan University of Technology

基　　金：国家自然科学基金(40872176);国家"十一五"科技支撑计划(2008BAC47B01);中国地质大学(武汉)优秀青年教师资助计划(CUGQNL0915)

摘　　要：单一的预测模型准确地预测滑坡变形发展趋势是不充分的,为解决这一问题进行了滑坡位移多模型预测结果的综合评判,从而优选最佳的模型。以三峡库区八字门滑坡为例,选取3种模型对其变形趋势进行预测,并将多元统计学中因子分析引入到对3种模型预测结果的综合评判中,建立了模型精度评价矩阵,计算了主因子得分及各模型综合得分。结果显示:对于监测点ZG110,PEARL生物生长模型为最佳预测模型,对于监测点ZG111,PEARL生物生长模型和BPNN神经网络模型为最佳预测模型。It is inadequate to use only one prediction model to actually predict the deformation trend of landslide.In order to solve this problem,it is necessary to make multiple predictions and establish the comprehensive evaluation method to obtain optimum model.In this paper three models are selected to predict the deformation trend of Bazimen landslide.The component analysis of multivariate statistic is drawn into the comprehensive evaluation of the three prediction results.The modeling error judgment matrix is established and the principal factors score and each model score are calculated.The results show that PEARL is the optimum prediction model for the monitoring points ZG110 and PEARL and BPNN are both the optimum prediction models for the monitoring point ZG111 of the landslide.

关 键 词：滑坡位移 多模型预测 因子分析 综合评判 

分 类 号：TD853.34[矿业工程—金属矿开采]