Combined prediction model of joint opening-closing deformation of immersed tube tunnel based on SSA optimized VMD, SVR and GRU

Abstract

Prediction of joint opening-closing deformation is crucial to ensure the operation safety of immersed tube tunnels. This paper proposes a novel deformation prediction model based on the framework of "decomposition-prediction-integration", which combines the Sparrow Search Algorithm (SSA), Variational Modal Decomposition (VMD), Support Vector Regression (SVR) and Gated Recurrent Unit (GRU). First, the deformation sequence is decomposed into trend, period, and residual components by VMD. In this process, a innovative evaluation index (EI) is proposed to guide the decomposition by combining the Root Mean Square Error (RMSE) and the Sample Entropy (SE), while the VMD is hyper-parametrically optimized by SSA under the EI to improve the decomposition quality. Secondly, the trend component is fitted and predicted using the Least Squares Method; the SSA-SVR model is used to establish the nonlinear response relationship between the period component and the influence factor; the GRU model optimized by SSA is applied to excavate the temporal characteristics within the residual component, which in turn achieves the prediction of the residual component. Finally, the opening-closing deformation prediction results are obtained by integrating the prediction results of each component. The application results demonstrate that the proposed model is excellent in terms of reliability and applicability, with smaller prediction errors and higher accuracy than the other six models. It provides an innovative method for joint opening-closing deformation prediction.