When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.
Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.
Structural Health Monitoring (SHM) plays a vital role in assessing in-situ the performance of a structure. There are several techniques available for System Identification, Damage Detection and Model Updating. SHM based on the vibration of structures has attracted the attention of researchers in many fields such as: civil, aeronautical and mechanical engineering. This Proposal focuses on the state-of-the art methods for Vibration Testing, Modal Analysis, Model Updating, and Damage Detection in structures. The objective of the thesis is to develop efficient methods in the above areas of SHM using ambient vibration testing. To develop and verify the proposed methods, several case studies are to be developed and implement new technique in multi setup merging by use of Random Decrement Technique. The method for system identification comprises time domain (Stochastic Subspace Identification), frequency domain (frequency domain decomposition) and time-frequency analysis (Complex continuous morlet wavelet) techniques. In addition, the preprocessing methods are required for some certain tests like ambient vibration where signal to noise ratio and vibration amplitude are quite low. The random decrement Technique (RDT) is a time domain procedure, where the structural responses to operational loads are transformed into random decrement functions, which are proportional to the correlation functions of a system’s operational responses, which can be considered equivalent to free vibration responses. Ambient vibration test of a structure usually produces noisy response and the existing modal identification techniques such as Frequency Domain Decomposition (FDD) and Stochastic Subspace Identifications (SSI) techniques often fail to produce accurate results in such case. Due to contamination of ambient vibration with white Gaussian noise, preprocessing may be required to clean up the vibration signal in order to detect the modal properties accurately. For Operational Modal Analysis (OMA), the multi-setup (or roving sensors) measurement techniques are required due to less number of sensors as compared to the number of degrees of freedom (DOFs) in a structure. The technique is based on selection of DOFs as reference, which is fixed during all measurements, and the rest of sensors are roved in each setup. There are three common methods have been used for merging multi-setup data include of Post Separate Estimation Re-scaling (PoSER) Post global Estimation Re-scaling (PoGER) and Pre Global Estimation Rescaling (PreGER). Due to some difficulties in PoSER and PoGER method such as high number of identification, pairing, pole averaging and so on, PreGER have been used recently as an alternative method because the merging process have been conducted before parameter estimation.
Damage in structural system is an important concern as it weaken the structure and reduces its functional capacity, and may even cause failure. In the past decades, special attention has been given to avoid the sudden failure of structural components by damage detection at an early stage. More specifically, SHM based on the vibration of structures has been explored by many researchers in order to obtain very efficient tools of great importance for the civil, aeronautical and mechanical engineering communities. Some damage detection technique such as, frequency based and mode shape based are applied in this proposal, the mode shape based method include Modal Assurance Criterion (MAC), Mode shape curvature, flexibility Matrix, uniform flexibility shape curvature, damage index method and Genetic Algorithm.
The measured vibration response of a structure and the results from its finite element (FE) model often some differences because of the idealization and assumptions in representing the structural system, material properties, support conditions etc in the FE model. There are a number of existing methods for updating such FE models and identifying system parameters such as stiffness and mass based on dynamic parameters response of a structure. The methods for model updating and damage detection have been classified into Physics-based methods where a mathematical model (e.g. FE model) of a structure needs to be constructed, and data-driven methods where an explicit FE model is used, only the data pattern are used for identifying the changes is a structure. The aim of this research is to compare the physics-based and data driven methods of model updating using the data from ambient vibration tests. The matrix update method has been utilized as a physics-based method of model updating, while the Artificial Neural Network has been used as data-driven methods for updating or correlating the FE models. The matrix update method correlates the models by solving the inverse problem using constrained optimization. The data-driven method, Neural Network was used to find the relationships between the structural frequencies and change in the sectional properties of the structure.