|کد مقاله||کد نشریه||سال انتشار||مقاله انگلیسی||ترجمه فارسی||نسخه تمام متن|
|672707||1459449||2016||5 صفحه PDF||سفارش دهید||دانلود رایگان|
• Novel technique of kinetic parameters determination with neural network models is proposed.
• Optimized artificial neural network successfully predicts the kinetic parameters even outside the training range.
• Constant-rate thermal analysis has high resolution ability between different reaction types.
Inspired by the recent victory of the computer program Alpha Go  over the world strongest human Go player and by the ability of this program based on artificial neural networks (ANN) for self-learning, we present the attempt to perform the thermokinetic analysis using ANN models. Various virtual combinations of kinetic parameters and reaction model types were used to generate thermoanalytical signals similar to experimental ones for neural network training. Results prove that trained artificial neural network models can be successfully used for the kinetic analysis. For methodological purposes, various types of input parameters have been considered, showing the weakness of analysis based on the single heating runs along with the high resolution ability of constant rate thermal analysis. The principal capability of neural networks to perform kinetic analysis within wide ranges of activation energy and pre-exponential factor for a number of reaction models has been demonstrated.
Journal: Thermochimica Acta - Volume 637, 10 August 2016, Pages 69–73