Computationally Intelligent Image Processing Techniques for Crack Detection in Structural Components from Imaged Data

This effort focuses on automatically detecting cracks in structural components from imaged data. Currently, fluorescent penetrant inspection (FPI) identifies cracks in engine components such as blades, disks, casings, etc. Alternatives to FPI are needed and advanced, non-destructive inspection (NDI) methods are under development to improve damage detection limits in an automated manner through the incorporation of computational techniques. Ultrasonic excitation is utilized in conjunction with infrared (IR) imaging to detect crack-like defects using Sonic Infrared (SIR) imaging. An automated process to detect cracks given a sequence of SIR images is developed. This process sorts the SIR images, performs a reference subtraction, calculates the largest image difference, creates structural elements, performs morphological filtering, and labels potential crack locations. The outlined crack detection method developed is applied to 14 industrial type blades from SIR image sequences of representative turbine engine components. Computationally intelligent image processing techniques were successfully able to identify cracks in all 14 blades, where the blades were previously assessed using FPI. False crack identifications due to localized heating of the holding mechanism of the part were also observed in addition to the actual crack locations. With the addition of other image processing techniques, these false-calls have the potential to be reduced. Using image processing techniques to automatically detect crack-like defects is a viable tool to use as an alternative method, or in conjunction with FPI, to improve damage detection, reduce manpower requirements, and chemical usage associated with FPI.