Intelligent nondestructive testing expert system for aircraft carbon/carbon composite brakes using infrared thermography and air-coupled ultrasound

This paper discusses the use of intelligent system approaches, a combination of fuzzy logic and artificial neural network (ANN) techniques, for the automated inspection of commercial carbon/carbon (C/C) composite aircraft brake disks. The hybrid fuzzy neural approach was applied to the infrared thermography and aircoupled ultrasonic testing (UT) results obtained for C/C composite samples. This approach in detecting and classifying discontinuities follows a general pattern recognition problem that is based on three steps: segmentation, feature extraction and classification. To achieve this, discontinuous regions were first segmented from the non-discontinuous regions. After that, a 3 × 3 processing window was considered within the infrared thermography and air-coupled UT results. Then, five linguistic labels for the center value and mean value of the 3 × 3 processing window were generated as: very low, low, medium, high and very high. Finally, a Mamdani inference was applied for the fuzzification, rule evaluation and aggregation, and defuzzification processes. It was demonstrated that the 25 fuzzy rules developed were able to identify and classify the discontinuous areas in the C/C samples. It was also shown that this approach greatly reduces the noise, leading to better discontinuity recognition. After the successful.