基于灰关联定权组合模型的基坑变形预测分析

Predictive Analysis on Deformation of Foundation Pit Based on Combined Model of Grey Relation Fixed Weight

将基坑变形预测分为横向预测与纵向预测,横向预测选用相邻监测对象为影响因素,并建立各影响因素的一元非线性回归模型;纵向预测分别建立回归分析模型、时间序列模型、灰色系统模型、BP网络模型。以单一预测模型与监测对象的关联度为依据,确定了单一预测模型的权重值,建立了灰关联定权组合预测模型。分别运用单一模型与组合模型进行数据预测,并对预测结果进行分析比较,组合预测模型预测结果更为稳定、准确,能较好地反映监测对象的变化趋势。

Prediction of foundation pit deformation will be divided into horizontal and vertical prediction. In horizontal prediction,we establish a simple non-linear regression model choosing adjacent monitoring subject as the affect factor. Meanwhile,vertical prediction include;regression analytical model,time series model,gray systemic model,BP network model. Based on correlation degree of simple prediction model and monitoring the object,we determine the weight value of the single forecast model and establish a fixed weight combination of gray correlation forecasting model. In this paper,we forecast data using the single and combined prediction model respectively,and the analytical comparing results showed that the combine forecasting model method was more stable and accurate,it also could reflect the trend of the monitoring subject better.