Book/Dissertation / PhD Thesis

FZJ-2017-06759

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Multivariate Methods for Life Safety Analysis in Case of Fire

Schröder, B. (Corresponding author)FZJ*

2017
Forschungszentrum Jülich GmbH Zentralbibliothek, Verlag Jülich
ISBN: 978-3-95806-254-2

Please use a persistent id in citations: http://hdl.handle.net/2128/15349  urn:nbn:de:0001-2017081810

Abstract: The assessment of life safety in case of a building re is based on thecomparison of the available safe egress time ($\textit{ASET}$) and the required safe egress time ($\textit{RSET}$). With regards to simulation experiments, this straightforward approach is accompanied by uncertainties including the underlying models, the specification of inputs, and the analysis of outputs. Concerning the two latter aspects, this thesis introduces methodological extensions in order to conduct $\textit{ASET-RSET}$ analyses in a multivariate fashion. For the specification of inputs, the multitude of possible scenarios is represented with the help of systematic sampling techniques. Uncertainties in terms of analysis are tackled with multi-criterial maps rendering both $\textit{ASET}$ and $\textit{RSET}$ in spacious environments. The subtraction of both maps is used to determine a measure of consequences. These methods are applied to a multi-level underground station which is investigated with numerical simulations based on the formation of two subsystems, namely $\textit{Fire}$ and $\textit{Evacuation}$. The analysis incorporates an ensemble of 8,640 combined fire scenarios and evacuation scenarios. Throughout the entire design space, more than 95 % of the scenario combinations account for less than half of the maximal observed consequences. This analysis is refined by agglomerative clustering in order to group all observations hierarchically. It becomes evident that the lowest margins of consequences are represented by two clusters covering approximately 75 % of all observations. The investigation of the parametric relations of all clusters allows for the systematic identification of the determining characteristics of fire and evacuation scenarios. In addition to the consequence measure derived from $\textit{ASET-RSET}$, fractional effective doses ($\textit{FED}$) are calculated to supplement the analysis. Within the clusters, the number of occupants exceeding common $\textit{FED}$ thresholds applicable to incapacitation corresponds to the introduced $\textit{ASET-RSET}$ measure. However, throughout the entire design space, this correspondence is not clear and needs further investigation.

Note: Bergische Universität Wuppertal, Diss., 2017

---

Contributing Institute(s):

1. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 511 - Computational Science and Mathematical Methods (POF3-511) (POF3-511)

Appears in the scientific report 2017

---

Database coverage:
;

---