PSO-LSTM BASED CONSTRUCTION SCHEDULE PREDICTION METHOD FOR SHIELD TUNNELING

With the development of shield method in water conservancy and hydropower tunnel construction, the duration prediction of shield excavation faces more complex environment and variable influencing factors. In order to consider comprehensive factors such as environment, pre-construction and shield excavation segmentation, this paper proposes a shield excavation prediction method based on particle swarm optimization long and short -term memory neural network (PSO-LSTM), and optimizes the long and short -term memory neural network (LSTM) by particle swarm optimization algorithm (PSO) to solve the problems of difficult to determine parameters of LSTM neural network model, low training efficiency and poor accuracy. The engineering example shows that the average error of the proposed simulation model is only 5.32%, which is smaller than the average error of other models. The real data proving that the proposed method can effectively predict the duration of shield excavation, which provides new data support for shield excavation duration control and resource allocation.