Boustras, Georgios

Two fire growth models for enclosures have been developed using a stochastic and a probabilistic approach. The first is based on probability tree analysis, while the second one is a typical Markov approach to the problem. A state transition model was built in order to provide the appropriate transition probabilities needed by the two models and a deterministic mainframe was used in order to close the set of equations in the probabilistic model. New criteria for the definition of the mean time spent in each of the fire stages are proposed and finally results from both approaches are compared to each other and with previous analysis. Results from a nationwide survey are used to furnish the models. The study concentrates on the development of a new risk assessment framework for use by the Cypriot manufacturing industry and it is funded by the Cyprus Research Promotion Foundation. An infamous catastrophe in the Cypriot manufacturing industry is used to illustrate the examples.

In the last few years a number of catastrophes have taken place in Cyprus; accumulating loses of millions of pounds, destroying valuable ancient pine tree forests and disrupting rural life for days and in a few cases weeks and months. A survey, aiming to gather information and opinions about the current fire safety situation in wildland fires was carried out. The aim of this study is to provide the reader with an adequate idea about the existing situation in Cyprus as far as safety is concerned in wildland areas.

Two fire growth models for dwellings (3 realms) have been developed using a stochastic approach. The first is based on probability tree analysis and the fundamental work of Aoki [1] and Ramachandran [2]. The second one is a typical Markov approach to the problem. A state transition model was built in order to provide the appropriate transition probabilities needed by the two models and a known deterministic mainframe (CFAST 3.1.6) was used in order to close the set of equations in the probabilistic model. New criteria for the definition of the mean time spent in each of the fire stages - 1) fire confined to room or area, 2) fire spread beyond room or area. 3) fire extinction or burn-out) - are proposed and, finally, results from both approaches are compared to each other and with previous analysis. The present models make an attempt to combine the randomness of stochasticity with the accuracy of a deterministic model based on the use of physical laws. As part of a probabilistic fire risk assessment tool, the models can be used to provide information for fire growth in buildings and be combined with a risk to life model to predict the risks to a human being in a fire situation.

This article comes as a natural continuation of the work presented in a previous article in this Journal [6]. In the previous article, two stochastic models of fire growth were presented and results were discussed. A known limitation of stochastic models is that of the integration of an accurate time frame in the equations. A reliable solution to the fundamental problem of the definition of the mean time spent in each of the fire stages has been provided. In this article a realistic fire scenario is presented and a full risk assessment is attempted. Different populations of inhabitants during the event are compared and analytical predictions per room are presented.