Frequency-based structural damage identification and dynamic system characterisation

Abstract

This thesis studies structural dynamic system identification in a frequency-based framework. The basic consideration stems from the fact that frequencies may generally be measured with higher accuracy than other pertinent modal data such as mode shapes; however only a limited number of frequencies may be measured in the conventional context of natural frequencies. Being able to measure extra frequencies is a key to the success of a frequency-based method. The main part of the thesis is therefore organised around the involvement of the so-called artificial boundary condition (ABC) frequencies to augment the frequency dataset for general structural damage identification. In essence, the ABC frequencies correspond to the natural frequencies of the system with additional pin supports, but may be extracted from specially configured incomplete frequency response function matrix of the original structure without the need of physically imposing the additional supports. In the first part of the research, a particular focus is placed on the actual extraction of these ABC frequencies from physical experiments through effective modal testing, data collection, data processing and analysis. The influences of key processes involved in a typical modal experimental procedure, including high-fidelity measurement of the (impact) excitation input, averaging, windowing, and an effective use of post-processing techniques, particularly the Singular Value Decomposition (SVD) technique, are scrutinised in relation to the extraction of the ABC frequencies. With appropriate implementation of testing and data processing procedures, results demonstrate that all one-pin and two-pin ABC frequencies from the first few modes can be extracted with good quality in a laboratory setting, and the accuracy of extracted ABC frequencies is comparable to natural frequencies of corresponding orders. A comprehensive study is then carried out to investigate the sensitivities of ABC frequencies to damages. Two-pin ABC frequency sensitivity is formulated by extending the expression of anti-resonance sensitivity. On this basis, the mode shape contribution is adopted as a criterion for the selection of more sensitive ABC frequencies to be employed in detailed parameter identification or finite element model updating procedures. The soundness of using ABC frequencies in structural parameter identification and the effectiveness of the above ABC frequency selection method are subsequently examined through case studies involving laboratory experiments and the corresponding FE model updating. Furthermore, a preliminary study is carried out to examine the possibility of formulating ABC frequency-based damage indicator, herein with an analogy to the mode shape curvature, for direct damage assessment. As an extended investigation in the general framework of frequency-based dynamic identification, in the last part of the thesis, a complex dynamic system, namely a railway bridge under moving loads & masses, is evaluated with regard to the various frequency characteristics involved. The variation of the natural frequencies of the bridge-moving mass system, as well as the presence of the apparent frequencies from the trainloads, are analysed in detail. Besides simplified theoretical analysis, a computational model is developed to simulate the combined bridge-moving vehicle/train system, where the vehicle mass is coupled with the bridge via surface contact. The model is verified by comparison with field measurement data and theoretical predictions. Parametric studies enable a clear identification of the correlation of the frequency contents between the response and the trainload, and provide new insight into the significance of the so-called driving and dominant frequencies. It is found that much of the dynamic response phenomena, including the resonance effect, may be explained from the view point of the frequency characteristics of the trainload pattern, which is governed primarily by the ratio between the carriage length and the bridge length. Finally, a resonance severity indicator (the Z-factor) is developed for the assessment of the resonance effect in the railway bridge response when the trainload moves at a resonance speed. Numerical results demonstrate that the proposed methods are effective for the determination of the critical speed and the resonance effects, including the situations where a significant carriage mass is incorporated.