Abstarct: Due to the increasing application of the tunnels in today's world and their closed and lengthy atmosphere, safety enhancement and fire accident control-related designs are many design engineer's concerns. In the interim, the airflow velocity control during the fire is directly related to the lives of the present people in the tunnel, and the lack of accurate supply of this flow and its proper velocity, also known as the critical velocity, can cause irreparable damages. The main focus of this thesis was to determine the critical and practical variables considering the proposal of a model closer to reality to estimate the critical velocity with a higher accuracy baxsed on the state of each tunnel and to take a step toward improving safety. The first step in this thesis was to identify 14 effective parameters on fire; some of these parameters were deleted. In the end, the tunnel length, cross-section, transverse displacement, geometric curve, and HRR parameters were considered the primary and input variables to the model, respectively, and by using the numerical simulations in different scenarios, the effect of the many parameters on the critical velocity was evaluated. The estimation methods were used to estimate the critical velocity in the second step by creating a physical model and simulating different scenarios. According to the obtained results, the coefficient of determination for the multiple linear regression, non-linear, and the artificial neural network was 0.84, 0.95, and 0.97, respectively, and the root-mean-square deviation was 0.082, 0.045, and 0.02 that indicate better performance of the artificial neural network model and is selected as the critical velocity estimator model. In order to evaluate the estimator model, in the third step, the actual fire data in two large-scale tunnels were compared to the final model's results. The estimator model's results estimated the critical velocity of 2.46 meters per second in the Memorial tunnel fire by approximately 0.08 decreasing difference. The critical velocity was determined to be 2.55 meters per second in the Runehamer tunnel fire by approximately 0.05 increasing difference. By using the obtained results, it can be found that the estimator model is reliable to a great extent and has shown an acceptable estimating ability.