Predicting the material behavior of recycled composites: Experimental analysis and deep learning hybrid approach

Abstract

The end-of-life issues associated with composite materials have inspired extensive investigations into recycling methods. However, recycled composites cannot be widely used owing to their poor reliability, which is caused by the random and significant variations in their mechanical properties. This accordingly study proposed and demonstrated a method for predicting the mechanical behaviors and fracture mechanisms of recycled composites. First, recycled carbon-fiber-reinforced polymers were manufactured using mechanical recycling and compression molding. Surface images of the resulting specimens were captured and tension tests subsequently conducted to obtain their mechanical properties. The images and test results were used to train convolutional neural networks to predict three mechanical properties and investigate the resulting stress-strain curves. Furthermore, the specimen fracture mechanisms were investigated using the Gradient-weighted Class Activation Mapping technique. The results indicate that the proposed approach can be effectively applied to analyze the mechanical behaviors of recycled composites and provide insights into their fracture mechanisms under specific stress conditions. These capabilities are expected to increase the reliability and utility of recycled composite materials.