Back to results list
Show full item record
Please use this identifier to cite or link to this item:
|Title:||A diagnostic imaging technique for identification of structural damage using hybrid features of Lamb wave signals||Authors:||Zhou, Chao||Degree:||Ph.D.||Issue Date:||2011||Abstract:||Diagnostic imaging using Lamb waves, such as Lamb wave tomography, has been attracting increasing preference because it yields an easily interpretable and intuitional quantitative map concerning the overall 'health' state of the structure under inspection. It is however envisaged that to construct a tomogram is often at the expenses of using a large number of sensors to scan the entire area carefully, therefore limiting its application for online structural health monitoring (SHM). To enhance the practicality of Lamb-wave-based SHM in conjunction with use of imaging techniques, alternative methods have been actively explored, exemplified by a recent technical breakthrough: probability-based diagnostic imaging (PDI). A PDI method attempts to describe a damage event using a binary grey-scale image. The field value at a particular pixel of the image is linked with the probability of damage presence at the spatial point of the monitored structure that corresponds to this pixel. Presenting damage in terms of its presence probability is an improvement over traditional nondestructive evaluation (NDE) or SHM which has an ultimate goal to define damage with deterministic parameters (e.g., location coordinates, size or length). That is because the underlying concept of 'probability' is more consistent with the implication of 'predicting' or 'estimating' damage. Nevertheless majority of the current PDI approaches, substantially based on canvassing changes in temporal information such as time-of-flights (ToFs) extracted from captured wave signals, fail to portray damage quantitatively including its orientation, shape and size, because difference in damage orientation, shape and size would not lead to pronounced changes in temporal information. In addition, when dealing with orientation-specific or sharp-angled damage featuring a dominant size in a particular dimension (e.g., a crack or a notch), a challenging issue is that such sort of damage often exerts strong directivity to wave propagation. As a result, information associated with damage may not be efficiently extracted from signals captured by sensors at certain locations, in the absence of prior knowledge of damage shape. Further down to a fine level, the periphery of structural damage such as delamination can be described as the continuum of a number of cracks with various lengths which shape the damage, although macroscopically the damage often presents relatively smooth boundaries. That implies ascertainment of orientation of individual cracks can facilitate depiction of damage shape and further severity. The effectiveness of most existing damage identification approaches has been verified by identifying singular damage in a structure. However, in practice an engineering structure is likely to evidence multiple instances of damage simultaneously. Rather than being a simple expansion from the approaches developed for singular damage, qualitative or quantitative identification of multiple damages is another tanglesome issue.
In this thesis work, a novel PDI approach was developed with an aim to circumvent the above-addressed deficiencies of current NDE and SHM. Systematic investigations were divided into the following three parts: First, three fundamental issues relevant to Lamb-wave-based-PDI were interrogated: a) effect of the orientation of damage with sizable length in a particular dimension (orientation-specific damage) on Lamb wave propagation, b) attenuation of Lamb wave as propagation and its compensation, and c) influential area of damage on wave sensing path. Both finite element (FE) simulation and experimental validation were conducted and results were compared. Results arising from these fundamental studies served as the knowledge basis to develop PDI approach in this study. Second, based on the correlations established in the above fundamental studies between (1) damage location and ToFs extracted from signals, and (2) intensity of signal energy scattered by damage and damage orientation, a PDI approach was developed, in conjunction with use of an active sensor network in conformity to a pulse-echo configuration. Relying on signal features including both the temporal information and signal intensity, and with the assistance of a two-level synthetic image fusion scheme, such an imaging approach was capable of indicating clearly the orientation of individual damage edges and further the shape of damage. The effectiveness of this approach was demonstrated by predicting selected orientation-specific damage including a triangular through-thickness hole (through FE simulation) and a through-thickness crack (through experiment) in aluminium plates. Nevertheless, as observed, when applied to detection of damage consisting of multiple edges of different orientations, this approach became unwieldy. Last, to overcome the deficiency of the above PDI approach using sole pulse-echo configuration (e.g. not capable of depicting damage of different orientations), improvement was made using hybrid signal features extracted from both the pulse-echo and pitch-catch configurations in the active sensor network. To supplement the possible insufficiency of signal features for high-precision identification, a novel concept of 'virtual sensing' was established, facilitating extraction of rich signal features. A hybrid image fusion scheme was developed, able to enhance the tolerance of the PDI approach to measurement noise and possible erroneous perceptions from individual sensing paths. The approach was validated by predicting representative damage scenarios including L-shape through-thickness crack (strong orientation-specific damage), polygonal damage (multi-edge damage) and multi-damage in aluminium plates. Accurate identification results for typical damage cases have demonstrated the effectiveness of the developed PDI approaches in quantitatively visualising structural damage, regardless of its shape and number, by highlighting its individual edges in an easily interpretable binary image.
Structural health monitoring.
Hong Kong Polytechnic University -- Dissertations
|Pages:||xx, 179 p. : ill. ; 30 cm.|
|Appears in Collections:||Thesis|
View full-text via https://theses.lib.polyu.edu.hk/handle/200/6190
Citations as of Jul 3, 2022
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.