Kaufland Case – Global Datathon 2018

Data Case Introduction

Failures of small parts in a machine can lead to a costly breakdown of the whole machine. Therefore, it is important to perform maintenance checks on machines.  The usual approach is to do maintenance in fixed intervals, but this leaves the risk that some parts could fail just before the maintenance is due if the interval is too long, or that the maintenance does not find anything wrong if the interval is too short. Furthermore, different parts may have different lifespans, so a fixed maintenance interval for the whole machine may also not be the correct way.

The approach that solves these issues is predictive maintenance. Here, historical data for a component is used to predict when it will fail. Therefore, you can do the maintenance very focused on only the part that is about to fail and only just before it will break, greatly reducing maintenance costs.

 

 

 

Goal

Kaufland would like to apply predictive maintenance to robot forklifts used in one of their distribution centers.

You will be provided with data from six different forklifts. Is it possible to tell which machine is more prone to failure than the others? Are parts deteriorating over the timespan you are looking at?  Are some parts especially notorious for failure? These are just a few of the exciting questions that can be answered by you!

Details

More information about the Kaufland case can be found at www.datasciencesociety.net/the-kaufland-case-iot-predictive-maintenance/