Prediction of different times of sulfur curing rubber using a recurrentneural network

This study uses artificial neural networks to model times of rubber sulfur curing. Two inputparameters were sulfur concentration and accelerator. Five output parameters representing differentcuring times were considered: ۱۰%, ۲۰%, ۵۰%, ۹۰%, and ۹۵% of rubber curing. A multilayer neuralnetwork was used for modeling, employing the LSTM training algorithm and nonlinear activationfunctions including Sigmoid, Hyperbolic Tangent, Leaky ReLU, ReLU, and Softmax. The optimalarchitecture, determined through iterative guesswork and error minimization, consists of three layerswith ۳۲:۱۶:۸ neurons. The modeling results are compared with experimental data, showing excellentconcurrence with a correlation coefficient (𝑅𝑅۲) of ۰.۹۹۸۴ and low error (MSE = ۰.۰۰۰۵۹). Thesefindings confirm the efficiency of artificial neural networks in accurately predicting different rubbercuring times.

کلیدواژه ها:

Modeling ، Neural Network ، Curing Times ، Rubber