ABSTRACT The plain woven composites have complex mesoscale structure characteristics, such as variation in the yarn width, thickness and gap values, random layer shift and nesting patterns. To reveal the individual effects of mesoscale structural characteristics on moisture diffusion behavior in plain woven composites, this study compares different representative volume element modeling methodologies based on CT images. To reveal the individual effects of layer shift and nesting on moisture distribution behavior, a modeling method based on feedforward neural networks (FFNN) and CT images is developed. It is found that the model considering all factors, variation in the yarn dimensions, random layer shift and nesting patterns, shows the closest alignment with experimental results. Larger yarn gaps lead to higher early‐stage mass gain values. The layer shift results in variation in early‐stage mass gain values due to the randomness in the moisture transfer path. Nesting contributes to a more complex mass transfer path and a larger fiber volume fraction, resulting in greater variation in early‐stage values and lower saturation values. Moisture absorption primarily causes thickness‐direction expansion, with maximum tensile stress concentrating at yarn boundaries, potentially leading to interface debonding.
Wang et al. (2026) studied this question.