TransRheo: A Physics-Informed Multi-Task Numeric Transformer for Polymer Nanocomposite Rheology
A new natural language embedding neural network method EINN is proposed. By combining the Transformers architecture with the scibert encoder, it achieves accurate viscosity prediction under extreme conditions, extrapolation scenarios, and sparse data situations.