An Artificial Intelligence-Based Melt Flow Rate Prediction Method for Analyzing Polymer Properties

Area of Science:

• Polymer Science and Engineering
• Materials Science
• Artificial Intelligence in Manufacturing

Background:

• Polymers are increasingly replacing traditional materials due to superior properties.
• Melt Flow Rate (MFR) is a critical indicator of polymer quality and processability.
• Current MFR measurement methods are time-consuming and not suitable for real-time industrial quality control.

Purpose of the Study:

• To develop an accurate and deployable artificial intelligence model for real-time prediction of polymer melt flow rate (MFR).
• To address the limitations of conventional offline MFR measurement techniques in industrial settings.
• To enhance polymer quality monitoring and processability analysis.

Main Methods:

• A dataset of 1044 polymer samples was utilized with six input features (reactor temperature, pressure, hydrogen-to-propylene ratio, catalyst feed rate) and MFR as the target variable.
• Min-max scaling was applied for input feature normalization.
• Two ensemble models, Kernel Extreme Learning Machine (KELM) and Random Vector Functional Link (RVFL), were developed and optimized using the Pelican Optimization Algorithm (POA).

Main Results:

• The proposed LAIML-MFRPPPA model achieved high predictive accuracy, with R² of 0.965, MAE of 0.09, RMSE of 0.12, and MAPE of 3.4%.
• The model demonstrated superior performance compared to traditional and deep learning models.
• SHAP-based sensitivity analysis identified melt temperature and molecular weight as dominant input features influencing MFR.

Conclusions:

• The LAIML-MFRPPPA model provides a robust and accurate solution for real-time polymer quality monitoring.
• This AI-driven approach facilitates efficient processability analysis and quality control in polymer manufacturing.
• The model's ability to predict MFR in real-time offers significant advantages for industrial applications.