Time-series prediction of shield movement performance during tunneling based on hybrid model

Abstract

This study presents a hybrid model based on the particle swarm optimization (PSO) algorithm and a long short-term memory (LSTM) neural network. PSO can determine the hyperparameters for the LSTM neural network. Using this approach, a framework for automatic data collection and application of the developed model during tunnel excavation was explored. The proposed model includes three stages: (i) data collection and pre-processing, (ii) hybrid prediction model establishment, and (iii) model performance validation. Pearson correlation coefficient is adopted to analyze the relationships between the influential factors and predicted object, which aids in feature selection for the developed model. A total of 1500 data sets, from a tunnel construction case in Shenzhen, China, were collected for training and testing the hybrid model. The results showed that the hybrid model with all the influential factors yielded the best performance. Thus, the developed model can provide a guideline for coping with measured data from an automatic monitoring system in earth pressure balance shield machines.