Data-driven load identification method of structures with uncertain parameters

Abstract

Dynamic load identification plays a crucial role in structural design and optimization. The majority of current studies are focused on deterministic structures. However, the structural parameters of actual engineering structures are unknown. It is essential to investigate the issue of dynamic load identification for uncertain structures since the existence of uncertain parameters can lead to errors between load identification results and actual load values. Therefore, in this paper, we propose a data-driven dynamic load identification method for structures containing some uncertain parameters. To start, the uncertain parameters are characterized by a set of closed interval vectors. Then a convolutional neural network (CNN) is introduced for the reconstruction of the interval of unknown load. Combining the interval analysis theory with Taylor expansion, the upper and lower boundaries of the supervised loads are obtained and used as training samples. Finally, the trained CNN model directly identifies the boundaries of the unknown load interval. The simulation results demonstrate that the proposed method has great accuracy in load identification and has good robustness to noise. We construct a simply supported beam structure for experiments to further validate the feasibility of the proposed method in engineering. Additionally, we discuss the effect of measurement point distribution and number of samples on the identification accuracy, which is beneficial for applications in engineering practice.