Team Hobo: Predictive Maintenance Through RL Employment

Inspiration

What it does

We traversed different ways of tackling the predictive maintenance problem, We did quite a lot of preprocessing, We did web scraping, used reinforcement learning and generated systems that can tell how often and when turbines are to be maintained to optimize the rewards (productive efficiency)

How we built it

We used interactive python, Flask, and ReactJS. Using a diverse set of frameworks from Selenium to Transformers.

Challenges we ran into

The original dataset didn't have features viable for predictive mainteance

Accomplishments that we're proud of

We filled the aforementioned gap through novel methods such as Q-Learning application in this area, which we saw no instances of online.

What we learned

To work as a team, to have fun, and to do basic and complex system design using reinforcement learning. We also learned some novel preprocessing ways such as creating embedding similarity matrices

What's next for Team hobo

We want to consider entrepreneurial options regarding this project, and eventually deploying it.

Built With

• python