Pattern Recognition for Aircraft Maintainer Troubleshooting

A significant maintenance problem is to prescribe the best corrective action for a problem; thus requiring a system to be capable of reasoning about a history of symptoms and corrective actions. Considerable time and cost savings could occur if maintainers were provided with the best corrective action given a problem. Since much of the symptoms and corrective actions are recorded as free-form text, a solution must be able to interpret that text, reason about the knowledge presented in the text, and select the best corrective action. COLLT (Computational Language and Learning Tool) will be a robust and extensible natural language and learning technology for CAMS that re-uses the Corrective Actions based on the same or similar Discrepancy. COLLT is built on computational linguistics, inductive logic programming, and machine learning technologies. COLLT derives normalized representations of the discrepancy/corrective action fields from CAMS and uses these to pattern match between Corrective Actions and Discrepancies. These normalized representations form the example set where machine-learning algorithms can compute generalizations about discrepancies and corrective actions based on CAMS. Hence, COLLT becomes more proficient over time. This innovative integration of technologies will result in a significant new capability for the aircraft maintainer in identifying the appropriate course of action.