The succinic acid is a microorganism common metabolite used in the food market which is produced exclusively by fermentation, and great attention has been given to the use of renewable rawmaterials for this purpose. This study aimed to determine the variables that influence the production of succinic acid by fermentation using Actinobacillus succinogenes strain (CIP 106512) through afractional factorial design and to test different architectures of artificial neural networks to model this process. Artificial neural networks are made of three layers and were the MultilayerPerceptron (MLP) type, with Backpropagation learning algorithm. Experimental data for learning and testing of networks were used, 13 and 6, respectively. The number of neurons in the hidden layer, learningrate and activation functions was varied. After evaluation of architectures, it was found that the sigmoidal activation function showed a better performance than the hyperbolic tangent and that thenumber of neurons and learning rate directly influence the error. The neural model with the lowest squared error was the network with the sigmoid function, learning rate 0.1 and 5 neurons in theintermediate layer. This work allowed to determine which variables most influence in the succinic acid production and in the construction of the neural model for this process.
O ácido succínico, metabólitocomum de microrganismos, utilizado no mercado alimentício é produzido exclusivamente por via fermentativa e grande atenção tem sido dada para o uso de matérias-primas renováveis para este fim. Estetrabalho teve como objetivo determinar as variáveis que influenciam na produção de ácido succínico por via fermentativa utilizando a cepa Actinobacillus succinogenes (CIP 106512) através de umplanejamento fatorial fracionário e testar diferentes arquiteturas de redes neurais artificiais para modelar este processo. As redes neurais artificiais (RNAs) utilizadas possuem três camadas e foram do tipo...
Ler documento completo
Por favor, assinar para o acesso.