Accelerated Development of Novel Biomass-Based Polyurethane Adhesives via Machine Learning

Area of Science:

• Polymer Chemistry
• Materials Science
• Biotechnology

Background:

• 2-Pyrone-4,6-dicarboxylic acid (PDC) derived from lignin offers a sustainable route to biomass-based polymers.
• PDC-based polyurethanes (PUs) exhibit promising adhesive properties, but require optimization for enhanced performance.

Purpose of the Study:

• To improve the adhesive strength of biomass-based polyurethanes (PUs) using 2-pyrone-4,6-dicarboxylic acid (PDC).
• To combine experimental approaches with machine learning (ML) for efficient optimization of adhesive properties.

Main Methods:

• A Taguchi L25 orthogonal design was used to synthesize 25 adhesive samples with varying polyols, isocyanates, and ratios.
• Adhesive strengths were measured after hot-pressing under diverse temperature and time conditions.
• Gaussian process-based Bayesian optimization (BO) and Random Forest regression were employed for data analysis and optimization.

Main Results:

• Bayesian optimization identified an optimal PDC-based PU adhesive formulation.
• The optimized adhesive achieved a significantly improved strength of 10.04 ± 1.26 MPa within five iterations.
• Random Forest regression validated the findings from Bayesian optimization.

Conclusions:

• Bayesian optimization effectively guides experimental parameters for enhanced adhesive performance.
• This ML-driven approach accelerates the development and optimization of novel biomass-based adhesive materials.
• PDC-based PUs show potential as high-performance, sustainable adhesives.