Abstract: To ensure the construction safety of tunnel's entrance segment, the causes of collapse at the entrance segment of Xiaozhaishan Tunnel are analyzed. On such basis, the status and development trend of surrounding rock stability are assessed using the cusp catastrophe theory and PSO-ELM (Particle Swarm Optimization based Extreme Learning Machine) model. The collapse treatment measures are systematically studied, and the treatment effects are analyzed to evaluate the effectiveness of measures. Results demonstrated that the major causes of collapse are deviatoric pressure and rainfall. The catastrophe characteristic values in the process of catastrophe analysis are less than zero, indicating that the surrounding rock of entrance segment is in an unstable state. In addition, deformation prediction results suggest that the deformation would continue, which is unfavorable for the stability of the entrance segment. Collapse treatment measures should be comprehensive and systematic to maintain the deformation of surrounding rock within controllable range. The effectiveness of various treatment measures is verified. The research findings offer a new idea for the prevention and control of hazards at such entrance segment of tunnels. Moreover, the collapse treatment measures should be implemented according to the practice of actual project. Targeted reinforcement should be carried out on surrounding rock in each construction stage, and then the macroscopic stress state of the tunnel support should be adjusted to ensure the stress balance. After treatment, the deformation of surrounding rock in Xiaozhaishan tunnel entrance is within controllable range, which verifies the effectiveness of the treatment measures.

Key words: tunnel subjected to deviatoric pressure, tunnel entrance, cause of collapse, deformation prediction, cusp catastrophe theory, PSO-ELM model, stability analysis, Xiaozhaishan