论文部分内容阅读
为了保证施工的正常进行、实现信息化施工,必须对基坑实际监测数据进行分析与预测。现有的趋势时间序列分析方法很难满足实际施工中高度非线性问题的拟合,预测误差较大。基于这点考虑,以天津某工程基坑施工地表沉降观测序列为例,在对原始数据进行分析的基础上,提出既可以考虑趋势时间序列,又具有高度非线性拟合性能的时序-投影寻踪回归模型。首先,通过比较分析几种时间序列方法的逼近误差和预测误差,寻求出一种逼近较好的时间序列预测方法。然后,将预测得到的时间序列和观测数据相结合,采用投影寻踪回归方法拟合。应用结果表明,该模型逼近性能良好,预测误差小,可为深基坑位移沉降的动态预测提供一条较好的途径,对基坑动态设计与信息化施工等方面具有重要的参考价值。
In order to ensure the normal construction and realize the information construction, the actual monitoring data of foundation pit must be analyzed and predicted. Existing trend time series analysis method is difficult to meet the actual construction of a high degree of non-linear fitting, prediction error larger. Based on this point, taking the observation sequence of surface subsidence in the excavation of a certain project in Tianjin as an example, based on the analysis of the original data, a time-projection finding method that can not only consider the trend time series but also have a high degree of nonlinear fitting performance Tracing regression model. First of all, by comparing the approximation errors and prediction errors of several time series methods, we find a better time series prediction method. Then, the predicted time series and observation data are combined and fitted by projection pursuit regression method. The application results show that the model has good approximation performance and small prediction error, which can provide a better way for dynamic prediction of displacement settlement of deep foundation pit and has important reference value for dynamic design and information construction of foundation pit.