A hierarchical multi-stage data fusion approach to non-destructive evaluation of aero engine components

Abstract

This paper describes a multi-stage image analysis system used to monitor the condition of aero engine components. A hierarchical framework, of analysis requirements was constructed then algorithms were developed to fulfilthe requirements. The components are inspected using the fluorescent dye penetrant inspection (FPI) technique. This allows surface-breaking discontinuities to be detected, helping to ensure quality standards are being met. The multi-stage algorithm developments automate the FPI inspection using a data fusion approach. Raw image data, inspection knowledge and extracted image data are all fused to reach an inspection decision. The paper describes the main developments within the hierarchical process from image acquisition through to image processing, region selection, feature extraction, knowledge encoding and the final component inspection classification. Finally, the commercial implementation and performance issues associated with the system are discussed.