Infrared-Driven Polymer Intelligence

Infrared (IR) spectroscopy offers direct insight into the vibrational modes of polymers, yet its full potential for predicting critical thermal properties remains underexplored. Herein, IR spectra is integrated with machine learning (ML) models, including support vector regression (SVR) and convolutional neural networks, to predict glass transition temperatures ((Formula presented.)) with high accuracy across diverse polymer formulations. Compared to fingerprint-based descriptors, IR-driven approaches capture subtle molecular variations such as impurities, additives, and colorants, that significantly influence the properties of polymers. By leveraging the rich spectral data, the models achieve improved predictive performance, with the SVR model achieving the highest prediction accuracy, evidenced by (Formula presented.) value of 0.94 and tighter residual distributions. The real-time process control is further demonstrated by applying these models to 3D printing, dynamically adjusting bed and nozzle temperatures to accommodate feedstock variability. This end-to-end system highlights the promise of IR-enabled ML in advancing polymer characterization, offering a robust path toward more intelligent, more adaptive, and more efficient manufacturing workflows.