Técnicas para obtenção de arrays lineares esparsos usando algoritmo genético
Carregando...
Arquivos
Data
2018-08-31
Autores
Souza, Julio Cesar Eduardo de
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O ultrassom pode ser utilizado para gerar imagens aplicadas aos ensaios não destrutivos (END) e diagnóstico médico, em que arrays são utilizados para obter imagens com melhor resolução lateral e contraste em relação a um único transdutor. No entanto, os arrays precisam ter o centro do seus elementos espaçados por uma distância (pitch) menor que λ/2, em que λ é o comprimento de onda gerado pelo transdutor, para que as imagens geradas por estes não apresente artefatos causados pelos lóbulos de espaçamento. Porém, para evitar circunstâncias como o aumento na complexidade eletrônica e tempo elevado para geração de imagens ultrassônicas, os arrays esparsos podem ser utilizados, os quais possuem os elementos espaçados por uma distância maior que λ/2 e as imagens geradas apresentam artefatos causados por lóbulos de espaçamento. Contudo, quando os arrays esparsos são utilizados, existem diversas combinações entre os pitches dos elementos para produzir imagens com diferentes qualidades, inviabilizando o teste de todas as combinações possíveis. Assim, neste trabalho, foi utilizado o algoritmo de busca genética para encontrar configurações de arrays esparsos que gerem imagens com boa qualidade, sendo proposto duas novas funções aptidão para avaliar os arrays esparsos. A primeiraé baseada no diagrama de radiação, e a segunda, na comparação entre PSFs (Point Spread Function). As configurações de arrays esparsos encontrados pelo algoritmo genético foram comparadas aos arrays esparsos disponibilizados na literatura utilizando o diagrama de radiação, PSF e imagens geradas a partir de dados experimentais disponibilizados pelo CSICConsejo Superior de Investigaciones Científicas para que se pudesse realizar uma análise da qualidade dos arrays encontrados. De modo geral, as imagens geradas pelos arrays obtidos pela metodologia desenvolvida neste trabalho apresentaram resolução lateral similar e artefatos com menor intensidade que as imagens geradas pelos arrays propostos na literatura e, ao comparar o array esparso encontrado pelo algoritmo genético com o array com mais elementos cujas distâncias são iguais a λ/2, o array esparso conseguiu gerar as imagens em menor tempo e com boas características. Deste modo, os resultados apontam que a metodologia desenvolvida no presente trabalho conseguiu encontrar configurações de arrays esparsos com qualidade adequada para a avaliação de falhas em ensaios não destrutivos e diagnóstico médico.
Ultrasound can be used to generate images for nondestructive testing and medical diagnostic, in which arrays are used to generate images with better lateral resolution and contrast in comparison to a single transducer. In order to avoid artifacts in ultrasonic images caused by grating lobes, arrays need to have the center of their elements spaced by a distance (pitch) less than λ/2, where λ is the wavelength generated by the transducer. However, in order to avoid electronic complexity and high time to create an ultrasonic image, sparse arrays can be used, which their elements have a pitch greater than λ/2 that generates images with artifacts caused by grating lobes. Although, sparse arrays have their elements with different pitches that produce images with different qualities which it makes impossible to test all combinations. Thus, in this work, the genetic search algorithm was used to find sparse arrays that generate images with good quality. In addition, two new fitness functions were proposed. The first one based on the beam pattern and the second one in the comparison of two PSFs (Point Spread Function). The quality of the sparse arrays found by the genetic algorithm was then compared to the sparse arrays proposed by different authors using the beam pattern, PSF, and images generated from experimental data provided by CSIC (Consejo Superior de Investigaciones Científicas). In general, the images generated by the arrays obtained by the methodology developed in this work presented similar lateral resolution and artifacts with less intensity than the images generated by the arrays proposed in the literature and when images generated by the sparse array found by the genetic algorithm were compared with images generated by arrays with more elements whose distance are equal to λ/2, the sparse arrays were able to generate images in less time with good characteristics. Therefore, the results show that the methodology developed in this work was able to find sparse arrays for the evaluation of failures in nondestructive tests and medical diagnosis.
Ultrasound can be used to generate images for nondestructive testing and medical diagnostic, in which arrays are used to generate images with better lateral resolution and contrast in comparison to a single transducer. In order to avoid artifacts in ultrasonic images caused by grating lobes, arrays need to have the center of their elements spaced by a distance (pitch) less than λ/2, where λ is the wavelength generated by the transducer. However, in order to avoid electronic complexity and high time to create an ultrasonic image, sparse arrays can be used, which their elements have a pitch greater than λ/2 that generates images with artifacts caused by grating lobes. Although, sparse arrays have their elements with different pitches that produce images with different qualities which it makes impossible to test all combinations. Thus, in this work, the genetic search algorithm was used to find sparse arrays that generate images with good quality. In addition, two new fitness functions were proposed. The first one based on the beam pattern and the second one in the comparison of two PSFs (Point Spread Function). The quality of the sparse arrays found by the genetic algorithm was then compared to the sparse arrays proposed by different authors using the beam pattern, PSF, and images generated from experimental data provided by CSIC (Consejo Superior de Investigaciones Científicas). In general, the images generated by the arrays obtained by the methodology developed in this work presented similar lateral resolution and artifacts with less intensity than the images generated by the arrays proposed in the literature and when images generated by the sparse array found by the genetic algorithm were compared with images generated by arrays with more elements whose distance are equal to λ/2, the sparse arrays were able to generate images in less time with good characteristics. Therefore, the results show that the methodology developed in this work was able to find sparse arrays for the evaluation of failures in nondestructive tests and medical diagnosis.
Descrição
Palavras-chave
Algoritmo genético, Array esparso, Dinamic time warping, PSF, Diagrama de radiação, Genetic algorithm, Sparse array, Beam pattern