Mahalanobis classification system for quality classification of complex metallic turbine blades

The complex geometry of metallic components combined with the variety of possible damage features limits the application of conventional NDT technologies. For parts with complex geometric shapes relevant product quality assurance tools are needed. Process Compensated Resonance Testing (PCRT) is an advanced and sensitive non-destructive evaluation method. It employs Mahalanobis Taguchi System (MTS) to classify the components as Good/Bad by evaluating the variations on resonance frequencies in Mahalanobis space. However, the process of feature selection and threshold determination in MTS is questionable. In the present paper, a two-stage Mahalanobis Classification System (MCS) approach is proposed coupled with binary particle swarm optimization procedure. The proposed MCS approach is applied to equiaxed Nickel alloy first-stage turbine blades with various possible defects. The obtained results demonstrate the high classification accuracy and evidence of the superior performance of the proposed approach.