Predictive Modelling and Optimisation of Rubber Blend Mixing Using a General Regression Neural Network.

Area of Science:

• Materials Science
• Chemical Engineering
• Artificial Intelligence

Background:

• Rubber blend mixing is critical for product quality.
• Real-time process control is essential for efficiency.
• Predictive modeling can optimize complex industrial processes.

Purpose of the Study:

• To develop an intelligent predictive system for real-time control of rubber blend mixing.
• To accurately predict key process parameters like viscosity, temperature, and energy consumption.
• To enable early detection of mixing progress for process optimization.

Main Methods:

• Implementation of a General Regression Neural Network (GRNN) model.
• Utilizing experimental data from a Brabender Plastograph EC Plus.
• Optimizing GRNN kernel width (σ) via 10-fold cross-validation.

Main Results:

• High predictive accuracy for viscosity, temperature, and energy consumption.
• Accurate evaluation of mixing progress from initial 10% of data.
• Achieved R² values close to 1 and low RMSE, confirming model reliability.

Conclusions:

• The GRNN-based system provides robust and scalable intelligent control for rubber mixing.
• The system enhances productivity and quality assurance in industrial applications.
• The predictive approach is applicable beyond rubber blending processes.