An expert system for power plant NDE

Imaging and inverse problems of electromagnetic nondestructive evaluation

“…Generally the measured data is compared with a previous known set of reference measurements which each correspond to a certain type and size of flaw, .e.g., a 1 mm long fatigue crack under a rivet The character of the anomaly of interest is then determined from the reference standard which most closely matches the test signal. This process is generally performed with a computerized calibration scheme which relies on signal processing and large amounts of a priori knowledge [8][9][10][11][12][13][14][15].…”

NDE data fusion using phenomenological approaches

“…Alternative approaches involve treatment of the characterization problem as a classification problem. Such approaches use such tools as neural networks [64,65,101] and pattern classification [102,103,117] for flaw identification and classification. A variety of transformations and signal processing techniques [63,108] are used for extracting features that are relevant for discriminating between signals.…”