Seismic intelligent retrofitting of aging steel structure using semi-active TMD with LSTM prediction and wavelet transform combined algorithm

Seismic structural control is critical for mitigating seismic damage to steel structures, particularly in the seismic-retrofitting and strengthening of damaged or aging structures. Conventional passive tuned mass dampers (PTMDs) are limited by their dependence on accurate structural information and specific external excitations to achieve optimal performance. When structural stiffness degrades owing to aging-related deterioration, PTMDs typically fail to maintain an effective vibration mitigation performance. Hence, to improve retrofitting performance and enhance applicability to stiffness-degraded structures, this study proposes a semi-active TMD (STMD) control algorithm that can dynamically adjust stiffness and frequency in response to structural movements. The proposed intelligent control algorithm uses a long short-term memory (LSTM) neural network to predict structural responses combined with instantaneous frequency identification via wavelet transform (WT). Two cases, i.e., a single-degree-of-freedom (SDOF) structure and a multi-DOF structure, are considered to evaluate the effectiveness of the combined LSTM-WT algorithm. First, the frequency-response curve of the SDOF structure is calculated. Subsequently, the response of a 10-story shear-type structure is analyzed under the excitation of 10 different earthquake waves. The results are compared with those obtained using an optimized passive TMD and an STMD with the WT algorithm. Furthermore, to demonstrate the control robustness and seismic-retrofitting capability of the proposed combined algorithm, the control effects of the three controllers are compared while considering a 20 % reduction in structural stiffness. The numerical results highlight the effectiveness and robustness of the combined LSTM-WT algorithm across various seismic excitation frequency ranges, thus confirming its potential for practical applications in the intelligent seismic-retrofitting of damaged and aging steel structures.