Vibration-based damage detection on a multi-girder bridge superstructure

Vibration-based damage detection (VBDD) techniques have been proposed as a potential form of structural health monitoring with which an entire structure can be evaluated simultaneously using relatively few sensors. Since these methods rely on the identification of small changes in dynamic properties (notably natural frequencies and mode shapes) to infer the existence and the location of damage, reliable estimates of these properties are essential for the successful implementation of VBDD schemes.;The research described in this thesis was primarily focused on an experimental investigation of the application of VBDD on a multi-girder bridge superstructure, with the objectives of identifying the most reliable test procedures, developing VBDD techniques that could be used for identifying the presence of damage, and evaluating the performance of VBDD techniques for such structures. The experimental investigation was supplemented by theoretical analyses and numerical verifications. The structure used for this investigation was a one-third scale model of a slab-on-girder composite bridge superstructure featuring four steel girders supporting a steel-free concrete deck, based on the North Perimeter Red River Bridge in Winnipeg, Manitoba.;The experimental tests were conducted in a well-controlled laboratory environment. Forced dynamic excitation was supplied by means of a feedback-controlled hydraulic shaker. Instrumentation used to measure the dynamic response included a closely-spaced grid of accelerometers mounted on the surface of the deck along the girder lines, as well as electrical-resistance foil strain gauges bonded to the girder webs. Damage cases investigated included damage to the steel girders, to the diaphragm members, to the lateral steel straps, and to the concrete deck.;A damage detection indicator was developed based on mode shapes that had been normalized to enclose an area of unity. The resulting area under the plot of the difference between two independently measured mode shapes was then used as the damage indicator. To demonstrate the features and verify the capability of the newly developed damage indicator in the absence of experimental uncertainties, a finite element model of the bridge superstructure was developed and used to generate theoretical data for the modal properties.;A database of pairs of independently measured mode shapes, in which both mode shapes in the pair were obtained with the structure in an identical condition, was used to ascertain the variability of the area of mode shape change indicator when different test procedures were followed. This allowed the definition of threshold values of the damage indicator for each set of test procedures, corresponding to the 90th or 95th percentile of the probability distribution of the damage indicator. When the damage indicator exceeds this value, the presence of damage can be inferred with a high level of confidence. A total of 28 different test protocols were investigated, which included two different excitation methods (resonant harmonic and white noise random), four different instrumentation schemes (accelerometers and strain gauges at various locations), and five different vibration modes (the lowest five).;Five commonly available VBDD indicators were also selected to identify the location of damage after its presence had been detected. The performance of the VBDD indicators were examined and evaluated while two different normalization schemes were adopted.;The newly developed damage indicator, the area of mode shape change, was shown to be capable of successfully identifying the presence of damage with a high confidence level using both numerical and experimental data.;Of the 28 test protocols investigated, those that used forced harmonic excitation in combination with the fundamental vibration mode consistently resulted in the lowest threshold values for the area of mode shape change, and therefore resulted in the highest sensitivity to the presence of damage.;The presence of most single damage scenarios could be identified with a relatively high confidence level (at least 90%) when harmonic excitation was used, regardless of the instrumentation scheme used to measure the mode shape (except for the scheme that used data from strain gauges near the top flanges of the girders), since the area changes in the fundamental mode shape due to damage exceeded the corresponding threshold values. Among the instrumentation schemes investigated, both acceleration measurements and measurements of flexural strains near the bottom flanges of the girders were able to identify the presence of damage with a high level of confidence.;In general, all VBDD methods selected could localize the damage investigated with varying degrees of accuracy when the fundamental mode was used, as long as the presence of damage had previously been detected with a high confidence level.