Damage assessment of composite material based on variational mode decomposition and BiLSTM

In industry, the failure of composite material may have negative consequences on the productivity, security, and the environment. To avoid such situations, the structural health monitoring (SHM) of the physical system used for tracking damage growth in these structures with minimal human intervention, predict future structural performance schedule maintenance. To overcome these problems, a new approach has been proposed based on bidirectional long short-term memory (BiLSTM) for the damage severity estimation of composite material. The proposed method focused on detecting and quantifying damage evolution in composite materials. Lamb wave (LW) is frequently used to evaluate the damage and its geometrical shape, and the test results need to be interpreted by trained experts. This approach is demonstrated on data collected from a run-to-failure tension-tension fatigue experiment measuring the damage progression in carbon fiber reinforced polymer (CFRP). The time frequency analysis used in this study with three kinds of neural networks RNN, LSTM, and BiLSTM to analyze the attributes with different severity of damage. Then, trained models are applied to identify the depth information of impact damage. VMD is a signal processing technique used to decompose signals into different band-limited intrinsic mode functions (IMFs), then the IMFs are used as input to feed the development model based on BiLSTM. Among all models, the proposed method based on VMD-BiLSTM has a good adaptive performance of modeling with respect to the efficiency and effectiveness by comparing with traditional learning techniques. Experimental results show that the proposed method based on VMD-BiLSTM can reflect effectively the damage assessment of CFRP and leads to a significant improvement of the predictive accuracy.