Study of dynamic response of crane system via surrogates based on Karhunen–Loève expansion and neural networks

Overhead cranes are widely used in construction sites and manufacturing/assembly lines for various tasks (loading, unloading, lifting, transporting,. . . ). This paper investigates the dynamic response of the main beam (i.e., girder) of an overhead crane by a surrogate technique based on truncated Karhunen–Loève (KL) expansion and neural networks. First, the physical modeling and the motion equations of the crane system are derived using the Lagrange equation. Then, the dynamic responses of the overhead crane system with a number of the input parameters (i.e., configurations) are estimated by the numerical Newmark-beta integral method. Finally, the surrogates based on the truncated KL expansion and neural networks are constructed for studying the dynamic responses of the girder within a limited reference dataset. It can be stated that with a convergence property, the reconstructed surrogate performs good predictability for characterizing the dynamic responses of the structure. The presented method allows us to study the dynamic analysis and optimization of the crane according to the design conditions in the actual applications.