An intelligent NDE system potentially improves flaw characterization in ultrasonic testing, or enhances detection of acoustic emission signals from growing cracks among a noisy background. The advent of high speed digital data acquisition of waveforms from the sensor, and availability of high performance personal computers for data processing, have fueled advances in building such systems. Data are typically acquired at a sampling rate up to 100 MHZ, and a resolution up to 12-bit. It is particularly difficult to enhance signals in continuous data acquired with a conventional strip chart recorder. However, signals acquired in discrete (digital) data may be enhanced by an appropriate DSP technique. A DSP technique evolved in future applies to digital data acquired in the past. Such digital simulations in a virtual' experiment minimize costly repeats of a real experiment. It permits feasibility of a forensic' detection of today's signal among the yesterday's noise.

Two main DSP techniques, to study digitized data, are based on the time correlation in the time-domain, and the fast Fourier transform in the frequency-domain. The acoustic background noise is examined by statistical techniques based on fluctuations and deviations. Pattern recognition techniques, combining statistical and DSP techniques, are then applied to enhance detection of signals buried in a highly noisy background. This paper presents specific examples of intelligent NDE systems, based on these techniques, for ultrasonic evaluation and acoustic emission testing.