Natural Rubber Blend Optimization via Data-Driven Modeling: The Implementation for Reverse Engineering.

Area of Science:

• Materials Science
• Polymer Chemistry
• Computational Modeling

Background:

• Traditional natural rubber formulation relies heavily on formulator experience, involving extensive trial-and-error.
• Optimizing rubber blends for specific applications requires balancing numerous components and processing parameters.

Purpose of the Study:

• To implement and evaluate predictive models for natural rubber blend formulation.
• To reduce experimental iterations and financial resources in developing rubber compounds.
• To reverse-engineer a footwear material based on target viscoelastic properties.

Main Methods:

• Incorporation of predictive models: linear regression, response surface method (RSM), artificial neural networks (ANNs), and Gaussian process regression (GPR).
• Training models with a small dataset to minimize experimental effort.
• Reverse engineering a footwear material using target properties: relaxation behavior, tanδ, and hardness.

Main Results:

• Response surface method (RSM), artificial neural networks (ANNs), and Gaussian process regression (GPR) achieved prediction accuracies of 90%, 97%, and 100%, respectively.
• Prediction accuracy generally increased with model complexity.
• Successful reverse engineering of a lightweight, high-performance footwear material.

Conclusions:

• Predictive modeling offers a powerful alternative to traditional experience-based rubber formulation.
• Gaussian process regression demonstrates exceptional accuracy for predicting compound properties.
• These methodologies provide versatile tools for formulating rubber compounds based on desired performance characteristics.