Stress Response Assessment of a Flexible Lining Structure in Tunnel Construction Using a Novel Hybrid Alternative Intelligent Model

Stress response assessment of flexible lining has always been a key problem during the construction and maintenance of tunnels in seismically active areas. There is a lack of reliable and efficient methods for calculating or predicting the maximum principal stress (MPS) of flexible linings in tunnel engineering. In this paper, a novel hybrid intelligent model, called the crayfish optimization algorithm (COA)-based kernel extreme learning machine (KELM) model, is utilized to precisely solve the aforementioned issue, i.e., the COA-KELM model. To further improve model performance, six types of data allocation are designed to train and test the proposed alternative models. Several indices and evaluation methods are utilized to determine the best model. The results show that the trained COA-KELM model achieves the highest prediction accuracy when the ratio of the training set to the test set is 90%:10%. Besides, the results of the parametric investigation and sensitivity analysis indicate that the damping layer thickness (DLT) is the most important feature for MPS prediction. However, simply increasing DLT is not necessarily beneficial for reducing MPS, especially when DLT exceeds 175 mm. In general, this paper provides a new solution based on artificial intelligence models for predicting the stress response of linings in tunnel construction.