Abstract:

Since thick silty soil distributes widely along the Zhengzhou-Wuhan Passenger Railway, the post-construction settlements should be controlled strictly. The key to embankment filling are regarded as the control over filling speed based on the settlement observation data to assure the stability of embankment; the prediction of post-construction settlements to determine the period of ballastless track construction; and the control over settlements after construction to the allowable limit. The reliability and pre-processing of settlement data, and the method of discriminating abnormal settlement data arepresented in this paper. The characteristics and applicability of settlement prediction models are discussed. The LSSVM (least squares support vector machine) is used in the pre-processing of settlement data. Hence a systematic prediction model for silty soil is established. The detailed prediction process is expounded by giving examples, and the prediction results show that the settlement tends to be stable; no large post-construction settlement would occur during the operation period, which suggest that the foundation treatment and construction method is successful.

Key words: silty soil , settlement prediction , de-noise , LSSVM