| This research consisted of identifying functional requirements for supporting vulnerability assessment, developing a formalized vulnerability representation schema, and developing an approach to reason about building systems and contents to support vulnerability assessment in building emergencies. Requirements elicited in this research are related to documenting, identifying and prioritizing contents, threats, building systems and/or spaces in order to support a set of core vulnerability assessment activities, identified in literature and in three case studies. Requirements were implemented in a prototype system, namely System for Intelligent Vulnerability Assessment (SIVA), and presented to ten subjects. SIVA fully supported 39 out of 52 user queries. The representation schema captures to what threats specific contents are vulnerable. It enables the representation of threats and how they can be traced in space, as well as a description of contents and content collections. The representation schema was tested for coverage in synthetic tests, using OmniClass for contents and the Emergency Response Guidebook for threats. The schema was also tested for extensibility in user tests interacting with SIVA. Such a representation schema is needed for supporting automated vulnerability assessment in building emergencies. The reasoning approach enables a first responder to perform flexible searches, as well as prioritize critical spaces and pieces of equipment that need to be protected in an emergency, leveraging existing building and content representations found in Building Information Models (BIM). This approach targets retrieving and prioritizing vulnerable content. It consists of a hybrid faceted-retrieval mechanism, based on three facets: threat, cost and location, in addition to graphs to represent networks of contents. The approach was tested for precision and recall in retrieving vulnerable content in two scenarios with ten experts interacting with SIVA. Results show increased precision and recall (97.7% and 91.8%, respectively) when all facets are used in the search, as compared to lower precision and recall when only retrieval facets are used or prioritization facets and graph are used. These results demonstrate that the hybrid approach more effectively retrieves and prioritizes contents that need to be protected in building emergencies. |
