On the Influences of Boundary Reflections and Piezoelectric Sensors to the Characteristics of Elastic Waves for Pattern Recognition Methods

Abstract

In this study, the influences of boundary reflections and piezoelectric sensors on the detection of propagating elastic waves emitted from newly formed damage surfaces are demonstrated using numerical and experimental models. The concepts of the displacement-based, time-domain equations of non-splitting perfectly matched layer (PML) and spectral finite elements are utilized to facilitate computationally efficient modeling for high frequency wave propagation in large scale structures. The PML is an artificial layer, which provides an exponential decay regardless of frequency or angle of incidence and does not produce any reflection from the interface. The numerical result under ball drop impact excitation is validated with experimental measurements. The boundary reflections within small scale-coupon specimens modify the received wave characteristics such a way that transferring the laboratory scale data to other geometries for pattern recognition methods becomes a challenge. The features used in the pattern recognition algorithms cannot be scaled up to large-scale structures unless the experimental geometry is designed through considering the target wavelength, and the waveform record-window is limited to prevent the influence of reflected waves as well as the transfer function of piezoelectric sensor is taken into account. A model based on selecting the sample size as a function of wavelength is developed. The influence of piezoelectric sensors to the output signals is demonstrated such that the selected piezoelectric sensor controls the waveform features, which further limits the test repeatability under different conditions.