The 1994 Northridge earthquake caused brittle fractures in the beam-to-column connections of numerous steel moment frame buildings. Since many buildings had little visible damage, most of these fractures were not detected until architectural finishes and fireproofing were removed in intrusive inspections weeks (or even months) later. Currently, the only way to determine if fractures have occurred in such cases is by intrusive inspection, which can be time-consuming and expensive. However, if a building is seismically instrumented, even sparsely (in some cases with a single tri-axial accelerograph at the roof only), the recorded acceleration response is available and can provide information that may indicate potential building damage. This report explores the possibility of going beyond traditional indicators of building damage that are obtained from accelerograms (such as fundamental period elongation) and examining indicators specific to fracture, impact and other types of sudden structural failure. Specifically, based on the equations of motion, sudden structural failures and impacts are expected to produce a transient dynamic response, both globally and locally, with frequency content higher than that of the predominant building response preceding it, due to the excitation of both member and structure higher modes by the failure/impact. In cases of substantial sudden damage (i.e. numerous connection fractures), this high-frequency, transient dynamic response is expected to be visible in the recorded acceleration response, provided certain conditions are satisfied.
Based on this expectation, a method for detecting steel moment connection fractures using fracture-generated high-frequency transient accelerations is proposed. The method is evaluated using a dataset containing strong motion data from the 1994 Northridge earthquake, and damage and inspection information from 24 steel moment frame buildings with varying degrees of connection fracture damage. The success rate of the method is high for heavily damaged buildings, relatively high for undamaged buildings, and moderate for moderately damaged buildings. However, the method fails for lightly damaged buildings. The success rate improves for moderately and heavily damaged buildings with the removal of records with excessive noise, indicating that too much noise can affect the success of the method. The proposed method is also evaluated in combination with more commonly used damage indicators, which gives a higher success rate at detecting damage occurrence.
Due to the very sparse instrumentation in most buildings in the dataset, the method is not able to determine specific fracture locations, though very general location determination (such as upper floors vs. lower floor) may be possible in the future in buildings with more accelerometers. However, it is anticipated that a very dense instrumentation scheme would be necessary to determine fracture locations down to the individual connections. Such high-density instrumentation schemes are not practical with current technology, though advances in cheaper and/or wireless instrumentation may permit such dense instrumentation in the future.
Download the text of this publication, 88 pages (5.9 MB).
For questions about the content of this report, contact Mehmet Çelebi
| PDF help
| Publications main page
Reports for 2005
This report is available only on the web.