Application of regularized ELM optimized by sine algorithm in prediction of ground settlement around foundation pit

The construction of the engineering is often accompanied by the excavation of the foundation pit, which will lead to the ground settlement around the foundation pit. It poses a serious threat to the surrounding buildings and even causes casualties. This puts forward higher requirements for the prediction accuracy of ground settlement around the foundation pit. The existing prediction model of ground settlement around foundation pit only considers the single factor of excavation days. However, the ground settlement around the foundation pit is affected by many factors. Obviously, the prediction model considering single factor is not accurate enough. To accurately predict the settlement of the ground around the foundation pit, the sine algorithm (SA) is used to optimize regularized extreme learning machine (RELM), and a prediction model of ground settlement around foundation pit related to excavation elevation, groundwater level, number of supporting layers and geotechnical parameters is proposed. The model is applied to the prediction of ground settlement around a pipe jacking receiving foundation pit in Foshan City, Guangdong Province, China. SA-RELM models based on time series and multiple factors are established, respectively. The prediction results are compared to BP neural network, ELM and SA-ELM model. The results show that the relative error, MAE, MAPE and RMSE of SA-RELM model based on multiple factors are the smallest, which are close to 0. The comprehensive analysis shows that the SA-RELM model has high accuracy and generalization ability, and the SA-RELM model based on multiple factors has higher accuracy in the prediction of ground settlement around the foundation pit.