Comparison of Damage Diagnosis Algorithms on a Spatial Frame Using Vibration Data

Abstract

Structural damage can be induced by short-term abnormal stresses as well as long-term natural aging. Detection of changes in a structure's ability to withstand subsequent loads can aid decisions on safety, repair, rehabilitation, and demolition. Dynamic property shifts can show internal cracks and minor damage before propagation or failure occurs, but only if a proper indicator is selected. In order to evaluate potential damage indices, a model three-story metal frame building was constructed and incrementally damaged. An as-built baseline was used for both sequential and cumulative comparisons to ten scenarios with removed bracing. Identical data sets are mined via twelve algorithms using six unique damage indicators, and each algorithm's performance is evaluated. Of all implemented schemes, frequency response function subtraction using frequency content differentials as a direct indicator is the most accurate damage detection scheme for the three-story test structure.