Pattern Recognition for Aircraft Maintainer Troubleshooting

The CAMS/GO-81 databases contain massive amounts of historical maintenance data that is unusable in the present form. The primary barrier to effectively using this data for troubleshooting purposes is the "free form" nature of the text fields that describe discrepancies and corrective actions. These text fields contain a wide array of "noisy" data and artifacts that prevent the effective query and search of the database without pre-processing and filtering of the data. The proposed approach seeks to develop a system that functions as a middleware application and enables the maintainer to retrieve useful legacy data in a format that is most useful for the maintenance environment. By using a data mining approach that is a hybrid collection of artificial intelligence (AI) techniques along with concepts from the field of natural language processing (NLP) it is possible to extract useful information from CAMS/GO-81. A combined hybrid approach is proposed since no single technique alone will provide the functionality that is necessary. The proposed approach will enable the Air Force to leverage this legacy data and to deploy a system that will evolve in strength as the system "learns" from the maintainers cognitive processes.