Using an artificial neural network (ANN) for prediction of thermal degradation from kinetics parameters of vegetable fibers

Vegetal fibers are prominent reinforcements for polymer composite materials, considering their properties and application possibilities. In particular, thermal degradation behavior is crucial for determining an application subjected to a temperature range. Methods to predict properties are a trend in materials science and have the main advantage of saving cost and time. For this reason, in the present study, an artificial neural network (ANN) approach was used to predict the thermal degradation curves. The heating rate of 10 °C·min^− 1 was carried out to train the network with 12 hidden layers and optimal training dataset of 60. Other heating rates were simulated and showed an excellent agreement with the experimental data. The coefficient of determination was R² > 0.99 for all sources of biomass, exhibiting appropriate predictive fit with error following the sequence: ramie (1.15 %) < kenaf (1.33 %) < curaua (1.83 %) < jute (1.97 %). In conclusion, ANNs can learn from their data and optimize processing, formulations, predict properties, and other input data combinations. The predictive curves present high reliability with the experimental fit allowing the prediction of the mass loss for different temperatures versus the heating rate set.