A constitutive model of composite laminates with matrix cracking is essential for the design of liner-less composite vessels. This paper proposes a data-driven approach to predicting stiffness degradation in cross-ply laminates with matrix cracking. A pixelated expression of the stacking sequence was adopted to deal with the difference in the number of cracked plies, as the features for machine learning (ML) should be uniform. The data set for ML training is established by using finite element analysis (FEA), where the material properties, stacking sequences and crack densities are random. The normalised stiffness of the cracked plies of [0n/90m]s and [0/90m/00.5]s laminates is obtained by using the proposed ML method, and they are then compared with the available experiments, a micro-mechanical analytical model and FEA simulations. The ML method successfully predicted the stiffness degradation of the laminates with different numbers of cracked plies and further explored the application of data-driven methods for mechanics of composite materials.
Read full abstract