Multidimensional dominant mapping eigenvalues fusion-driven method for structural damage identification and crack tip energy characterization

Structural damage identification serves as a critical safeguard, enhancing the safety, longevity, and reliability of engineering structures. This article proposed a structural damage identification approach based on the multidimensional mapping eigenvalue analysis (MMEA) and multiple eigenvalues fusion of CNN-LSTM (MEFCNN-LSTM) model. With the combination of the dominant mapping eigenvalues of various damage indices, the proposed approach forms the Multiple Eigenvalues Fusion Damage Index (MEFDI). This is a new approach to evaluate the impact of eigenvalues on indices, which is a combination of analysis of sensitivity, standard regression coefficient and variance. The MMEA approach was utilized to compute six key mapping eigenvalues, serving as input for the MEFCNN-LSTM model. This model extracts features from the input through the CNN layers, and then analyzes the eigenvalues using the LSTM layers. Following data integration with the MEFCNN-LSTM model, the MEFDI was developed. To verify the effectiveness of the structural damage identification and crack tip energy characterization, the main beam of crane structure was selected as a case study. The model identified the damage location and severity of the crane main beam. The average damage identification accuracy was 98.36 %, with a range of [95.62 %, 99.80 %]. Moreover, the model was also capable of characterizing the crack tip energy. The average accuracy of characterization was 92.93 % at single location and 90.61 % at multiple locations, compared with numerical analysis. The proposed method resolves traditional limitations like incomplete feature correlation and inefficient data integration, providing a robust solution for precise structural damage assessment and crack tip energy characterization.