Non-destructive Techniques for Classifying Aircraft Coating Degradation.

摘要

In this research non-destructive techniques were examined as possible methods of determining aircraft coating degradation. Single Value Decomposition(SVD)-Linear Discriminant Analysis(LDA) algorithms were applied to measured spectra. When applied to infrared emittance spectra only 52% classification accuracy was achieved. When applied to Raman spectroscopy a higher classification accuracy of 70.4% is attained when using the same SVD-LDA algorithm. However the best performing measurement was using infrared reflectance classification accuracies were 100%, 99.83% and 94.4% when using the Bomem FTS, DRIFTS and Telops respectively for one of the sample sets. For DRIFTS data a more accurate fingerprint region was identified 865.6 - 1238.7 cm -1 decreasing classification error by 50%. Feature selection was applied to determine filter locations for multi-spectral measurements. Simulating the optimal and commercially available filters accuracies of 95% and 94% were achieved using 5 filters. Infrared reflectance produces high classification accuracy when using the DRIFTS, Bomem FTS, Telops and a multi-spectral imager.