Made available in DSpace on 2014-12-17T15:01:14Z (GMT). No. of bitstreams: 1
AdrianaKVG.pdf: 2228086 bytes, checksum: 5a5ddc6253972ac4ee82c43ae0f08d6c (MD5)
Previous issue date: 2008-07-27 / The petroleum industry, in consequence of an intense activity of exploration and production, is responsible by great part of the generation of residues, which are considered toxic and pollutants to the environment. Among these, the oil sludge is found produced during the production, transportation and refine phases. This work had the purpose to develop a process to recovery the oil present in oil sludge, in order to use the recovered oil as fuel or return it to the refining plant. From the preliminary tests, were identified the most important independent variables, like: temperature, contact time, solvents and acid volumes. Initially, a series of parameters to characterize the oil sludge was determined to characterize its. A special extractor was projected to work with oily waste. Two experimental designs were applied: fractional factorial and Doehlert. The tests were carried out in batch process to the conditions of the experimental designs applied. The efficiency obtained in the oil extraction process was 70%, in average. Oil sludge is composed of 36,2% of oil, 16,8% of ash, 40% of water and 7% of volatile constituents. However, the statistical analysis showed that the quadratic model was not well fitted to the process with a relative low determination coefficient (60,6%). This occurred due to the complexity of the oil sludge. To obtain a model able to represent the experiments, the mathematical model was used, the so called artificial neural networks (RNA), which was generated, initially, with 2, 4, 5, 6, 7 and 8 neurons in the hidden layer, 64 experimental results and 10000 presentations (interactions). Lesser dispersions were verified between the experimental and calculated values using 4 neurons, regarding the proportion of experimental points and estimated parameters. The analysis of the average deviations of the test divided by the respective training showed up that 2150 presentations resulted in the best value parameters. For the new model, the determination coefficient was 87,5%, which is quite satisfactory for the studied system / A ind?stria de petr?leo, em decorr?ncia de uma intensa atividade de explora??o e produ??o, ? respons?vel por grande parte da gera??o de res?duos, os quais s?o considerados t?xicos e poluentes ao meio ambiente. Dentre estes, encontra-se a borra oleosa formada durante as etapas de produ??o, transporte e refino de petr?leo. Este trabalho teve como prop?sito recuperar o ?leo presente na borra oleosa por processo de extra??o, a fim de que este pudesse ser utilizado como combust?vel ou retornar em alguma corrente do processo de refino. A partir dos ensaios preliminares foram selecionadas as vari?veis independentes que exercem maior influ?ncia no processo de extra??o, s?o elas: temperatura, volume de solvente, volume de ?cido e tempo de extra??o. Inicialmente, determinou-se uma s?rie de par?metros para caracterizar a borra oleosa. Posteriormente, projetou-se um extrator para operar com a borra de petr?leo. Foram aplicados dois planejamentos experimentais: fatorial fracionado e Doehlert. Os ensaios foram realizados em processo batelada, de acordo com as condi??es dos planejamentos experimentais aplicados. Atrav?s dos par?metros de caracteriza??o constatou-se que o res?duo oleoso ? constitu?do predominantemente de material org?nico (36,2% de ?leo), 16,8% de cinzas, 40% de ?gua e 7% de compostos vol?teis. A efici?ncia m?dia do processo de extra??o foi de 70%. Entretanto, a an?lise estat?stica mostrou que o modelo quadr?tico n?o se ajustou bem ao processo, indicando um baixo coeficiente de determina??o (60,6%). Isto ocorreu devido ? complexidade do material estudado. Para obter um modelo que melhor se ajustasse aos resultados obtidos experimentalmente, utilizou-se a ferramenta da modelagem matem?tica, redes neurais artificiais (RNA), a qual foi gerada, inicialmente, com 2, 4, 5, 6, 7 e 8 neur?nios na camada oculta, 64 dados experimentais e 10000 apresenta??es (intera??es), verificando-se menores dispers?es entre os valores experimentais e calculados para o n?mero de 4 neur?nios. Com base na an?lise dos desvios m?dios do teste e treinamento evidenciou-se que o n?mero de 2150 apresenta??es foi o melhor valor considerando a propor??o de pontos experimentais e par?metros estimados. Para o novo modelo, o coeficiente de determina??o foi de 87,5%, mostrando-se bastante satisfat?rio
| Identifer | oai:union.ndltd.org:IBICT/oai:repositorio.ufrn.br:123456789/15746 |
| Date | 27 July 2008 |
| Creators | Guimar?es, Adriana Karla Virgolino |
| Contributors | CPF:08580772800, http://lattes.cnpq.br/2621516646153655, Melo, Josette Lourdes de Sousa, CPF:10200720406, http://buscatextual.cnpq.br/buscatextual/visualizacv.do?id=K4787094H6, Pacheco Filho, Jos? Geraldo de Andrade, CPF:32681089468, http://lattes.cnpq.br/6315186407922891, Teixeira, Ant?nio Carlos Silva Costa, Chiavone Filho, Osvaldo |
| Publisher | Universidade Federal do Rio Grande do Norte, Programa de P?s-Gradua??o em Engenharia Qu?mica, UFRN, BR, Pesquisa e Desenvolvimento de Tecnologias Regionais |
| Source Sets | IBICT Brazilian ETDs |
| Language | Portuguese |
| Detected Language | English |
| Type | info:eu-repo/semantics/publishedVersion, info:eu-repo/semantics/masterThesis |
| Format | application/pdf |
| Source | reponame:Repositório Institucional da UFRN, instname:Universidade Federal do Rio Grande do Norte, instacron:UFRN |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0026 seconds