Deep learning based load and position identification of complex structure

Load identification is crucial for structural health monitoring. However, the traditional load identification methods based on the system response function. These methods calculate load using the inverse solution of structure dynamic response and system characteristics. Therefore, the traditional methods are only applicable to linear structures, and have some shortcomings such as ill-posedness and huge computational cost. In this paper, a deep learning based identification method is proposed to identify the static load amplitude and position of bulkhead plate rapidly. Firstly, the loading experiment is carried out. The raw signal is preprocessed by normalization and temporal segment. Secondly, we design the Bi-directional Long Short-Term Memory and Convolutional neural network (BiLSTM+CNN) to realize static load identification. Finally, the model is applied to the identification of dynamic loads to prove its generalization capability. It is demonstrated that the proposed deep learning model reveals a fast convergence and a high degree of accuracy in the identification of the load and its position. Moreover, the model can be applied to dynamic load identification.