Abstract
The multiscale model based on micro-mechanics failure theory is modified to consider complex internal structures, including a fiber random arrangement pattern and interface with the clustering method. Then, a feed-forward-neural-network (FFNN)-based damage evolution method is developed to evaluate the macroscale property degradation. The progressive damage analysis of open-hole laminates under compression is conducted to validate the modified multiscale method. The predicted results reveal that the interface results in the premature initiation of damage, and the fiber random arrangement pattern contributes to the decrease in the predicted compression responses. The developed FFNN-based method aimed at degradation results in an increase in the predicted compression strength. For the fiber random distribution pattern, the increase in percentage of predicted compressive strength is 6.0%, which is much larger than the value for the fiber diamond distribution pattern.