Demonstration of model-assisted probability of detection framework for ultrasonic inspection of cracks in compressor blades

Abstract

The fatigue crack detection capability of nondestructive testing was quantitatively evaluated using the model-assisted probability of detection (MAPOD) process. A framework based on conventional MAPOD with dimension reduction was studied. By conducting a metamodel-assisted sensitivity analysis, the dimension of the physics-based model was reduced. The factors selected based on the sensitivity analysis were partitioned for both computational and physical experiments. Subsequently a transfer function was constructed. Additionally, a linear model of the relationship between the ultrasonic signal and crack size was obtained and calibrated by the transfer function by conducting physical experiments. The probability of detection curve and the corresponding detectable crack size were evaluated using the calibrated model. Furthermore, the improved MAPOD process was demonstrated by the fatigue crack detection in a J85 engine compressor first-stage blade tang hole.