Accurately predicting the performance degradation of composite laminates remains a considerable challenge, primarily due to the multi-physics coupled phenomena, intricate damage evolution, and the inherent limitations of conventional mathematical formulations. Deep learning, with its outstanding capacity for pattern recognition and complex mapping, offers a robust solution to circumvent the limitation of physics priors. In this work, a novel Transformer-LSTM network is presented specifically for the prediction of the performance degradation of composite laminates. During this process, the Transformer-LSTM network is systematically constructed to optimize its performance for the unique features of composite performance degradation. A series of experiments are implemented to verify the effectiveness of the model on several high-dimensional and nonlinear composite degradation dataset. This work emphasizes the potential of advanced adhibition of deep learning to predict the performance degradation of composite laminates with self-data, which provides a novel insight for the accurate prediction of nonlinear and high-dimensional degradation data in actual applications.

Composite, performance degradation prediction, data-driven, Transformer-LSTM network