published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

University of Macau / Faculty of Science and Technology. / Department of Computer and Information Science

Jianhua Chen. / Thesis (Ph.D.)--Chinese University of Hong kong, 1998. / Includes bibliographical references (p. 127-[134]). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Mode of access: World Wide Web. / Abstracts in English and Chinese.

Dissertação (Mestrado em Engenharia Mecânica) - Centro Universitário da FEI, São Bernardo do Campo, 2014

Tese (doutorado)—Universidade de Brasília, Faculdade de Tecnologia, Departamento de Engenharia Civil e Ambiental, 2015. / Nos últimos anos, percebe-se uma constante preocupação da comunidade científica pela busca de técnicas eficientes para identificação de danos em estruturas. Isto resultou no desenvolvimento progressivo da Área de Integridade Estrutural. Neste sentido, o objetivo deste trabalho é contribuir com mais testes estáticos e dinâmicos, com uma metodologia de identificação de danos e com um índice de dano para auxiliar no processo de identificação de danos em pontes rodoviárias e ferroviárias de aço e de concreto armado. Para isto, são escolhidos os métodos baseados em wavelets. Tais métodos podem detectar singularidades presentes nos parâmetros modais ou deslocamentos causados pelo dano e, consequentemente, não requerem a condição da estrutura antes do dano. Além disso, podem ser aplicados tanto em resposta estáticas quanto dinâmicas. Foram realizados testes experimentais e análises numéricas em vigas e em um modelo reduzido de uma ponte ferroviária em aço visando obter as respostas estáticas e dinâmicas, para em seguida, aplicar as Transformadas de Wavelet e o índice de dano proposto baseado na curvatura da energia dos coeficientes de wavelet. A partir dos estudos realizados, concluiu-se que a metodologia proposta utilizando as transformadas de wavelet após a interpolação e a regularização dos dados e o índice de dano proposto podem ser utilizados como uma alternativa às técnicas tradicionais de detecção de danos, visto que as mesmas foram capazes de localizar a posição do dano para diversas situações. / In recent years, it can be observed a constant concern of the scientific community by the search for efficient techniques for damage identification of structures thus enabling the progressive development of the Structural Health Monitoring (SHM). In this sense, the objective of this work is to contribute with more dynamic and static tests, with a damage identification methodology and a damage index to help the damage identification process in beams, roadway and railway bridges of steel and reinforced concrete. For this, methods based on wavelets were chosen. Such methods may detect singularities present in the modal parameters or displacements caused by the damage and therefore do not require the condition of the structure before damage. Furthermore, it can be applied both in static as well as dynamic response. Experimental and numerical tests were carried out on beams, in a scale model of a steel railway bridge and a real reinforced concrete bridge to obtain the static and dynamic responses, to then apply the Wavelet Transform and the proposed damage index based on the curvature energy of the wavelet coefficients. From the studies, it was concluded that the proposed methodology using wavelet transforms after interpolation and smoothing in the data and the proposed damage index could be used as an alternative to traditional techniques for detection of damage, since they were able to detect the position of the damage for many situations.

A thesis submitted to the Faculty of Science, School of Computational and Applied Mathematics University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Doctor of Philosophy. Johannesburg, South Africa, July 2015. / Reinforcement learning agents solve tasks by finding policies that maximise their reward over time. The policy can be found from the value function, which represents the value of each state-action pair. In continuous state spaces, the value function must be approximated. Often, this is done using a fixed linear combination of functions across all dimensions. We introduce and demonstrate the wavelet basis for reinforcement learning, a basis function scheme competitive against state of the art fixed bases. We extend two online adaptive tiling schemes to wavelet functions and show their performance improvement across standard domains. Finally we introduce the Multiscale Adaptive Wavelet Basis (MAWB), a wavelet-based adaptive basis scheme which is dimensionally scalable and insensitive to the initial level of detail. This scheme adaptively grows the basis function set by combining across dimensions, or splitting within a dimension those candidate functions which have a high estimated projection onto the Bellman error. A number of novel measures are used to find this estimate. i

Wavelets are functions that are useful for representing signals and approximating other functions. Wavelets sets are defined in terms of Fourier transforms of certain wavelet functions. In this paper, we provide an introduction to wavelets and wavelets sets, examine the preexisting literature on the subject, and investigate an algorithm for creating wavelet sets. This algorithm creates single wavelets, which can be used to create bases for L2(Rn) through dilation and translation. We investigate the convergence properties of the algorithm, and implement the algorithm in Matlab.

Involuntary muscle contractions (spasms) are a major secondary consequence of spinal cord injury. These spasms disrupt mobility and the ability to perform daily activities. The rhythmic repetitive muscle contractions of clonus are one kind of spasm. In this study an algorithm was developed to automatically detect the start and end times of EMG bursts during clonus. These measures were used to calculate the duration of EMG bursts, clonus frequency and the intensity (root mean square) of each EMG burst, parameters that characterize clonus. This algorithm relied on the technique of intensity analysis (Von Tscharner 2000). Filters were created by non-linearly scaling a Mother (Morlet) wavelet to produce envelopes of the EMG in different frequency bands. The intermediate frequency band (80-190 Hz) enveloped the EMG best and was used to detect the EMG bursts during clonus. To detect the EMG bursts, an intensity threshold and time separation threshold were imposed on the algorithm to eliminate multiple peaks caused by the baseline EMG, motor units or EMG changes. Window regions were extended between the midpoints of identified EMG peaks then resized to 50 ms on either side of each identified EMG peak. The start and end times of EMG bursts were at 5% and 95% of the energy contained in a window region, respectively. A motor unit threshold constraint was used to eliminate motor unit potentials at the beginning and end of clonus. The algorithm output from 31 spasms in long term (24 hr) EMG data recorded from 8 paralyzed leg muscles of 7 subjects with a chronic cervical spinal cord injury were compared to that generated by two independent human operators. The algorithm was as good as a human operator at identifying EMG bursts (p = 0.946), clonus frequency (intra class correlation coefficient (p = 0.949), contraction intensity (p = 0.997) and the durations of each burst of EMG during clonus (p = 0.852). On average the algorithm was 574 (SE 238) times faster than manual analysis by two people (p <= 0.001). Analysis of clonus in one 24 hour dataset from the right medial gastrocnemius muscle with the algorithm showed that clonus was more prevalent and stronger during awake versus sleep time. This algorithm can be used to analyze long term recordings accurately with limited user intervention. The algorithm may also be a prospective diagnostic tool to judge the effectiveness of interventions such as drugs like baclofen that are used to mitigate clonus.

The ability to efficiently and sparsely represent seismic data is becoming an increasingly important problem in geophysics. Over the last thirty years many transforms such as wavelets, curvelets, contourlets, surfacelets, shearlets, and many other types of ‘x-lets’ have been developed. Such transform were leveraged to resolve this issue of sparse representations. In this work we compare the properties of four of these commonly used transforms, namely the shift-invariant wavelets, complex wavelets, curvelets and surfacelets. We also explore the performance of these transforms for the problem of recovering seismic wavefields from incomplete measurements.

No description available.