Abstract

Vibration-based techniques for Structural Health Monitoring (SHM) utilize the dynamic response of a structure measured using a set of sensors to identify the modal properties and potential structural deterioration or damage. Signal processing tools are widely used for analyzing and diagnosing these response signals obtained from a structure. Any changes in the dynamic characteristics of a structure indicates deterioration in the structure. However, a direct comparison of the vibration signals or modal properties at different periods of time may not be sufficient to identify the damages and their locations. Therefore, it is essential to analyze the vibration signals to extract the morphologies of the changes in these response signals and correlate them with the possible location of structural damage. Anomalies can exist in a structure’s response in the form of damage due to loss of stiffness in the structural and non-structural elements and these anomalies alter a structure’s behaviors and modal properties. For some decade researchers have been using experimental and operational modal analysis to estimate modal parameters to measure structural damages in experimental and operational conditions respectively employing vibration-based response of structures. In case of ambient vibration data, processing and estimating the exact location of damage are complex for being non-stationary and non-linear. In this work, a novel method is developed where response signals obtained from structures are decomposed into Intrinsic Mode Functions (IMF) using Empirical Mode Decomposition (EMD) technique. Those IMFs are then processed with Hilbert-Huang transform (HHT) to obtain their corresponding Hilbert Spectra (HS), which allows the estimation of the time-varying instantaneous properties of those response signals. Then for system identification, Singular Value Decomposition (SVD) is performed to obtain the Singular Hilbert Spectra (SHS), and for damage detection, Marginal Hilbert Spectrum (MHS) is estimated. This leads to estimation of coefficients to calculate associated damage indices (DI). Simultaneously, modal frequencies of the structure are obtained from time-frequency-amplitude domain plot of the Singular Hilbert Spectra (SHS). Subsequently, Finite Element (FE) model is constructed for the healthy and damaged structure and tuned moment of inertia of the assembled members, which correlate with the natural frequencies obtained during testing. Joint displacements obtained from the linear modal analysis of the healthy and damaged structures lead to estimation of Displacement Mode Shape (DMS) and Curvature Mode Shape (CMS), and their absolute difference between the healthy and damage structure help to locate damage from the model. Thus, a hybrid method comprising MHS-based damage index along with numerical model-based joint curvature mode shape is proposed. This proposed method is verified using experimental tests conducted on: (a) a cantilever steel beam, (b) five-storey scaled frame, and (c) an eight-storey full scale steel building. The present thesis shows that the results obtained from system identification and damage detection manifest the advantages of the HHT-based technique for health monitoring of structures. It is observed from the results that the damage can be localized effectively in single and multiple damage cases from the DI values obtained using the proposed method. Simultaneously, modal frequencies are estimated as a by-product of the damage detection algorithms. The present investigation indicates that MHS is a promising tool for SHM to assess any deterioration or anomaly by comparing with a baseline model. Data-driven techniques for SHM has been mostly limited to small scale tests in laboratories. To address the challenges in application of data-driven methods for real life structures, it is necessary to incorporate physics-based analysis of structure along with the non-model-based method. For data-driven methods it is absolutely necessary to employ modern instrumentation methodologies to obtain vibration-response from structures. To highlight the present work and advantages of the modern wireless sensors for vibration measurement, two different case studies are presented: (a) modal analysis of pre-cast concrete shear wall with wired Piezotronics sensor, and (b) modal analysis of sixteen-storey Concordia University EV building with modern wireless sensors and data acquisition (DAQ) wirelessly connected with android device. The wireless vibration sensors are found to be quite easy and flexible to use for capturing the frequency response of a structure to detect a range of modal frequencies. However, cost of those sensors could pose a challenge and positioning of optimum of sensors is important in testing of tall buildings or long span bridges. To mitigate the problem of limited number of sensors, multi set up and roving sensors methods were used in the past. That could be time consuming and error prone. Further research is required to make the process more efficient. Manual estimation of frequency band width from the SHS plot, complexity in estimating DI from MHS and time expensive manual iteration in the numerical modeling are some of the limitations for the current work.