A comparison of different NDE signal processing techniques based on waveform entropies applied to long fiber-graphite/epoxy-plates

This study compares di?€erent approaches for imaging the internal architecture of graphite/epoxy composites using backscattered ultrasound. Two cases are studied. In the first, near-surface defects in a thin graphite/epoxy plates are imaged. The backscattered waveforms were used to produce peak-to-peak, logarithm of signal energy, as well as entropy images of di?€erent types. All of the entropy images exhibit better border delineation and defect contrast than either the peak-to-peak or the logarithm of signal energy. The best results are obtained using the joint entropy of the backscattered waveforms with a reference function. Two di?€erent references are examined. The first is a reflection of the insonifying pulse from a stainless steel reflector. The second is an approximate optimum obtained from an iterative parametric search. The joint entropy images produced using this reference exhibit three times the contrast obtained in previous studies. These plates were later destructively analyzed to determine size and location of near-surface defects and the results are found to agree with the defect location and shape as indicated by the entropy images. In the second study, images of long carbon fibers (50% by weight) in polypropylene thermoplastic were obtained as a first step toward ultrasonic determination of the distributions of fiber position and orientation.