Application of Integrated Forecasting Model Based on Multiple Variables in Tunnel Vault Settlement

Abstract

The settlement of tunnel vault, which is a complicated process that includes combined actions of multiple factors together, is difficult to calculate accurately with a mathematical model. In the actual construction process, the surrounding rock often changes, so we must strictly control the tunnel vault settlement to meet design requirements. Traditional prediction models can obtain only a rough estimate of the vault settlement by using the monitoring data to fit and forecast its changes. To solve the problem, this paper constructs a time-series model, GM (1, 2) model, and BP model with input variables to predict the settlement of the vault. The effectiveness of the model is verified by examples, which show that the GM (1, 1) prediction model of a single variable is completely ineffective. To avoid the lower prediction accuracy of a single model and use the complementary advantages of different models, this paper establishes an integrated prediction model based on the above three models, and its weighting coefficient is determined by the application of the entropy method. The model is applied to Baimiaozi Tunnel for testing, which provides high-speed travel from Baoji to Hanzhong (two cities in Shanxi Province). The integrated model is effective in practice and has a high prediction accuracy.