Abstract: Buoyancy frequency method, N percentage method, integral ratio method, are three common methods to determine smoke layer height in tunnel fire, the calculations of three methods are often biased from visual observations.In order to study and explain deviations of calculation results from visual observations, which were found in three common determination methods of smoke layer height in tunnel fire, a long tunnel model was established based on FDS, and six types of fire source power were set. Then, vertical visibility distribution was used to verify deviation between existing determination method and visual results before visibility mutation method was proposed. The results show that the deviation calculated fromabove four methods ranks as visibility mutation method (1.63%) < buoyancy frequency method (4.88%) <N percentage method (N=30, 7.88%) < integral ratio method (8.50%), among which results of integral ratio and buoyancy frequency method are higher than visual ones. It is also found that visibility value in vertical direction of tunnel remains constant and then decreases abruptly as the height increases, and the greater fire power is, the faster it declines, and the lower visibility value there is.

Key words: tunnel fire, numerical simulation, fire dynamics simulator (FDS), smoke layer height, visibility distribution