Pattern Recognition for Aircraft Maintainer Troubleshooting

This project aims toward automated troubleshooting of aircraft maintenance supported by unprecedented connectivity: between the depot and the field, between maintainers and technical orders and other documents, and between novice and expert maintainers. Advances in semantic processing will automatically connect, for the first time, work units in the field with bills of materials and work control documents. The troubleshooter will converse with the maintainer in natural language as well as more conventional forms and text, analyzing the conversation, extracting topics, and then finding and presenting relevant topics from the large collection of technical orders.The front end of the troubleshooter will be designed to operate on devices as small as a pocket PC or smart phone, allowing point-of-task support in the field and in the depot.The dialog manager will tolerate a wide range of diction and will interact and choose data to present appropriately to the experience level of the maintainer. Novices will get extra help and exposure to best practices. Experts will do their job faster, while the troubleshooter learns from their experience, sharing their best practices with less experienced maintainers. The troubleshooter will be designed to learn on the job, improving service over time to all maintainers.