Inspiration
Given Prompt: Drone Delivery Fleet Maintenance
A drone startup is experiencing mid-air failures, but they only investigate after a crash. Design a tracking tool for technicians to help them log wear-and-tear before and during flight, and manage and improve system performance and/or reliability.
How we built it
We built the Drone Fleet Reliability Monitor as a full-stack predictive maintenance platform powered by a robust Python statistical engine in app/qc.py that handles I-MR charts with all eight Nelson rules, EWMA tracking, and maximum likelihood Weibull life estimations. This engine feeds a high-performance FastAPI backend (app/api.py) which delivers real-time telemetry, cost-benefit curves, and dynamic upload parsing to a responsive React and Vite frontend dashboard designed to give maintenance technicians instant clarity on fleet health, part countdowns, and proactive replacements.
Challenges we ran into
We successfully integrated rigorous statistical process control—implementing all eight Nelson rules, EWMA slow-wear detection, and maximum likelihood Weibull life estimation with right-censoring—into a seamless, production-ready FastAPI and React application that transforms complex telemetry into actionable, technician-friendly maintenance insights.
Accomplishments that we're proud of
We are especially proud that our control charts successfully predicted special causes and provided early warnings for every drone component needing replacement before in-flight failure, catching critical anomalies that a standard Weibull distribution alone could not explain. Beyond the technical triumph of building a fully functional, production-ready predictive maintenance platform, we deeply value the incredible learning experience and teamwork that brought this project to life, ultimately delivering a polished solution that fulfills the true purpose of the hackathon.
What we learned
We are especially proud that our control charts successfully predicted special causes and provided early warnings for every drone component needing replacement before in-flight failure, catching critical anomalies that a standard Weibull distribution alone could not explain. Beyond the technical triumph of building a fully functional, production-ready predictive maintenance platform, we deeply value the incredible learning experience and teamwork that brought this project to life, ultimately delivering a polished solution that fulfills the true purpose of the hackathon.
What's next for Drone Delivery Fleet Maintenance
Through sharing our prototype with classmates and stakeholders, we gathered invaluable real-world feedback that will guide our next phase of improvements. Reviewing what other teams built also inspired new ideas such as adding a quick-export summary report for shift handovers and refining our mobile layout so technicians can easily log readings right from the flight line on their tablets. Integrating these insights will make the tool even more intuitive and robust for everyday fleet maintenance.
Log in or sign up for Devpost to join the conversation.