Study On Human Behaviour In Building Fire

This paper is the main content of the co-project with City University of HK entitled: Research on evacuation models in building fires. The author is the principal investigator for this co-project.Human behaviour in building fire is the key point of performance-based fire safety design and assessment of buildings. The study will help to learn the characteristics of the human behaviour in building fire and enlarge human behaviour database in China. This work provides practical fundamental for the performance-based fire safety design and assessment of buildings in China, as the performance-based fire safety design and assessment of buildings has been used gradually in the world.This work focuses on the human behaviour in fire in China. Based on the existing studies of the human behaviour in fire, post-fire survey on fire evacuation in buildings have been first performed on the 169 fire incidents in Shaanxi, Shanxi and Henan province in China. 274 questionnaires were collected. Frequency distribution statistics was conducted about their personal information, environmental information and human behaviour information in fire. Analysis has been done on the human behaviour and reaction duration of the investigated persons inside the buildings with different ages, different gender, different educated level, different fire experiences, different location inside buildings, different activities, different health conditions and different approaches of fire alarm. Three highest distribution values of the first human behavior were concluded about fire reporting, fire identifying and fire warning to others. This paper have first obtained the range of reaction duration and the characteristics of frequency distribution under different situations of the investigated persons inside the buildings in China.This paper has first used the artificial neural network (ANN) software, Neuroshell4.0, to study the human behaviour in building fire in China. Based on the principle of that whether the human behaviour is benefit to fire evacuation or not, seven input variables of training sample and one output variables were first normalized. Fire reporting, fire identifying and fire extinguishing selected as the human behaviour output variables, this work has developed the 50 iterative error bar charts through 50 iterative training checkouts. The training results of the artificial neural network (ANN) have been analyzed. The network training results are acceptable for the highly discrete human behaviour in building fire.Comparative studies have been first preformed on the human behaviour in building fire under the Chinese culture background, compared with the research literature under the UK, USA, HKSAR and Japan culture background. Evaluating the research in broader range, this paper has concluded the commons and differences characteristics of the human behaviour in building fire under different culture background. The research provides references for the performance-based fire safety design and assessment of buildings in China.