Vibration-based structural health monitoring technique using statistical features for data stability assessment and damage localization

Abstract

A statistical vibration-based damage identification algorithm to assess the stability of the measurement data, detect and locate damage in civil structures, where variability in response and modal parameters due to measurement noise and environmental influence is often inevitable, is presented in this paper. The algorithm utilizes the statistical correlation of the magnitudes of frequency response function (FRF) of target sensors relative to a reference sensor to assess the consistency of the data while the slopes of lines of fit are exploited for damage localization. Through numerical simulation and experimental investigation of a flexural structure using accelerometers and "point" strain gauges, and long gauge fiber Bragg gratings (FBG) sensors, the importance of the technique for civil SHM is established and presented in an easy-to-interpret graphical format for effective implementation of results. The proposed method using long-gauge FBG sensors is suitable for practical civil SHM with limited number of sensors and where variability in response and modal parameters due to measurement noise and environmental influence is often inevitable. Also, the ability to effectively manage and make sense of enormous amounts of data collected under continuous monitoring process for an effective diagnostic and/or prognostic system is an added advantage.

Identification of Damage in Reinforced Concrete Columns Under Progressive Seismic Excitation Stages

“…The mode shapes are normalized with respect to the measured response at the FBG sensor furthest away from the maximum moment point at the base. The detailed procedure of normalization of modal macro strain (MMS) vectors can be obtained from Li and Wu [2008], and the change in normalized MMS algorithms can be found in Adewuyi and Wu [2010b] and Wu and Adewuyi [2009].…”

Quantification of cracks in beams on the Pasternak foundation using Haar wavelets and machine learning

“…The statistical methods, such as correlation functions and coefficients, autoregressive models, machine learning methods, properly address the modern-type vibration-based methods. The description of each method can be found in a great number of articles, for example, [19,32,43,46]. Mahmoud and Kiefa [25] investigated a steel cantilever beam with a single edge crack.…”