Polymer property prediction and optimization using neural networks.

Area of Science:

• Materials Science
• Computational Chemistry
• Polymer Science

Background:

• Predicting polymer properties is a complex, nonlinear challenge, particularly for high molecular weight engineering plastics.
• Experimental investigation of monomer modifications is often impractical due to the vast number of possibilities.
• This limitation hinders the design of novel polymers with specific end-use properties.

Purpose of the Study:

• To demonstrate the prediction of modified monomer properties using neural networks.
• To leverage existing polymer databases for accurate property prediction and optimization.
• To overcome experimental limitations in polymer design.

Main Methods:

• Utilized a comprehensive database of polymer properties.
• Employed various neural network architectures, including backpropagation networks and self-associating maps.
• Selected and classified networks based on prediction accuracy, training speed, and generalization capabilities.

Main Results:

• Neural networks accurately predict properties of modified monomers.
• Specific networks were identified for precise prediction of distinct polymer characteristics.
• Networks were categorized by training efficiency and generalization performance.

Conclusions:

• Neural networks offer a viable computational approach for predicting polymer properties.
• The developed method facilitates the accurate prediction and optimization of monomer modifications.
• This computational strategy accelerates the design of new polymers with tailored properties.