Statistical approach for structural damage diagnosis using bootstrap hypothesis testing

Abstract

In this paper, a statistical approach for structural damage diagnosis that uses uncertain frequency response functions (FRFs) was presented. Structural damage was detected from the changes in FRFs from the original intact state. The measurements are always contaminated by noise, and sufficient data often are difficult to obtain; those making difficult to detect damage with a finite number of data. To surmount this, we introduced the bootstrap hypothesis testing to statistically prevent identification error due to measurement noise. The proposed method iteratively zooms in the damaged elements by excluding the elements which were assessed undamaged from damage candidates step by step. The proposed approach was applied to numerical simulations using 2-dimensional frame structure and its efficiency was confirmed