This article investigates the accuracy and applicability of existing analytical correlations for predicting drag forces on trains in tunnels, particularly focusing on the nose, side, and base drag coefficients. Originally formulated for constant-velocity trains under steady conditions, these correlations have not been thoroughly scrutinized or tested under varying scenarios such as train acceleration and entry/exit from tunnels. This study uses Computational Fluid Dynamics methods to evaluate these drag coefficients for axisymmetric train models of varying sizes and compares the results with predictions from one-dimensional models. The dynamic mesh technique simulates train motion, while the unsteady Reynolds Averaged Navier Stokes equations, coupled with a two-equation turbulence model, capture the turbulent flow behavior. The study examines two train lengths and velocity profiles (constant speed and acceleration). Findings reveal that side drag predictions by 1D models are acceptable if known under open atmospheric conditions. However, significant discrepancies arise in predicting base and nose drag. The analytical correlation underestimates nose drags at tunnel entry but overestimates it when the train is in the middle and at the end, resulting in a lower average error. For base drag, the correlation significantly underestimates values during tunnel transit, with Computational Fluid Dynamics results indicating an error exceeding 50%. Consequently, the 1D simulation tools based on these correlations underpredict the induced flow rate caused by the train's piston effect in tunnels.