Structural Dynamic Load Prediction Method Based on Long Short-term Memory Network

Objectives To address the limitations of traditional surrogate models in handling time-dependent dynamic processes and heterogeneous data, this paper proposes a dynamic load surrogate model method based on long short-term memory (LSTM) networks. Methods The surrogate model comprises two modules: load feature encoder and load response decoder. Firstly, the LSTM in the load feature encoder performs feature extraction on the time series of dynamic external loads. Then, the extracted load features are combined with the structural parameter features. The LSTM in the load decoder conducts further feature extraction and finally generate output, thus considering the heterogeneous data input of dynamic external load time series and one-dimensional structural parameter features comprehensively to predict the time history of internal force responses. The model’s accuracy is evaluated using a finite element simulation dataset and compared with other surrogate model methods. The model is also employed for the correlation analysis between structural parameters and internal force responses, serving as an example for the application of surrogate models in rapid calculation of a large number of cases. Results The results show that the average accuracy of the dynamic load surrogate model can reach 98%, which is higher than other methods, and the calculation speed is faster than that of the finite element method. Conclusions The proposed method addresses the issue of heterogeneous data involving both time-series and non-time-series features, offering advantages of high accuracy and efficiency, therefore effective for fast iterative computation tasks.