A signal model adapted ESPRIT algorithm for joint estimation of spatial and temporal parameters in vibrational analysis of cylinders

This paper deals with the joint estimation of spatial mode shapes and temporal frequencies from transient vibrations measured by a uniform circular array of sensors encircling a cross-section of a cylinder. The geometry allows each 2D mode to be interpreted as two real planar waves impinging from mirrored directions with respect to the broadside of a uniform linear array. Algorithms for joint estimation tend, in the presence of noise, to produce signal roots that express planar waves impinging from only approximately mirrored directions. A recently published 2-D ESPRIT algorithm is modified to take advantage of the geometry, and to force the estimated planar waves to impinge from exactly mirrored directions. This reduces the estimation errors; the main advantage is more easily interpreted results. Memory requirements and complexity are significantly lower for the modified algorithm, since the decomposition of a large data matrix may be broken down into several small ones.