Agency Partners
The business partner was a subsidiary of a top provider in paper converting and packaging solutions. Our project aimed to tackle ongoing problems with crucial paper forming machines that were causing frequent disruptions. Our main aim was to carefully understand what was causing these machines to fail every so often. We wanted to make them more reliable and reduce the amount of downtime.
Gabor
Project Manager
/ Agency /
USA
Manufacturing
In a busy production environment with a hundred-meter-long machine full of real-time sensors, we faced the challenge of making sense of the data. Despite having no clear target and lots of noise in the data, we took on a detailed analysis. We used various advanced analytical models to find patterns in the data’s fluctuations with the aim to step in before problems happened.
We successfully set up a reliable alert system that works about 85% of the time to warn about upcoming issues. We also created a predictive maintenance system to spot key factors that could lead to machine failures, in order to take action before problems occur and reducing disruptions.
In the pilot project, we managed to create the alert system and identify key factors leading to machine breakdowns. However, because the system is complex, we realized that we need to look into more solutions for a complete system implementation to keep operations running smoothly.
We ran out of memory during the algorithm training period and we rebuilt the whole infrastructure in AWS. The data pipeline runs in a Metaflow framework that gives us the flexibility of operating either in the cloud or on premise. Data operations and the models are coded in Python.
We've identified 100+ world-class agencies, drop us a line if you'd like to work with them.
