Study On Uncertainty Of Occupant Safety Egress Under Building Fires

There are many uncertainties in the building fire protection design process due to the duality of fire occurrence and development. As one of important aims of building fire protection design, the occupant egress safety also contains uncertainty. Many uncertain parameters are involved in the calculation of ASET (Available Safety Egress Time) and RSET (Required Safety Egress Time). Currently, uncertainties often are always ignored or handled with conservative values or safety factors in the calculation of ASET and RSET. These uncertainties can not be quantified by an effective method. To properly deal with these uncertain parameters is a key technique deserved studying, also is the basis for credibility and acceptability of the design and assessment results by people. The systematical method for uncertainty analysis of occupant safety egress in single-room building is discussed in this dissertation. The primary studies are demonstrated as following:Occupant safety egress uncertainty analysis method based-on MonteCarlo simulation is established. Based on fire dynamics, the time prediction models associated with ASET and RSET are constructed: prediction model of the time at which hot smoke layer reach critical height (i.e. ASET) and prediction model of detection time. The uncertain parameters associated ASET and RSET are discussed. These uncertain parameters are described by probability distributions. The uncertainties of ASET and RSET were also investigated by MonteCarlo simulation technique. The results of the case study demonstrate that it works well in analyzing the uncertainty of occupant safety egress.We developed the parameter sensitivity analysis method of occupant safety egress. To reduce the amount of uncertain parameters and decrease uncertainty analysis workload, and guarantee the precision of uncertainty analysis at the same time, we utilized the single-factor sensitivity analysis method to investigate sensitivity of individual parameter at the specific benchmark value. Based on the results of sensitivity analysis, the parameters with lower sensitivity were treated as constants. Then, the uncertainty analysis was carried out. A case study is presented, firstly, the uncertainty of ASET and RSET are studied respectively with all uncertain parameters considered. Then, the parameter sensitivities were analyzed by the sensitivity method reported by us. Thus the primary uncertain parameters with higher sensitivity can be found out. We investigated how they affected ASET/RSET through increasing the number of them. When all the primary uncertain parameters are considered, the calculated results were consistent with that when all the uncertain parameters were considered. It suggested that the method of occupant safety egress parameter sensitivity is correct and efficient.The reliability analysis model was established, which mainly includes: (1) using reliable probability (reliability index) as a measure to describe the safety level of occupant egress under building fire. The principle of occupant safety egress reliability analysis is introduced here, and the reliability is consequently deduced. Aiming at the defects of FOSM calculation method in accuracy and efficiency, this dissertation presents a reliability index calculation method based on PSO optimization method. (2) For present researches, the distributions of parameters are assumed as theoretical distribution, which might be inconsistent with the actual range of parameter in reality. Moreover, the reliability of egress is always less than 1. Aiming at this problem, this dissertation adopts truncated probabilistic distribution to describe the uncertain parameters with considering actual ranges of parameters, and the probability density function (PDF) of truncated probabilistic distribution was thus deduced. We present the random numbers generation method based on the PDF of truncated probabilistic distribution and establish occupant egress reliability Monte- Carlo calculation method for the case that the parameters follow truncated probabilistic distributions. (3) The sensitivity analysis method of reliability index to uncertain parameter was established, which can be used to find the parameters with more significance before calculation of reliability index by comparing the sensitivities of parameters. The parameters with smaller sensitivities can be treated as constants (benchmark value), which can effectively reduce the number of uncertain parameters and decrease calculation load during the uncertainty analysis. A case study was presented. The results demonstrate that the parameter sensitivity analysis method is effective and practicable.