Volumetric inspection of moderately thick austenitic stainless steels by multifrequency eddy currents

Abstract

This paper describes the initial phase of a project to develop eddy-current methods to inspect welds joining sections of austenitic stainless steel pipe having walls up to 13 mm (0.5 in.) thick. The objective of this phase was to demonstrate the feasibility of detecting and characterizing flaws in austenitic stainless steel base metals. These materials and welds present challenging eddy-current problems because of their relatively large thickness and ferromagnetism. Multiparameter analysis shows that a reflection coil probe operated with three discrete driver frequencies and phase detection can locate and size a cracklike defect in a single conductor in the presence of variations in conductor resistivity, permeability, and thickness and in the probe-conductor spacing (liftoff). Experiments were performed with a modular three-frequency instrument. Flat-plate specimens of types 304L and 347 stainless steel machined to 12.7 to 15.9 mm thickness simulated pipe walls; saw-cut slots 10 to 30% of nominal specimen thickness simulated cracklike defects. The same slots were used in duplicate experiments as near-side (directly under the test probe) or far-side (in the face opposite the probe) defects. Flaw detection and characterization capability was demonstrated by a series of experimental measurements fitted to specimen properties by least squares techniques. The quality of the fit determined the expected accuracy of measurement. Comparison of accuracy estimates determined the best choice of operating frequencies. From the 1,2,5 sequence of frequencies between 0.5 and 20 kHz, the optimum set of operating frequencies was selected to be 0.5, 2, and 10 kHz. Estimates of measurement accuracy for combined near- and far-side defect cases were: plate thickness, 0.74 mm; probe liftoff, 0.03 mm; defect location (depth of material above defect), 3.48 mm; and defect size (vertical slot depth), 1.09 mm. A few property values were back-calculated from instrument readings; the errors in these values were somewhat larger than the measurement accuracy estimates because of instrument drift and the absence of calibration circuits.