Study On Uncertainty Of Occupant Evacuation Time In Fire Safety Design

In fire safety engineering design, prescriptive-based and performance-based fire protection methods are widely adopted. However, due to the unpredictability of fire occurrence and human behavior, evacuation time is subjected to a high degree of uncertainty. Building design parameters and other various indices are specified in prescriptive-based building codes that lack flexibility. In performance-based building fire protection design, safety factor-based approach and conservative method are adopted to deal with the uncertainties. However, the selection of safety factors is to a degree dependent on the fire experts, to be further verified. In addition, it is difficult to carry out a cost-effective building fire protection design by adopting the conservative method. In order to solve the problems mentioned above, the methods for the uncertainty analysis of evacuation time and optimization of exit parameters for evacuation time under uncertainty are proposed, which is based on the polynomial chaos expansion in this thesis. The main research is focused on as follows.Due to the complexity of occupant evacuation, there are various factors that affect significantly evacuation time. According to the analysis for the evacuation process and the degree of difficulty in the controllability of parameters, parameters associated with evacuation time are divided into uncertain parameters and exit parameters. Uncertain parameters of evacuation time can not be exactly determined due to the randomness of fire and coccupant evacuation, such as the number of evacuees and pre-movement time. And exit parameters of evacuation time are associated with geometric characteristics of exits, such as exit width, exit spacing, the number of exits and exit location.In order to reduce the computational burden for uncertainty quantification of evacuation time, a method coupling computer evacuation models to polynomial chaos expansion is proposed. In order to demonstrate the proposed method, a fire compartment of commercial buildings is presented as a building evacuation environment. And the results suggest that the proposed method can take a reduced significantly computational cost to quantify the uncertainty of evacuation time, whose accuracy is similar to that from the Monte Carlo simulation of computer evacuation models. Meanwhile, uncertainties of evacuation time, which involves the uncertainty of pre-movement time and does not consider the uncertainty of pre-movement time, are compared. It can be found that the safety factor is not increased with the number of uncertain parameters.In addition, in order to take a low computation cost to quantify the effect of uncertain parameters and exit parameters on evacuation time, a method for the sensitivity of evacuation time to parameters is proposed, which combines the polynomial chaos expansion and variance decomposition method. The linear sensitivity indices of evacuation time to parameters are analyzed based on the first order polynomial chaos expansion of evacuation time. Due to the nolinear nature of evacuation system, the nonlinear sensitivity indices of evacuation time to parmaters are studied based on the second order polynominal chaos expansion of evacuation time. And the proposed method of the sensitivity of evacuation time to parameters is compared with the sample-based methods for sensitivity analysis. Through the sensitivity analysis of evacuation time to parameters in one single fire compartment, it can be obtained that the results from the proposed sensitivity analysis method is consistent to that from the sample-based sensitivity analysis method, whose computational cost is reduced significantly. When the exit width is not large enough, the sensitivity of evacuation time to uncertain parameters is affected significantly by exit width. However, when occupant density is large, the sensitivity of evacuation time to exit parameters is affected less by occupant density.Furthermore, in order to ensure the safety performance, economy and functional fitness of buildings with low a computational burden, the method coupling the polynomial chaos expansion of evacuation time to the two-stage nested Monte Carlo simulations is proposed to determine the optimal value of exit parameters under uncertainty. The exit parameters considering in this thesis are exit width, exit spacing and exit location. Through the analysis of occupant evacuation in one-room fire compartment, it can be seen that the proposed techniques for optimization of evacuation time under uncertainty can be used to determine the optimal exit width, exit spacing and exit location under uncertain parameters, whose computational cost is reduced significantly. With the increase in the reliability of the value for evacuation time, the minimum evacuation time will increase. For different reliability of the value for evacuation time, the minimum exit width is obviously different. The mean and standard deviation of evacuation time will decrease significantly with the increase in exit width. However, when exit width is large enough, the mean and standard deviation of evacuation time is affected less by exit width, which can regarded as constants. When exit width is designed in the range of0.1-0.5m for every100persons, exit width has a small effect on safety factor of evacuation time and the reliability of a certain safety factor of evacuation time. The exit location which is determined by minimizing the maximum travel distance is not optimum for evacuation time. With the increase in the reliability of the value for evacuation time, the optimum exit spacing tends to increase, and the difference between the optimal exit location and exit location determined from minimizing the maximum travel distance is increased. When exit spacing is small, exit spacing has a small effect on the safety factor. However, when exit spacing increases to a certain value, the minimum safety factor of evacuation time at certain reliability level will begin to decrease. Furthermore, the safety factor corresponding to a certain reliability of the value for evacuation time has a tendency to decrease with the increase in the difference between the optimal exit location and exit location determined from minimizing the maximum travel distance.