Quantitative relations between curing processes and local properties within thick composites based on simulation and machine learning

Abstract

Overheating is almost inevitable during the curing of thick polymer matrix composite parts, which always induces degradation of the mechanical properties. To explore the relationship between the local process variables and the property distribution of interlaminar shear strengths and compression strengths inside thick composites, experiments and relative simulations were conducted herein. Based on machine learning techniques, a convolutional autoencoder (CAE) was used to evaluate the spatial distributions of temperature, cure degree, and stress during autoclave curing process of thick composites. The results demonstrate a strong linear relationship between the spatial distribution of stress with the property values of interlaminar shear strengths and compressive strengths. This indicates that the stress distribution history strongly impacts the mechanical properties of thick laminates, which is usually neglected in previous studies that only concerns the stress magnitude.

Keywords

• Machine learning
• Autoclave process simulation
• Thick composites
• Mechanical properties
• Stress distribution