Session: 01-08-01: Vibration and Acoustic Measurements, Signal Processing, and Test Facilities

Paper Number: 67591

Start Time: Wednesday, 01:00 PM

67591 - Spatial Localization of Air Inclusions in Carbon Fiber T-Beam, by Use of Wavelet Entropy Time Series From Hammer Tap Test Localization of Air Inclusions in Carbon Fiber T-Beam, by Use of Wavelet Entropy Time Series From Hammer Tap Test. 

Many critical safety structures such as aircrafts are constructed using extensively carbon fiber; therefore, non-destructive testing (NDT) of carbon fiber structures is imperative. The following paper investigates a new method of NDT based on the analysis of a vibration signal after exciting a carbon fiber specimen with a pulse. The analysis proposed in this paper consists of transforming the signal using the Haar Wavelet Transform, segmenting the Transform in the time domain and calculating for every frame its Informational Entropy, as defined by Shannon (Wavelet Entropy Time Series). The use of Wavelet Transform in combination with Shannon’s Informational Entropy (Wavelet Entropy) has been proven to identify irregularities in signals, mostly when comparing them to healthy signals using the Kullback–Leibler Divergence. Wavelet Entropy Time Series requires no additional signals for comparison.

In this paper, notions such as Wavelet Entropy and Wavelet Entropy Time Series are introduced and defined. The presupposition that Wavelet Entropy Time Series can single out and localize damaged areas of a structure is backed up both theoretically and experimentally.

 

On a theoretical standpoint, by applying an abrupt force at a specimen (e.g. conducting a hammer tap test), an elastic wave that resembles a pulse is generated. This pulse when interacting with geometric discontinuities, such as cracks or air inclusions, will get scattered. The lesser the distance of the discontinuity to the source of the excitation, the more potent the scattering. This results in a dispersion over time of the energy distribution of the wave and therefore, the energy of the signal acquired by an accelerometer. The argument made in this paper is that by calculating the Wavelet Entropy Time Series, the aforementioned dispersion can by quantified by viewing its trend. When exciting an undamaged section of the specimen, the energy distribution of the signal tends to shrink which according to the definition of Shannon's informational entropy causes the Wavelet Entropy Time Series to decay. For a damaged section, the opposite is expected, meaning an increasing trend in the Wavelet Entropy Time Series.

 

To prove this experimentally, a carbon fiber T-beam containing air inclusions was subject to hammer tap tests across its length. Accelerometers were mounted along its long axis. For every excitation, these accelerometers measured the vibration signals, which were then analyzed using Wavelet Entropy Time Series, and viewing its trend. The analysis was capable of localizing 2 locations of interest where the wavelet entropy time series follows an increasing path, thus making us suspect the existence of air inclusions.

 

Non-destructive testing (NDT) of composite specimens can be a tedious task, as such, any new NDT method can be a new tool in the arsenal of an NDT inspector. The applicability of wavelet entropy time series for NDT could also probably extend outside of carbon fiber to other composite materials and even in metal structures. The method, should it be developed commercially, could be a low-cost, reliable testing technique that requires just a computer, an accelerometer with relatively small sampling frequency of 5000Hz as well as little to no specialization on behalf of the tester.

 

Presenting Author: Spyridon Brouzas National Technical University of Athens

Authors:

Spyridon Brouzas National Technical University of Athens
Ioannis Georgiou National Technical University of Athens