Performance investigation of fire protection and intervention strategies: Artificial neural network-based assessment framework

Abstract

This study aims to develop an integrated framework to investigate and assess fire safety strategies, including protection and intervention. The proposed framework is also able to predict potential fire impacts and recommend optimal fire strategies; a case study of British Columbia was considered. The results showed that 58% of the fires were detected by the visual sighting; besides, smoke alarms were not activated for 2% of the fires. Handheld extinguisher, sprinkler system, standpipe and hose systems were able to extinguish 17%, 4%, and 3% of the fires, respectively. Moreover, 30% of the fires were found to have damages beyond the building of origin. Regarding the dollar loss, 26% of the fires caused a very high level of building dollar loss, 20% had a high level of content dollar loss, and 18% of the fires caused a high level of total dollar loss. The investigated variables are found to be statistically significant, with a p-value < 0.05. The proposed ANN model was able to predict fire impacts very strongly, with R2 equal to 85%. Finally, the optimal set of predictors for minimal fire impacts were determined. These results assist decision-makers to enhance and invest in specific fire safety strategies accordingly.