Digital image processing for the interpretation of radiographs used for condition monitoring

摘要

Digital image processing of radiographs has shown itself to be a useful aid to the analysis of indications and the verification of indication changes within the context of condition monitoring. With the help of special algorithms, objective and fast computer-based evaluation of crack-like indications is possible, even when the indications lack contrast and are in an area of wide density variation. The indications can be described in units of measurement: contrast, half-width and length. In contrast to previous ultrasonic methods, a threshold can be set which takes into account length and contrast; this allows discrimination between relevant indications and noise or geometric inexactitudes. This threshold can be set with great sensitivity (minimum contrast ΔD = 0.01 - determined over 5 pixels and length 1 td 10 mm). Radiographic indic'ations,of defects affecting wall thickness (cracks and undercut) can be detected in areas including excess penetration and mismatch. In cases of plant monitoring, radiographic indications can be compared quantitatively in terms of contrast, half-width and length in order to establish a reliable body of evidence.