Forecasting the Construction Duration of Ballastless Track in High-Altitude Tunnels under Unexpected Disturbance

Abstract

As railway development accelerates in plateau regions, ballastless track construction in tunnels encounters challenges from confined spaces, complex human-machine coordination, time-varying labor productivity, and unpredictable disturbances, limiting the ability of conventional methods to capture the dynamics of construction duration. This study proposes a hybrid simulation framework integrating system dynamics and agent-based modeling (ABM-SD). The framework captures time-varying labor productivity under high-altitude and continuous working conditions, formalizes interactions and constraints among personnel, equipment, and transportation, employs Monte Carlo–based personnel speed distributions to capture stochastic volatility and Bayesian prior networks to model low-probability high-impact disturbances, and introduces an adaptive meshing strategy to balance computational cost and efficiency, enabling the derivation of probabilistic duration distributions. Validation using a plateau tunnel project in Southwest China shows that the construction duration distribution predicted by the disturbance-integrated ABM-SD model is slightly right-skewed and follows an approximately normal distribution, with an average of 47.80 days. The actual construction duration is 48.63 days, corresponding to the 84th percentile of the cumulative probability, with a relative prediction error of approximately 2%. A comparative experiment demonstrates that the simulation results of the proposed model align more closely with actual engineering outcomes than those of traditional models. The adaptive meshing strategy accelerates simulation by approximately 3 to 6 times, with only a slight loss in simulation accuracy, enabling efficient large-scale probabilistic analysis.