Please type a search term (at least two characters)
Statistical methods for conformity assessment when dealing with computationally expensive systems: application to a fire engineering case study
Ensuring public safety in shopping malls, car parks and theatres in the event of a fire is an essential part of building safety regulations. Safety strategies include reducing the likelihood of fires starting or limiting its spread by using fire suppression systems, such as sprinklers. This approach is supported by a second line of defence which consists of ensuring that occupants are not trapped in the smoke and may escape easily to a place of refuge. It is therefore vital that emergency exit signs remain visible for as long as possible.
Fire safety engineering is the application of science and engineering principles to protect people and their environments from the destructive effects of fire and smoke. In a building fire smoke can accumulate very quickly potentially obscuring exit signs and preventing safe evacuation. Computer models based on complex mathematical simulations and computational fluid dynamic tools are used to evaluate new building designs, and check the reliability and practicality of evacuation paths under different emergency conditions.
Using fire modelling to improve safety
Fire simulation modelling is complex with hundreds of possible parameters that effect how a fire generates heat or spreads. Numerous complex simulations are required, and each may take weeks of computer time to reach a conclusion on whether a building is fire safety compliant. Speeding this process whilst upgrading the number of inputs that models can handle in a single iteration would improve fire safety in large public buildings.
Using statistical modelling to improve safety
The EMRP project Novel mathematical and statistical approaches to uncertainty evaluation improved uncertainty calculations for fire safety evaluations. The project modelled the movement of smoke and the flow of heat, in a building in order to assess its safety and produced accurate short-cuts to reduce the time taken to evaluate a buildings safety without compromising accuracy.
The project team based at the Laboratoire national de métrologie et d'essais (LNE) in France, investigated the use of modelling and statistical approaches to evaluate building safety. Initially simulating the fire of a relatively simple building – a cube - with some doors and exits open. The team then investigated variables, such as the energy released at the start of the fire, the time over which the fire burns, and the environmental conditions (parameters such as the temperature both inside and outside the building, and atmospheric pressure) and how these effect simulation outcomes. Fire engineers gave practical advice to ensure that the statistical boundaries in the simulation were realistic. Conformity criteria were set to see if the simulation would provide information on whether the building was fire regulation compliant or not. By running many simulations based on different scenarios a data base was built up that can aid the prediction of how a fire will evolve in a building, and help to assess the safety of the building viewed as a probability of conformity with an associated uncertainty.
Based on this probabilistic approach, it has been possible to reduce computation times for complex fire safety evaluations, using iterative software to achieve accurate results. This is based on combining smartly designed simulations of increasing reliability (from those that take minutes, to hours to days or weeks). The major goal is to enable the fire safety of a building to be reliably assessed against regulatory requirements using a fast iterative process based on improved simulations.
The approach can also be used to assess which parameters are the most influential on a building’s safety and to predict the probability of conformity of a building and its uncertainty for a particular fire.
The work has been recognised and rewarded with two prices at the French annual conference on computer experiments (GRD Mascot Num) emanating from CNRS, and by the IAFSS (International Association for Fire Safety Science) which reflects the interest of the fire safety engineering community in probabilistic approaches to get more reliable conformity assessment.
The EMRP project Novel mathematical and statistical approaches to uncertainty evaluation used a probability framework to make an assessment on whether building occupants would be able to safely evacuate it in the advent of a fire. The simulation generated is an example of how applying probability and risk assessments to complex mathematical problems can lead to rapid yet credible simulations of safety for regulatory purposes.
EURAMET announces the launch of the Stage 2 EMPIR Call 2017 more
Delivering confidence in construction and the built environment more
Statistical methods for conformity assessment when dealing with computationally expensive systems: application to a fire engineering case study more
New process to prove correctness of calculations with complex, numerical algorithms implemented in measurement software developed by EMRP project NEW0... more