A Laser Doppler Vibrometer (LDV) measures the laser Doppler frequency shift and converts it to the velocity at a point of a structure along the laser beam direction. In commercially available scanning LDV, the laser is redirected by a pair of orthogonal mirrors from one point to another, measuring the responses at these points sequentially. Continuous-Scan Laser Doppler Vibrometry (CSLDV) is built on scanning LDV; the laser sweeps continuously over a structure while recording the response along the laser path. The continuous-scan approach can greatly accelerate modal testing, providing spatially detailed vibration shape of the structure at tens or even hundreds of points in the time that is required to measure the vibration at a single point. However, extracting vibration shapes from CSLDV measurements is challenging because the laser spot is continuously moving. This technical difficulty and the equipment cost have become the major barriers that prevent the widespread use of CSLDV. Several algorithms to extract vibration shapes have been developed since CSLDV was introduced. Ewins et al proposed a polynomial approach that treats the vibration shape along the laser scan path as a polynomial function of the laser position. The polynomial coefficients were found from the sideband harmonics in the frequency spectrum of the acquired velocity signal. Allen et al proposed a lifting approach that collects the measured responses at the same location along the laser path. The reorganized measurements appear to be from a set of pseudo transducers attached to the structure. Hence, the well-established conventional modal identification routines can be applied to process CSLDV measurement. Algorithms based on linear time periodic system identification theory were explored as well. These algorithms are based on the fact that the measured velocities along the laser path are the responses of a special liner time periodic system when a closed, periodic laser scan pattern is employed. For the first time, this work compares these signal processing techniques employed in different applications using the same set of data obtained from a cantilever beam. The noise and uncertainty in the reconstructed vibration shapes are discussed in order to present the advantages and disadvantages of each method.
Learning Signal Processing System Using a Neural Network Model for Laser Doppler Velocimetry
