Representation of the Impact of Smoke on Agent Walking Speeds in Evacuation Models

This paper addresses the problem of reproducing the effect of different visibility conditions on people’s walking speed when using evacuation models. In particular, different strategies regarding the use of default settings and embedded data-sets are investigated. Currently, the correlation between smoke and walking speed is typically based on two different sets of experimental data produced by (1) Jin and (2) Frantzich and Nilsson. The two data-sets present different experimental conditions, but are often applied as if equivalent. In addition, models may implement the same data-sets in different ways. To test the impact of this representation within evacuation tools, the authors have employed six evacuation models, making different assumptions and employing different data-sets (FDS+EVAC, Gridflow, buildingEXODUS, STEPS, Pathfinder and Simulex). A simple case-study is simulated in order to investigate the sensitivity of the representation of two key variables: (1) initial occupant speeds in clear conditions, (2) extinction coefficients. Results show that (1) evacuation times appear to be consistent if models use the same data-sets and interpret the smoke vs speed correlation in the same manner (2) the same model may provide different results if applying different data-sets or interpretations for configuring the inputs; i.e. default settings are crucial for the calculation of the model results (3) models using embedded data-sets/assumptions require user expertise, experience and understanding to be employed appropriately and the results evaluated in a credible manner.