Multi-objective optimization of mechanical properties in PLA/SCG/silane composites using synthetic data and XGBoost

Area of Science:

• Materials Science
• Polymer Science
• Sustainable Materials

Background:

• Polylactic acid (PLA) composites offer biodegradable alternatives.
• Spent coffee grounds (SCG) present a sustainable reinforcement filler.
• Silane coupling agents (VTMS) can enhance filler-matrix compatibility.

Purpose of the Study:

• To optimize the composition of PLA/SCG composites using a data-driven approach.
• To enhance mechanical properties, specifically tensile strength and Shore D hardness.
• To develop a sustainable bio-composite with improved performance.

Main Methods:

• Fabrication and testing of 75 PLA/SCG/silane composite samples.
• Development of a multi-output XGBoost regression model for property prediction.
• Augmentation of training data using jittering, Gaussian noise, and KDE.
• Application of NSGA-II for multi-objective optimization of tensile strength and hardness.

Main Results:

• Achieved tensile strengths ranging from 26.5-57.9 MPa and Shore D hardness of 77.5-80.8.
• XGBoost model demonstrated high predictive accuracy (R^2 = 0.884 for tensile strength, R^2 = 0.908 for hardness).
• Identified Pareto-optimal compositions favoring higher PLA content and moderate SCG and silane.
• Optimal formulation yielded 53.33 MPa tensile strength and 80.06 Shore D hardness.

Conclusions:

• The combined XGBoost-NSGA-II framework provides an efficient strategy for optimizing bio-composite performance.
• This data-driven approach minimizes experimental effort while maximizing desired properties.
• Optimized PLA/SCG/silane composites represent a promising sustainable material solution.