Vibration-Based Quality Assessment of Metallic Turbine Blades Considering Measurement Uncertainty

Assessing the structural quality of metallic turbine blades is a challenging task due to their complex geometry and wide range of possible defect features. Process Compensated Resonance Testing (PCRT) is an effective quality assessment tool that uses a broadband sinusoidal swept input to excite the resonant modes of the component and employs a supervised learning algorithm to interpret the resonant modes and as such to determine the structural quality of the component. Our previous work has mostly centered on the exploitation of classifiers that are based on Mahalanobis distance. However, in practice, the measurement uncertainty may lead to bias in the trained classifier, potentially resulting in misclassification of the turbine blade. In this study, the concept of interval Mahalanobis space is introduced in the classifier in order to cope with measurement uncertainty. The resulting Integrated Interval Mahalanobis Classification System (IIMCS) classifier employs BPSO to screen those resonant frequencies that contribute favorably to the system and analyzes the sensitivity of resonant frequencies to measurement uncertainty under a Monte Carlo simulation scheme. The developed classifier was applied to an experimental case study of equiaxed nickel alloy first-stage turbine blades with a range of defect features, showing a good and robust classification performance.