隧道下穿时基于傅里叶时间序列预测临近结构沉降发展

隧道下穿临近结构时,目前仅能基于现有沉降判定结构是否安全,缺乏对未来短期内沉降发展的预测方法.通过测量某隧道右线下穿临近某水库时水库逐小时的沉降,分析得出水库短期沉降主要由隧道下穿引起的趋势项和每日温度波动导致的周期项组成,并根据以上特点提出一种基于傅里叶时间序列的结构短期沉降预测方法.该方法采用傅里叶级数和线性函数对周期项和趋势项分别进行拟合,并通过数据的新陈代谢保证一旦沉降发生突变后,该方法所得预测沉降对突变具有较好的灵敏度.采用该水库在同一隧道左线下穿时的沉降实测数据对该方法和GM(1,1)灰色模型的预测精度进行比较.结果表明,该方法的预测精度高于GM(1,1)灰色模型.

A forecasting method based on Fourier time series for short-term settlements in adjacent structures during an under-passing shield tunnel construction

Short-term settlement increment of adjacent structures is an important index to control the speed of excavation of an under-passing tunnel. Due to lack of effective method to forecast the short-term settlement increment in the time period between the just finished settlement measurements and the next one, it is difficult to carry out some efficient actions to prevent sudden change in settlement. Because the manual settlement observation cannot provide enough data in everyday measurements to show the detailed variations in settlements, the settlement-time curve of an adjacent structure during an under-passing shield tunnel construction has been measured. Analysis shows that the short-term settlement can be divided into a trend part and a periodic part, which are caused by the under-passing shield tunnel construction and environmental temperature cycling, respectively. Then a first-order transformation and the Fourier expansion can be used to fit the trend part and the periodical part in the time period, respectively. The sum of the two parts is the predicted value of short-term settlement. Another characteristic of this method is self-correcting forecasting by replacing the oldest known data by the newest known data to make this method sensitive to sudden change in settlements. At last, a series of practical monitoring settlement from a reservoir near an under-passing shield tunnel construction are used to compare the forecasting precision between the proposed method and the Gray model. Results from the comparison show that the forecasting settlements from the proposed method are more accurate than those from the Gray model.