Study on the Healing Efficiency of Hydrogen‐Bonded Self‐Healing Polyurethane Based on Machine Learning Techniques

To address data scarcity in machine learning‐assisted design of high‐performance polymers, this study develops a novel data‐augmentation strategy. Using hydrogen‐bonded self‐healing polyurethane as a case study, a series of equations describing the self‐healing mechanism were derived based on the self‐healing mechanism of hydrogen‐bonded self‐healing polyurethane. The study generated extensive data by utilizing these equations. The Elastic Net method was employed for feature selection, performing correlation analysis and removing highly correlated features. The selected features were used as inputs for a transformer model to predict the effect of temperature and time on self‐healing efficiency of polyurethanes. The model achieved high accuracy for both linear (R 2 = 0.941, RMSE = 4.972%) and branched (R 2 = 0.948, RMSE = 4.496%) structures. Predictions for individual samples also showed high accuracy, with R 2 = 0.929 and RMSE = 2.746% under isothermal conditions, and R 2 = 0.984 and RMSE = 1.666% under fixed time conditions. These results demonstrate that this study provides a viable solution to the issue of data scarcity in machine learning‐assisted polymer design. The equations and features analyses reveal that chain diffusion and the rate of hydrogen bond formation significantly influence the self‐healing process of hydrogen‐bonded polyurethanes.