Estudos sobre a estrutura e propriedades de fibras de coco e curauá do Brasil
Resumo
Resumo: O desenvolvimento de materiais obtidos de fontes renováveis tem sido uma necessidade corrente nos últimos anos para minimizar os problemas ambientais do mundo. O emprego destes materiais vem sendo objeto de diferentes estudos e pesquisas visando sua aplicação potencial como substituto em materiais derivados do petróleo, devido ao alto custo destes últimos, impacto ambiental e poluição gerada. As fibras naturais vêm apresentando um grande potencial de aplicação em diversos setores industriais. Por isso, a compreensão de suas propriedades e entendimento delas em termos de sua estrutura é de fundamental importância para definição de novas aplicações. Neste trabalho, são avaliadas as fibras de curauá e coco do Brasil em termos das propriedades físicas, mecânicas e comportamento térmico, e a simulação das propriedades mecânicas através de redes neurais artificiais. Os valores de tensão na ruptura e módulo elástico apresentam redução com o aumento do diâmetro, em ambas as fibras. Nas duas fibras em estudo, o aumento da distância entre garras levou a redução da tensão e deformação na ruptura, enquanto aumento do módulo elástico para ambas as fibras. Por outro lado, através do aumento da velocidade de ensaio, ambas as fibras apresentaram um aumento da tensão de ruptura. O módulo elástico e deformação na ruptura não sofreram variação com o aumento da velocidade em nenhuma das fibras. As fibras de curauá apresentaram maior resistência mecânica, representada pela maior tensão de ruptura (500 – 3000MPa) contra o coco (100 a 250MPa) e pelo maior módulo elástico 50GPa contra 3GPa. As fibras de coco apresentam maior deformação na ruptura 30% quando comparado ao curauá (4%). A análise térmica das fibras mostrou os processos de degradação dos constituintes das fibras (celulose, hemicelulose e lignina), enquanto que a análise dinâmico-mecânica mostrou que a rigidez das fibras é bastante influenciada pela presença de água. A difratometria de raios-X das fibras permitiu avaliar a cristalinidade das fibras naturais, pontando uma cristalinidade de 40% para as fibras de coco e de 50,2% para o curauá. As propriedades mecânicas das fibras naturais foram simuladas através de redes neurais artificiais (RNA). Foram empregadas redes com 13 neurônios em 3 camadas ocultas, usando como parâmetros de saída as propriedades mecânicas (tensão de ruptura, alongamento na ruptura e módulo elástico). Os sinais de entrada empregados foram a o ângulo helicoidal, quantidade de celulose, cristalinidade da fibra, condições de ensaio (velocidade e distância entre garras) e diâmetro das fibras. Também, duas das propriedades mecânicas foram empregadas como parâmetros de entrada enquanto era simulada a terceira. Para cada propriedade simulada, uma nova rede neural foi definida, com a mesma arquitetura que as outras. A correlação encontrada entre os dados experimentais e os dados simulados foi maior que 90% para todas as pro riedades. Abstract: The development of new materials obtained from renewable sources has become a need over the last decade in order to mitigate the environmental issues of today’s world. The search for this kind of materials has been subject of many researchers aiming for the replacement of the materials derived from petroleum in view of their high cost, environment impact and pollution they create. The potential of applications of natural fibers has been increasing in several manufacturing sectors. Therefore, it is of great importance to understand its physical and other properties and to know its structure in order to explore new applications. This work oriented towards these aspects of characterization. The present studies concentrates on the structure and properties of the Brazilian curauá and coconut fibers in terms of their physical, mechanical properties and thermal behavior and the mechanical properties simulation by artificial neural networks. It is observed that the values of both Young modulus and tensile strength decrease with increasing diameter of the fiber in the case of both the fibers. In the case of the two fibers studied, it is observed that increase in gauge length, both the tensile strength and strain at break decreased, while the Young modulus increased for both the fibers. On the other hand, with varying strain rate, higher tensile strengh is observed with increasing strain rate. But, both Young modulus and strain at break were not influenced by the strain rate. Curauá fibers showed higher mechanical resistance, as can been seen by higher tensile strength (500 – 3000MPa), while the strength values of 100 to 250MPa is observed for coir. Also, higher Young modulus is observed for curauá (50GPa), while it is 3GPa for coir. Coir fiber showed higher strain at break 30% against 4% for curauá fiber. The thermogravimetric analysis of the fibers demonstrated the decomposition profiles of the constituents of the fibers (namely, lignin, cellulose and hemicelluloses), while the dynamic-mechanical analysis showed that the stiffness of the fibers is greatly influenced by the uptake of water. The X-rays diffraction spectra revealed the crystalline behavior of both the natural fibers, showing a 40% crystallinity index for the coconut fibers and 50,2% for the curauá. The mechanical properties of these two lignocellulosic fibers were simulated through Artificial Neural Networks (ANN). Three ANNs with thirteen neurons in three hidden layers were used; each one had a mechanical property as an output (Strain at break, Tensile strength and Young modulus). The inputs were the microfibrilar angle, cellulose content, fiber crystallinity, test conditions (strain rate and gauge length) and fiber diameter. Also two of the strength properties were used as inputs while the third one was simulated. For each of the simulated properties, one new ANN was defined, with the same structure as the others but with different inputs and outputs. The correlation found etween the experimental data and the simulated data was better than 90% for all the properties.
Collections
- Teses & Dissertações [9149]