Sigma Xi Lecture

James J. Filliben
Statistical Engineering Division, NIST
Green Auditorium, Administration Building
Teusday, November 22, 2005, 10:30 AM

Abstract

The Congressionally-mandated NIST Federal Building and Fire Safety Investigation of the World Trade Center Disaster is coming to completion. The buildings' degradation immediately prior to collapse was extremely complicated, with structural, thermal, dynamic, and stochastic interdependencies existent across both time and space. Four pre-collapse stages (a simplification of reality) will be discussed: aircraft impact, fire spread, thermal propagation through insulation, and structural deformation. Engineering issues and the statistical methodologies to address these issues will be discussed. A major challenge in the statistical analysis of the World Trade Center was the relatively meager amount of data--little physical evidence existed on important events occurring in the core of the WTC buildings. In this regard, the study was both assisted and complicated by reliance on computational engineering virtual data--primarily in the form of NIST's FDS (Fire Dynamics Simulator) and phase-specific FEA (finite element analysis) computational models. As analyses progress from component to sub-assembly to global models, such computational models require characterization, sensitivity analysis, and validation--it will be shown how statistically designed experiments played a major role in this regard. Various other statistical analysis techniques (e.g., complex demodulation for assessing post-impact building oscillation frequency and--indirectly--building damage) will also be discussed. This talk will emphasize the statistical methodologies employed. Detailed engineering conclusions and recommendations resulting from the Federal Building and Fire Safety Investigation of the World Trade Center Disaster are presented in the investigation's (10,000 page) final report, due for current release.