TransRheo: A Physics-Informed Multi-Task Numeric Transformer for Polymer Nanocomposite Rheology

Overview

TransRheo studies polymer nanocomposite rheology prediction under sparse data and extreme conditions. The framework combines natural-language knowledge encoding, numeric Transformer modeling, empirical knowledge, and physical constraints.

Method

Monotonicity constraints aim to improve prediction stability at low shear rates and low temperatures, including out-of-distribution conditions.

Status

The manuscript is under first-round review at Interdisciplinary Materials.