Predictive maintenance models are difficult to build off of bad data. The current life cycle of an issue requires maintenance technician to input data after the work is complete through a poor UI which has led to poor predictive models. Talking to the mechanics we learned about the challenges they face everyday, and realized the data being collected is not only useful for the government, but the mechanics themselves. Our voice-driven interactive assistant tracks jobs from start to finish while documenting activities, personnel, and findings. It does so by focusing on key words to collect data entry which, compared to natural language processing, is scalable to DoD size operations in terms of processing. Based on previous collected information on both the specific plane and the model, manufacturer databases, trouble shooting tree history, etc., the system helps mechanics narrow down on root problems as well as possible solutions. In addition, our system streamlines workflow by acting as an intermediate between shifts and auto-correcting information. We aim to integrate the data input and extraction into the workflow of a mechanic in a way that is useful to them and incentivizes them to input more accurate data. We envision the AI computer system working hand-in-hand with the technician to diagnose and fix problems. Presentation Material (please view on desktop):

Share this project: