Application of machine learning to reveal relationship between processing-structure-property for polypropylene injection molding

Revealing relationship between processing–structure-property using machine learning (ML) is beneficial to shorten the development cycle of polymer materials, which could be a promising solution to break through the traditional mode of trial-and-error. In this work, the quantitative data of processing field, polymorphic structure and mechanical properties of samples are successfully obtained based on microfocus wide-angle X-ray diffraction and mechanical property measurements, which provides the dataset for the application of machine learning. An initial step toward predicting the contents of polymorphic structure (α and β) and mechanical properties of isotactic polypropylene injection molding is taken by applying four algorithms of machine learning to establish regression models. The importance of processing descriptors is further analyzed. The prediction models for the contents of polymorphic structure and mechanical properties demonstrate satisfactory performance (R2max = 0.95). Compared with the prediction for the content of α-crystals, the results of feature importance analysis indicates that the injection pressure and shear rate of filling end in the processing descriptors charge higher contribution to the prediction for the β-crystals. And mold temperature plays an important role in predicting mechanical properties, which is consistent with the results of the signal-to-noise (S/N) ratio analysis. [Display omitted] •Machine learning is facile tool to investigate the relationship of processing-structure-property for injection molding.•The contents of polymorphic structure and mechanical properties are predicted successfully.•The analysis of feature importance shows the contribution of processing descriptors to the prediction results.