Kinetic study by thermal decomposition of antiretroviral drugs, Efavirenz (EFV) and Lamivudine (3TC), usually present in the HIV cocktail, can be done by individual adjustment of the solid decomposition models. However, in some cases unacceptable errors are found using this methodology. To circumvent this problem, here is proposed to use a multilayer perceptron neural network (MLP), with an appropriate algorithm, which constitutes a linearization of the network by setting weights between the input layer and the intermediate one and the use of Kinetic models as activation functions of neurons in the hidden layer. The interconnection weights between that intermediate layer and output layer determines the contribution of each model in the overall fit of the experimental data. Thus, the decomposition is assumed to be a phenomenon that can occur following different kinetic processes. In the investigated data, the kinetic thermal decomposition process was best described by R1 and D4 model for all temperatures to EFV and 3TC, respectively. The residual error adjustment over the network is on average 10 3 times lower for EFV and 10 2 times lower for 3TC compared to the best individual kinetic model that describes the process. These improvements in physical adjustment allow detailed study of the process and therefore a more accurate calculation of the kinetic parameters such as the activation energy and frequency factor. It was found E a = 75.230 kJ / mol and s -1 for EFV and E a = 103.25 kJ / mol and s -1 for 3TC.

publicado na web em 16/10/2017 KINETIC AND THERMAL DECOMPOSITION STUDIES OF RIGID POLYURETHANE FOAMS MODELED BY ARTIFICIAL NEURAL NETWORKS. Kinetic models of solid thermal decomposition are traditionally used for individual fit of isothermal experimental data. However, this methodology presents unacceptable errors in some regions of the data. To solve this problem, a neural network was adopted in this work. The implemented algorithm uses the rate constants as predetermined weights between the input and intermediate layer and kinetic models as activation functions of neurons in the hidden layer. The contribution of each model in the overall fit of experimental data is calculated as the weights between the intermediate and output layer. In this way, the phenomenon is better described as a sum of kinetic processes. Two rigid polyurethane foam samples: loaded with Al 2 O 3 and no inorganic filler were used in this work. The R3 and D2 models described the thermal decomposition kinetic process for all temperatures for both foams with smaller residual error. However, the network, combining the kinetic models, presented residual errors on average 10 2 times lower compared to these individual models. The determined activation energy is 12.44 kJ mol -1 higher for the loaded foam. This result corroborates the use of this material as flame retardant, even with the presence of a small amount of charge in its structure.Keywords: thermal decomposition; rigid polyurethane foams; polyurethane; artificial multilayer neural network. INTRODUÇÃOEstudos cinéticos de decomposição térmica no estado sólido são de grande interesse científico e industrial.1 Três metodologias relevantes no tratamento teórico de cinética de decomposição em sólidos são: (i) descrição do processo por modelos cinéticos, 2 (ii) associação e correção dos modelos cinéticos por rede neural artificial 3,4 e (iii) modelo isoconversional. 5A metodologia de ajuste por modelos cinéticos considera que as reações de decomposição térmica de sólidos ocorrem na interface do produto-reagente. Este processo cinético, com base na formação e crescimento de núcleos, pode ser estudado por análise de dados de decomposição térmica nos quais a redução da massa é medida num intervalo de tempo à temperatura constante. Os núcleos de reação são preferencialmente formados em imperfeições da estrutura e os modelos cinéticos são usados para explicar as isotermas experimentais, em que a fração de decomposição é medida ao longo do tempo. Um determinado modelo é escolhido devido a sua melhor correlação para ajustar os dados experimentais. No entanto, em muitas situações, o erro residual do modelo não é aceitável na descrição do processo total, embora, para regiões específicas de decomposição, o modelo possa ser apropriado, sendo apenas uma primeira aproximação. Os modelos podem ser representados por uma equação geral do tipo: 2 (1) em que α representa a fração de decomposição, t o tempo, k a constante de velocidade. Os parâmetros de ajuste p e q definem o modelo físico.Em alguns trabalhos já ...

No abstract

Recebido em 11/6/12; aceito em 10/9/12; publicado na web em 1/2/2013 Methane combustion was studied by the Westbrook and Dryer model. This well-established simplified mechanism is very useful in combustion science, for computational effort can be notably reduced. In the inversion procedure to be studied, rate constants are obtained from [CO] concentration data. However, when inherent experimental errors in chemical concentrations are considered, an ill-conditioned inverse problem must be solved for which appropriate mathematical algorithms are needed. A recurrent neural network was chosen due to its numerical stability and robustness. The proposed methodology was compared against Simplex and Levenberg-Marquardt, the most used methods for optimization problems.

No abstract