Problems in using experimental data for dynamic substructuring of a lumped parameter system

In experimental dynamic substructuring two main problems are defined: addition of substructures (coupling) and subtraction of substructures (decoupling). Decoupling can be important in built-up structures where some components (critical subsystems or joints) cannot be removed or accessed easily. Whilst addition of substructures often leads to satisfactory results even in relatively complex cases, subtraction of substructures is a source of problems even in apparently trivial applications. Some critical issues of decoupling (such as ill-conditioning around a discrete number of frequencies) have been highlighted and verified by using simulated data corrupted by random noise. In this paper, experimental data acquired on a lumped parameter benchmark system are used to check previously highlighted problems both in coupling and decoupling, and to look for additional issues (systematic errors, inconsistencies, etc.) that can not be observed from simulated data.