Determinação do tamanho de nanopartículas de polimetacrilato de metila utilizando redes neurais artificiais
Artificial neural networks are computational systems that can be used to solve complex math and engineering problems, being mostly used in non-linear problems. Recently, the introduction of the artificial neural network area into the nanotechnology has been increasing. Nanotechnology is an area with...
| Autor principal: | Baumbach, Fernanda Pegoraro |
|---|---|
| Formato: | Trabalho de Conclusão de Curso (Graduação) |
| Idioma: | Português |
| Publicado em: |
Universidade Tecnológica Federal do Paraná
2020
|
| Assuntos: | |
| Acesso em linha: |
http://repositorio.utfpr.edu.br/jspui/handle/1/11511 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: |
Artificial neural networks are computational systems that can be used to solve complex math and engineering problems, being mostly used in non-linear problems. Recently, the introduction of the artificial neural network area into the nanotechnology has been increasing. Nanotechnology is an area with vast applications due to the specific characteristics of the nanoparticles, especially its size in the range of 100nm. It is crucial for the application of nanoparticles that they meet the requirement of having they size in the 100nm range, therefore, the use of artificial neural networks for the prediction of particle size may has an huge contribution to the area. In this context, this study aimed the training of artificial neural networks using data from MMA polymerization, where various parameters of the reaction were variated, for example, the surfactant and the initiator. I was also aimed to predict the final particle size using the better trained network varying the initial amount of surfactant used in the reaction. The neural network was built with varied parameters. In total there were 48 structures where the transfer functions in the hidden and output layers were varied along with the number of neurons in the hidden layer and the training function. A single type of neural network was used, the feedforward network, with 13 parameters in the input layer and one in the output layer, being it the medium particle diameter. Amongthe trained networks, the best performance was the one trained with the algorithm Levenberg-Marquardt backpropagation, using 20 neurons in the hidden layer and logsig function on the hidden layers along with a linear function in the output. Mostly, the networks with the best performance were the ones trained with the Levenberg-Marquardt backpropagation and Resilient Backpropagation algorithms. Finally, the prections using the best performance network had results in accordance with the literature. |
|---|