Dynamic analysis of coupled structures using experimental data

The work in this thesis is concerned with substructure coupling techniques which incorporate data readily available from modal tests. Two coupling techniques are investigated in detail, namely the Impedance and Modal Coupling techniques. In the former, different procedures for reducing models are described and the effects of the corresponding incompleteness at this stage mainly relating to the coordinates are investigated in order to detect and understand the main sources of errors in the predicted dynamic behaviour of different case studies. Additionally, alternative coupling algorithms are proposed to overcome numerical errors arising due to redundancy in the set of connection coordinates. In the second technique Modal coupling the investigation is concentrated on the effects of truncating the number of modes. A refined approach is presented for including residual flexibility equivalent to the omitted modes by incorporating a dummy flexible system between two components. A common problem in all the investigated coupling techniques is that their validity may be reduced by using experimental data which can be measured rather than the data that ought to be measured. One of the most critical areas here is the formulation of meaningful constraint equations to express the actual physical connections between components. Sometimes, the number of measured junction coordinates can be excessive, thus provoking numerical difficulties during the coupling process. On other occasions, there are extreme situations where a lack of information causes a meaningless representation of the actual connection properties. Both situations are dealt with in the present work. The former is investigated by making use of a well established mathematical technique the Singular Value Decomposition which permits a confident inversion of ill-conditioned matrices and, additionally, detects the redundant coordinates responsible for the coupling numerical failures, when combined with a QR factorization. The latter aspect is related to the possibility of accurately measuring rotational coordinate