A joint stiffness identification method based on finite element modeling and frequency response functions

Accurate finite element (FE) modeling of mechanical structures is extremely difficult with unknown joints or boundary conditions. An alternative joint stiffness identification method that involves a hybrid of FE model and frequency response functions (FRFs) is presented. Firstly, the joint stiffness is assumed by experience and the mechanical structure is modeled with the FE method. Secondly, the FRFs at the concerned nodes of the structure are simulated and measured, respectively. Then the norm of residual FRFs between the simulations and measurements is calculated. Finally, a sensitivity-based iterative algorithm is derived for minimizing the norm of residual FRFs and the least square method is used to solve over-determined iterative equation. The joints stiffness parameters are identified through the iteration process, while the FE model is updated simultaneously. The proposed joint stiffness identification method is applied on a clamped beam assembly. The first three natural frequencies calculated by the FE model are compared with the measured values. The largest relative error of the simulation deceases from 16.7 % to 2.5 % after the joint stiffness parameters are identified, which demonstrates the effectiveness of the presented method.