2026 年 99 巻 6 号 p. 148-154
This study presents a data‑driven approach to optimize PPS/elastomer blend melting processes. Although elastomer dispersion improves impact strength, process design is hindered by thermal degradation and morphology changes. Infrared sensing and spectroscopic measurements were integrated with machine‑learning models to extract key descriptors. Melt‑temperature profiles obtained from a twin‑screw extruder enabled random‑forest regression to identify elastomer content and the temperature at the first kneading zone (T1) as dominant variables. Raman and near‑infrared spectra captured PPS degradation and dispersion states. Incorporating these descriptors into a Bayesian optimization framework efficiently identified high‑performance processing conditions, demonstrating the utility of data‑driven informatics in optimizing PPS/elastomer systems.
