【作者】 姜蓬；

【导师】 廖光煊； 蒋勇；

【作者基本信息】 中国科学技术大学 ， 安全技术及工程， 2009， 博士

【摘要】 火灾事故造成的损失不断增大,火灾调查的重要性不言而喻。但是,火灾调查工作中的一些调查和分析方法仍然停留在经验和半经验的水平,火调工作人员的经验和水平等诸多因素均可能对火灾调查结论的准确性产生影响。同时,随着公民法律意识的增强,也对火灾调查提出了更高的要求,如何提高火灾调查中分析方法的准确性和说服力,如何从理论和定量分析的层次开展火灾调查工作,为火灾调查研究提出了新的要求。为改进电气火灾事故调查中常用的金相分析方法,本文引入数字图像处理方法对金相图进行分析和处理,采用和改进了人际交互式半自动图像分割算法Live-wire,并利用成功提取金相图中的各晶格的轮廓。借用数字形态学方法对提取出的晶格轮廓图进行分析和处理,有效去除了金相图中冗余信息,并将每一个晶格单独提取出来。接着,本文分析和讨论了多种描述子参数对实现晶格分类的作用和影响,并最终确定采用离心率、圆形度和晶格标记方差三种描述子作为晶格类型自动分类的主要参数依据,采用聚类分析方法,对两幅金相图中的共40个晶格进行分类,结果表明,能够对大部分的晶格实现正确分类,但分类中仍存在一些问题。接着,引入fisher判别法,对聚类分析结果进行再处理,结果表明,fisher判别法能最终实现对金相图中胞状晶和柱状晶两大类晶格的自动分类,准确度较高。随后,本文从理论角度出发,结合物理实验,对火灾调查中常用的火灾图痕方法进行了分析研究。本文第四章对火灾烟气在壁面黏附的机理做出分析,分析了火场中烟气颗粒的受力模型,分析和模拟火场中烟气颗粒与壁面碰撞和黏附动力学行为,并提出临界逃逸速度概念,用于判断烟气颗粒是否能够黏附在壁面上。随后,本文第五章开展实验研究,研究火源功率,燃烧时间对壁面烟熏图痕的影响,结果表明,随着火源功率的增大,烟熏图痕覆盖的范围逐渐增大,相同位置的烟熏浓度也逐渐变大,同时,烟熏图痕的结构特征也发生变化:较小火源功率情况下,烟熏图痕由V型痕和倒U型痕构成,火源功率增大后,在原烟熏图痕的上部逐渐出现漏斗型烟熏图痕,同时,烟熏浓度不断增大,继续增大火源功率,图痕中部的倒U型痕逐渐被V型痕和漏斗型图痕共同覆盖,整个图痕可以认为是由典型的V型痕和漏斗型图痕构成。将实验结果与模拟结果进行对比,发现,第四章提出的烟熏图痕预测模型能较好的预测特定燃料和特定壁面材料的烟熏图痕,精确度较高。本文第六章以实际发生的某古民居电气火灾事故为研究对象,借助第二章研究方法,分析研究现场提取铜导线残留物的金相图,确定火灾原因为电热作用引起,排除了电路短路的可能性。继而,采用计算机模拟手段,对火灾过程进行数值重构,重现了火灾过程,得到了现场勘测所无法得到的诸多信息。分析和研究了火场中关键部位温度、烟气浓度及FID指数等参数的发展变化趋势,经与证人证言的对比,证明重构结果与实际火灾场景相符,同时,提出针对该类建筑的消防改进方案,并对改进后的方案进行火灾预测,将预测结果与实际火灾过程进行对比,证实了改进方案的有效性和可行性,完善和拓展了火灾调查结论。更多还原

【Abstract】 It is important to conduct fire investigation research since fire cost more and more damage.It is unlucky that most investigation and analysis methods are still experience based which means that there is little standard for judgment.Different investigators may make different conclusions on the same fire.At the same time, improvements of legal awareness and idea of the public also ask for precision and clarity of the fire investigation.It is important to improve technology and method of fire investigation and find new way to make correct conclusion based on a tighter work.To improve metallographic analysis in electrical fire investigation,digital image processing methods are adopted.A human-computer interaction algorithm which name is Live-wire is used and improved and with this work,crystal outlines are captured.To eliminate off redundancy information and digital date from the metallographic image,digital morphology methods are adopted and so the precise outline of crystals are captured.Then,effects of several segment descriptors on crystal categories classification are analysis and discussed.With the help of principal component analysis,key segment descriptors such as eccentricity,degree of circularity and variance of crystal checkmarks are determined to be used as the final parameters of the automatic classification.Cluster analysis results show that most crystals are digested correctly which there are still little problems in the classification. Some crystals are mis-classificationed.To improve the results,fisher discrimination method is adopted and the results show that cylindrical crystals and afterbirth-like crystals in two different metallographic images are distinguished correctly.To improve theoretical and the formation mechanisms of smoke pattern in fire investigation,the fourth chapter conducts research in the light of mechanical model of a single smoke particle.The kinematical model of smoke particle is given.This chapter also offers a conception of critical escape velocity of the smoke particle. While the velocity of the smoke particle after the collision with the wall is greater than the critical velocity,the particle will escape;while the critical velocity is greater, the particle will stay on the wall.The fifth chapter designed validation and usability experiments,the results show that as the HRR and combustion time of the fire source rise up,smoke pattern change from "V pattern + U pattern" to "V pattern + funnel pattern".By the comparison between numerical simulation results which used particle adhesion model and experimental results,it can be concluded that the particle adhesion model can give predict smoke patterns numerically and the accuracy is acceptable.The sixth chapter conducted a numerical reconstruction of a typical heritage building fire which is caused by electrical malfunction.By using method of the second chapter,metallographic image of copper wires which is gathered from the fire site is processed and the analysis results show that it is electric heat which ignited the combustibles near the socket.By using computational numerical simulation method,a fire reconstruction is conducted.The way fire and smoke spread is given and the accuracy is validated with the comparison with testimony of witness.Temperature and smoke density of key spot of the heritage building are given,and FID parameter is used to help find out how CO and CO₂ kill people in this fire.In order to improve safety and extend available safe egress time,water sprinkles are set in the second numerical case and it can be seen from the results that with the effect of the water sprinkles,temperatures of key spots are much more lower and so the smoke density, people have more time to egress.更多还原