SEASONAL FROZEN TURFY SOIL SUBGRADE SETTLEMENT PREDICTION STUDY ON GREY AND BP NEURAL NETWORK THEORY

With the development of highway and railway construction in northern China, expressways inevitably pass through the turfy soil area. Turfy soil in seasonal frozen area has high water content, organic matter and low decomposition degree etc, so that theoretical calculation cannot meet the precision requirements of practical engineering. Therefore, this paper firstly analyzes the particular settlement mechanism of turfy soil subgrade in seasonal frozen area, and then puts forward to the optimized grey settlement prediction model and the two-dimension double-layer BP neural network settlement prediction model. The two simulation models not only take the disturbance due to complicated engineering geological properties of turfy soil and seasonal freezing in northern China into account, but also introduce filling condition as one of engineering factors to deep learn subgrade settlement changes law. This paper takes the actual settlement monitoring data of seasonal turfy soil subgrade from jilin to yanji close to Changbai Mountain for instance, both the fitted and the predicted results of two models are comparatively analyzed which shows both of two models have high accuracy, but each has its own advantages and disadvantages. thus, this paper further summarizes the similarities and differences of two models in this kind of engineering, which in the meantime provides some reference for the application of the multi-factor prediction model to such turfy soil subgrade settlement prediction in northern China.