【作者】 王德勇；

【导师】 廖光煊； 朱霁平；

【作者基本信息】 中国科学技术大学 ， 安全科学与工程， 2013， 博士

【摘要】 消防救援对火灾的控制作用显著,主要体现在消防出动时间和消防战斗时间,前者反映消防救援的及时性,后者反映消防救援的有效性。本文基于江西省2000--2010年火灾数据,在空间维度上分析了消防出动时间与消防战斗时间的内在联系,以及消防战斗时间与火灾损失之间的关联规律；在时间维度上分析了消防出动时间和消防战斗时间与火灾损失的时间依存关系；最后综合分析了空间和时间两个维度上消防出动时间和战斗时间与火灾损失之间的量化规律,并在此基础上发展了建筑火灾应急救援序贯决策模型。在消防时间的空间维度统计规律方面,发现不同火灾场所消防出动时间存在差异,这种差异主要是其地理位置造成的。战斗时间受火灾场所和城市区域双重因素的影响,表现在市区和县城火灾的战斗时间显著小于集镇和郊区,厂房和仓库火灾的战斗时间大于其他场所,消防出动时间和战斗时间之间存在一定的相关性,出动时间15分钟内,平均战斗时间随出动时间线性增长,出动时间大于15分钟后,平均战斗时间随出动时间增长趋势逐渐变缓。在消防时间与火灾损失的关系方面,发现不同战斗时间下城市建筑火灾“频率—过火面积”满足幂律分布,幂函数指数可表征控火能力,指数的绝对值越大则小火发生概率越大,大火发生概率越小。该值与战斗时间负相关,表明控火能力随着战斗时间的增长而降低。不同火灾场所幂函数指数存在差异,随着战斗时间的增大,住宅的控火能力衰减最快,公共娱乐场所、商业场所和厂房次之,仓库衰减最慢。不同城市区域火灾的“频率—过火面积”幂律分布存在差异,城市市区、县城城区和郊区农村满足“一段式”幂律分布,而在集镇镇区火灾满足“两段式”幂律分布。在时间维度上,发现建筑火灾过火面积及消防时间均存在时间标度性特征,且随着阈值的变大,分形特征逐渐消失,时间序列过渡为泊松分布。根据艾伦因子,过火面积≥200m2的城市建筑火灾表现为泊松分布,当出动时间大于60分钟后,其时间标度性消失,当战斗时间大于110分钟时,其时间标度性消失,表现为泊松分布。进一步进行地时间序列分析表明：周平均出动时间的滞后阶数为1阶时,是火灾的Granger原因,而周平均战斗时间在滞后4阶时是火灾的Granger原因；从长期来看,平均出动时间、平均战斗时间与平均过火面积变化之间存在均衡关系,且均为正稳定关系；平均出动时间和平均战斗时间的突变均会对平均过火面积产生正的影响,且影响持续时间约为1个月；相较于平均战斗时间,平均出动时间对平均火灾损失的贡献度更大。敏感性分析表明对于不同火灾场所,住宅的出动时间对过火面积最敏感,而战斗时间的敏感性最小；对于不同城市区域,城市市区的出动时间敏感性最大,同时城市市区和县城城区火灾的过火面积对战斗时间的敏感性要大于集镇镇区和郊区农村火灾。基于上述消防时间关联分析,考虑了应急救援车辆的动态特征、火灾发展的动态特征及火灾发生的时序特征,建立了针对建筑火灾应急救援的动态序贯决策模型,模型引入火灾重要度的概念解决火灾场所和城市区域对消防时间和过火面积的影响,采用双层两目标规划方法,考虑当前并发火灾损失最小和并发火灾总体损失最小两个目标,实现了局部最优和整体最优。实证分析表明本文发展的建筑火灾应急救援决策模型适用于复杂应急救援条件下的救援决策,能够为决策者提供最优的资源调度决策方案。更多还原

【Abstract】 Well performed fire rescue can significantly control fires. This is represented by fire attendance time and fire fighting time. The former reflects the timely ability of fire rescue, while the latter reflects the effectiveness of fire fighting. This paper, based on the urban fire data from2000to2010in Jiangxi province, on the spatial dimension, the inter connection between fire attendance time and fire fighting time and the correlation between fire fighting time and fire loss; on the time dimension the dependency relationship between fire fighting time and fire loss. Finally it summarizes the statistical law based on the two dimensions and developed a sequential decision model of building fire emergency rescue.The differences of the first attendance time at difference places are caused by the different location. Fire fighting time is confined by both the fire location and the urban area, which is manifested in that fire fighting time in the city and county was significantly less than that in towns and suburbs, fire fighting time in workshop and warehouse is more than that in other places. There is a certain correlation between fire attendance time and fire fighting time. If the fire attendance time is within15minutes, the average fire fighting time grows linearly with it, while it expends more than15minutes; the increasing trend of fire fighting time becomes slow.Under different fire fighting time, the city building fire "frequency-burned area" satisfies the power-law distribution. Power function index represents the fire control ability. If the absolute value of the index is large, the probability of small fires is large and the probability of large fires is small. Meanwhile, this value of index has a negative correlation with fire fighting time, that is to say the fire control ability decreased with the growth of fire fighting time. Power function index can be different at different locations. As the fire fighting time increases, the control ability in residence decreased fastest, which followed by the public entertainment places, commercial places and factories. The control ability in warehouse decreases the slowest."Frequency-burned area" power law distribution of regional fire in different cities act differently. Urban areas of cities and county and suburban rural area meet "the one part form" power law distribution, while that of township fire follows "the two part form" power law distribution.Burned area and fire attendance time of building fire have time scale characteristics. With the increasing of the threshold, the fractal characteristic gradually disappears, and the time series transports to Poisson distribution. According to Allen factor, if the burned area is larger than200m2, city building fire performance follows Poisson distribution. When the fire attendance time is larger than60minutes, the time scale disappears and performed as Poisson distribution. The same presents when fighting time is more than110minutes.Time sequence analysis showed that:the average weekly fire attendance time becomes the Granger effect of fire if it lags one order, the average weekly fire fighting time becomes the Granger effect of fire if it lags4orders. In the long run, there is an equilibrium relationship among the average fire attendance time, the average fire fight time and the average burned area and all of the relationships are stable. Average fire attendance time and average fire fighting time positively influence average burned area for about a month. Compared with the average fire fighting time, the average fire attendance time contributes more to the average fire loss. Sensitivity analysis shows that burned area is most sensitive to the attendance time in residential places. And the fire fighting time has the least effect on the burned area. For different city zone, the burned area is most sensitive to fire attendance time in city urban area. And the sensitivity of burned area in city and county urban area to fire fighting time is greater than that of fires in the rural town and suburban town.Based on the correlation analysis of fire attendance time, this paper has set up a building dynamic sequential decision model of fire emergency rescue, considering the dynamic characteristics of the emergency rescue vehicle, fire development and the temporal characteristics of the fire presence. This model introduces the concept of fire important degree, which can solve the impact of fire location and urban areas on fire rescue time and burned area. And it adopted a double layer two goal programming method, considering the current minimum concurrent fire loss and minimum concurrent fire total loss as two targets to achieve local optimum and global optimum. Empirical analysis shows that the building fire emergency rescue decision model developed in this paper is suitable for rescue decision under complex emergency rescue condition, and can provide optimal resource scheduling decision scheme.更多还原