A Comparison Between Predicted and Actual Behaviour of Domestic Smoke Detectors in a Realistic House Fire

Abstract

This report describes an investigation into various methods for predicting the activation times for domestic ionisation and optical smoke detectors when detecting flaming fires. The prediction methods studied were the temperature correlation method where the detector is assumed to activate when the gas that surrounds it rises a certain temperature above ambient, the pseudo-heat detector method in which an RTI is assumed for the smoke detector, and Heskestad's method which is based on optical density. The activation times predicted by these methods were then compared with those that were recorded during actual fire tests within a full size two-storey test house. The house was modelled on the field modelling software FDS. Two test fires were considered in this study. Both were lounge chair fires in a room on the ground floor of the house, and were flaming fires under vitiated conditions. The energy output reached by the two fires was between 600 and 1000 kW. Data collected during these tests included gas temperatures, optical density and the mass loss history of the chairs. There were also a number of radiators throughout the house that were left on for some time to heat the house before the fires were ignited. It was found that there was little difference between the temperature correlation and the pseudo-heat detector methods when a value for RTI of 1 m½ s½, was considered. It was also found that the use of an activation temperature of 20°C above ambient within the temperature correlation method was the most appropriate method for predicting the activation times for the detectors. This value for activation temperature is high compared with other researchers, but it is likely that the radiators within the house created airflow currents that had an effect on the behaviour of the detectors.