A critical evaluation of SEMM-based joint identification procedure to reduce the error propagation effects

Mechanical joints play a critical role in many engineering structures, but identifying their characteristics can be challenging, particularly when the joint interface (coupling DoFs) cannot be directly measured. In these cases, an existing iterative procedure based on the use of the System Equivalent Model Mixing (SEMM) expansion technique and substructure decoupling can be used to identify the joint properties by taking measurement on other accessible DoFs (internal DoFs). Despite the potential of this procedure, it is prone to large error propagation. In addition, in the SEMM expansion a weighted pseudo-inverse operation is needed to ensure the convergence of the iterative procedure when both coupling and internal DoFs (extended interface) are involved in the decoupling. This paper focuses on the detection of the error sources in the process and on the definition of some strategies to limit error propagation. The use of internal DoFs only (pseudo-interface) in the decoupling is proposed. This avoids the use of coupling DoFs affected by the expansion error. Furthermore, two strategies are proposed to improve the conditioning of the procedure when using the extended interface. Both strategies are based on the Truncated Singular Values Decomposition (TSVD). It is shown that the weights clearly indicate the number of singular values to be retained in the matrices to be inverted. The proposed improvements are validated on a laboratory benchmark. Measurements on the benchmark are performed to validate the strategies with experimental data. In addition, the Monte Carlo method is applied using noise-polluted numerical data to evaluate the potential of the proposed strategies to mitigate the error propagation in the SEMM-based joint identification procedure.