Examination for Predicting Consolidation Settlement by Measurement Records

—Prediction of settlement by using neural network model—

Abstract

The earthfill structure such as embankments, which are constructed for the preservation of the agricultural land, has shown large settlement in the middle of the construction and after construction in the lowland area on the coast of the Ariake Sea, and then the long term settlement of those buildings is measured. The hyperbolic method is one of the most famous methods that predicting the settlement by using measurement records and is used extensively both domestically and internationally. In this paper the neural network model for predicting settlement by using measurement records in early stage is examined. Using the learning pattern that focused on the convergence of settlement rate, the prediction values are good agreement with the measured values. As a result, having the model learn the data that has a suitable regularity, the proposed method can provide the early prediction of settlement with high accuracy.