Detection of Spring Support Locations in Elastic Structures Using a Gradient-Based Finite Element Model Updating Technique

The Location of Spring Supports From Measured Vibration Data

“…The eigenvalue derivative with respect to the updating parameters, and hence the sensitivity matrix, may now be calculated analytically. The advantages gained compared to the earlier studies [1,3] are as follows:…”

“…Therefore, the system mass and sti!ness matrices, and thus the eigenvalues, were not a continuous function of the updating parameters. This lack of continuity meant that the eigenvalue derivatives and the sensitivity matrix could not be computed analytically, and had to be estimated numerically [1]. This method was used for solving typical problems encountered in nuclear power plants and was also tested on a small laboratory set-up [3].…”

“…This paper presents the theoretical formulation for estimating support location, including the FE modelling of the problem and the validation of the method through the simulated and experimental examples cited in the earlier studies [1,3].…”

“…Sinha et al [1] presented a technique to locate simple massless translational spring supports in elastic structures, based on the measured natural frequencies of the structure. A gradient-based "nite element (FE) model updating technique [2] was used based on a correlated FE model.…”

“…Such an approach requires the formulation and computation of the sensitivity matrix ("rst order derivatives) of the cost function with respect to the updating parameters. Sinha et al [1] used the support locations as updating parameters, but only allowed the supporting springs to be attached at nodes of the FE models of the beams. Therefore, the system mass and sti!ness matrices, and thus the eigenvalues, were not a continuous function of the updating parameters.…”