Improving Hybrid Model Accuracy for Structural Analysis: The Effects of Parameterized Joint Stiffness on System Equivalent Model Mixing

Abstract

System equivalent model mixing (SEMM) is a hybrid model that integrates experimental data and numerical analysis to obtain an unmeasured degree-of-freedom (DoF). SEMM extends the frequency response functions obtained from experimental data to a finite element method (FEM) model, resulting in a highly accurate and precise hybrid model that enables the measurement of joint behavior and characteristics. By leveraging the precision of experimental data and the expansion of numerical models, SEMM can acquire frequency response functions from inaccessible joints. This approach provides researchers with more complete information on joint behavior and characteristics than traditional modeling approaches that rely solely on either experimental or numerical data, allowing for more accurate and comprehensive analysis. A common method for implementing SEMM involves using a virtual point transformation (VPT) that accounts for the rotational degrees of freedom of the joint by employing three tri-axial sensors near the joint, or by attaching a transmission simulator (TS) to the joint to mitigate the influence of rotational degrees of freedom. However, the accuracy of the expansion DoF using SEMM is determined not only by the initial boundary condition of the analytical model but also by the boundary condition error between the experiment and the analytical model. Therefore, the accurate establishment of system boundary conditions in the numerical model is crucial for SEMM modeling to improve the reliability and correctness of the resulting hybrid model. This paper describes the development of hybrid models for structural analysis by incorporating parameterized joint stiffness to optimize boundary conditions in numerical models and enhance the accuracy of the resulting hybrid models. The accuracy of the proposed approach based on SEMM is evaluated using the frequency response assurance criterion (FRAC). The results show that the FRAC is sensitive to changes in boundary stiffness and that the accuracy of FRFs is directly proportional to the similarity between the numerical and experimental boundaries. To further investigate the impact of the measured DoF, we applied classical sequential sensor placement optimization techniques to select the optimal DoF in the experiment and evaluated their performance accordingly.