2004

Tillander, Kati

This study is the first relatively broad statistical survey utilising the statistical data collected in the national accident database, Pronto. As a result valuable new information relating to fire risks is obtained and quantitative methods for fire risk assessment of buildings are presented. This work is a step forward in the field of risk-analysis-based fire safety design and overall a step towards a better understanding of the anatomy of fires. The use of statistical information is a good objective way of attempting to characterise fires. This study concentrates on ignition frequency, economic fire losses and fire department operation in the event of building-fires. Ignition frequency was derived as a function of total floor area for different building categories. The analysis showed that the variations of ignition frequency are dependent on initial floor area distributions of the buildings hit by fire and at risk. For engineering design purposes, the generalisation of the theory starting from the initial floor area distributions, leading to a sum of two power laws, was found suitable. The parameters and partial safety coefficients for the model were estimated for three building groups. The model is suitable for determining the ignition frequency of buildings with a total floor area of between 100 and 20 000 m2. The elements describing the fire department operation were analysed on the basis of statistical information. In the presented approach, the buildings in which fire safety depends completely on automatic extinguishing systems can be distinguished from those in which the fire department is able to arrive at the fire scene early enough to have a good chance of saving the building. The most important factor affecting the performance of the rescue force was found to be the travel time to the fire scene. Thus, to make the task easier for the fire department, special attention must be paid to rapid fire detection and locating of the fire seat. Delays in these actions lengthen the total response time and reduce significantly the chances of the fire department successfully intervening in the progress of the fire. Economic losses were considered as consequences of the fires. The analysis showed the dependency of loss and value-at-risk of the building on the floor area. Clear local peaks were detected for both the ignition frequency and fire losses. A more detailed analysis of residential buildings where the phenomenon was most apparent revealed that the peaks were located around the floor-area region where the dominant building type of the building stock, and thus the compartmentation manner, changed. With small values of the total floor area of the building, the rise of the loss was very steep, but levelled off to substantially slower growth with large values. A natural explanation for the behaviour is compartmentation. Both the ignition frequency and the fire losses should therefore be examined in relation to the size of the ignition compartment, which would be a significantly more appropriate descriptor than the total floor area of the building. Hence, it is essential that the information becomes available to the Finnish accident database, in which it is not at the moment included. The analysis shows that the type of building and compartmentation, rather than the material of the load-bearing member itself, was the factor having the greatest effect on the risk of fire. The use of the information gathered was demonstrated through a simple example case in which the fire risk was assessed using the time-dependent event-tree approach. This study concentrates on the utilisation of statistics to collect information and gain an understanding of the elements affecting fire risks in buildings. Many of the methods used are well known in other application areas; the available statistical data