Intelligent Modelling of the Real Dynamic Viscosity of Rubber Blends Using Parallel Computing

Area of Science:

• Materials Science
• Polymer Engineering
• Computational Science

Background:

• Predicting rheological behavior of rubber blends is crucial for product manufacturing.
• Complex, nonlinear processes in rubber processing challenge traditional analytical modeling.
• Artificial intelligence offers a promising approach for accurate rheological modeling.

Purpose of the Study:

• To develop a highly efficient artificial neural network (ANN) model for predicting rubber blend rheological test results.
• To optimize the ANN model using a novel training algorithm and fast parallel computing.
• To analyze dynamic viscosity-time curves of styrene-butadiene rubber blends under various conditions.

Main Methods:

• Utilized a Generalised Regression Neural Network (GRNN) to analyze 120 dynamic viscosity-time curves.
• Optimized the GRNN model by constraining the training dataset fitting error to below 1%.
• Employed parallel computing with multiple cores to accelerate repetitive calculations.

Main Results:

• Achieved excellent agreement between predicted and measured generalization data, with an error below 4.7%.
• Demonstrated the high generalization performance of the developed ANN model.
• Significantly reduced total computation time through parallel computing.

Conclusions:

• The developed ANN model accurately predicts the rheological behavior of rubber blends.
• The novel training algorithm and parallel computing enhance model efficiency and speed.
• This AI-driven approach provides a reliable tool for optimizing rubber processing and production.