Developing hybrid ELM-ALO, ELM-LSO and ELM-SOA models for predicting advance rate of TBM

Abstract

Accurate prediction of TBM performance is very important for efficient completion of TBM construction tunnel project. This paper aims to predict the advance rate (AR) of tunnel boring machine (TBM) using three hybrid models by combining three swarm intelligence optimization algorithm (Ant Lion Optimizer (ALO), Loin swarm optimization (LSO) and Seagull optimization algorithm (SOA)) and the Extreme learning machine (ELM) model, namely ELM-ALO, ELM-LSO and ELM-SOA model respectively. The dataset consists of 1, 286 samples from the Pahang Selangor Raw Water Transfer (PSRWT) tunnel project in Malaysia, and containing six parameters (rock quality designation (RQD), uniaxial compressive strength (UCS), rock mass rating (RMR), Brazilian tensile strength (BTS), thrust force per cutter (TFC) and revolution per minutes (RPM)) from rock mass and machines. In order to evaluate the prediction performance of different hybrid models, the root mean square error (RMSE), the mean absolute error (MAE), the mean square error (MSE), the determination coefficient (R2), the sum of square error (SSE), and the variance accounted for (VAF) were adopted as the performance indicators. The results show that ELM-LSO is the best model to predict AR. Sensitivity analysis shows the importance of all considerations to AR. TFC, RPM and RMR are the three most important parameters. But this is not absolute, more parameters need to be taken into account in AR prediction. Meanwhile, the ELM-LSO model proposed in this paper can be used as a new method to predict AR.