Parametrically-upscaled continuum damage mechanics (PUCDM) model for plain weave woven composites: Part I model development

• Published in: Composite structures 2022-09, Vol.296, p.115825, Article 115825
• Main Authors: Xiao, Yanrong, Zhang, Xiaofan, Ghosh, Somnath
• Format: Article
• Language: English
• Subjects: Damage evolution; Machine learning; Parametrically-upscaled continuum damage mechanics (PUCDM) model; Representative aggregated meso-structural parameters (RAMPs); Woven composites
• ISSN: 0263-8223; 1879-1085
• DOI: 10.1016/j.compstruct.2022.115825

This is the first of a two-part paper that develops a validated 2-level parametrically-upscaled continuum damage mechanics (PUCDM) model for multiscale modeling of damage evolution in plain weave woven composites, characterized by multiple levels of hierarchy. The two-level PUCDM models, viz. PUCDM-1 and PUCDM-2, bridge three levels of composite hierarchy. The PUCDM-1 model corresponds to the constitutive model for yarns in the mesoscopic woven RVE with representation of the morphology and mechanisms of underlying unidirectional fibers in level-1 of the hierarchy. The PUCDM-2 model is used for modeling the structural scale composite behavior with morphological and mechanisms representation from the underlying levels 2 and 1. Functional forms of constitutive coefficients in the PUCDM-1 and PUCDM-2 models, in terms of representative aggregated micro- and meso-structural parameters (RAMPs), material properties, and damage states, are generated by machine learning (ML) methods. The ML algorithms operate on databases of volume-averaged response variables, obtained from the simulation of lower-level RVEs. The validation tests, comparing PUCDM model results with homogenized variables from lower scale analysis demonstrate very good accuracy. The two-level PUCDM models are capable of predicting the stress–strain and damage evolution with high computational efficiency and accuracy. •Two-level Parametrically-upscaled continuum damage mechanics (PUCDM) model.•Multiscale modeling of deformation and damage in plain-weave woven composites.•PUCDM-1 model for mesoscale fiber yarn modeling incorporates microscale descriptors.•PUCDM-2 model for macroscale modeling bridges mesoscale descriptors and properties.•High accuracy and computational efficiency in multiscale analysis of woven composites.