Structural dynamic modification (SDM) problems involve local modifications with lumped masses, stiffeners, dampers and beam elements, etc., to improve the dynamic behaviour of existing structures. Many proposals have been presented earlier [1-3] which identify the optimal parameter of the system. The need for quantification of identified optimal parameters for a desired natural frequency has been stressed by Snoeys et al. [4]. Reanalysis techniques for finding the geometrical dimensions of structural members have been investigated by Wang [5]. The modification techniques reviewed above offer no guarantee that the chosen modification is the most appropriate. Depending upon the dynamicist's intuition and other factors, the modification may be acceptable, but not necessarily optimal. This is the reason why sensitivity analysis together with modification techniques form useful tools for the optimum modification of the dynamic behaviour of mechanical structures. These techniques, based on the calculation of derivatives of the modal parameters, provide the most effective parameter changes such that the desired dynamic behaviour can be obtained in one modification cycle. The calculation of derivatives of modal parameters in turn requires the sensitivity derivatives of the spatial parameters, namely mass and stiffness matrices of the structure [6, 7]. Usually, these matrices are dynamically condensed corresponding to the co-ordinates of interest [8]. The finite difference calculation of sensitivity derivatives is easy to implement but the accuracy of this method is greatly influenced by the choice of step size. The present work addresses itself to analytical determination of sensitivity derivatives of dynamically condensed mass and stiffness matrices of structures which are modelled with beam elements [8]. These derivatives are used to find derivatives of eigenvalues and thereby predict optimal values of modifications such that the modified structure has the desired shifted natural frequency. For a dynamicist, it is desirable that none of the natural frequencies coincide with excitation frequencies. The present note deals with shifting of an undesirable natural frequency in an optimum way. The proposed algorithm has an advantage that a shift in natural frequency and its prediction is accurate even when the magnitude of the desired frequency shift is large. A first step in this direction is the sensitivity analysis to identify effective parameters for modification. The present work utilizes the sensitivity derivatives of the natural frequency and also quantifies the effectiveness of various parameters. Further, optimal modification is approached as one of the two possible requirements.
Read full abstract