Optimized combined forecasting model for hybrid signals in the displacement monitoring data of concrete dams

Abstract

The chief aim of this study is to solve the problem of effective utilization of hybrid signals in the displacement monitoring data of concrete dams. Therefore, an optimization analysis scheme for complex and irregular monitoring data is proposed. The prototype monitoring data are divided into three main components and one residual component by classical statistical calculation. In consideration of the fuzzy generation principle of aging component and residual component, a nonlinear autoregressive model with exogenous inputs (NARX) based on singular spectrum analysis (SSA) is constructed. Moreover, the support vector machine (SVM) regression model optimized by artificial bee colony (ABC) algorithm is used to predict the water pressure component and temperature component with clear driving factors. Accordingly, a combined prediction model for hybrid signals in the displacement monitoring data of concrete dams is established. The engineering example shows that the model constructed in this study can identify the nonlinear time-frequency characteristics of the prototype monitoring signals better than other models.