ALPLA uses Crate to analyze sensor data to increase factory efficiency

ALPLA is a world leader in the area of packaging solutions, producing high-quality packaging for Coca-Cola, Unilever, and other brands in the food, drinks, cosmetics and cleaning industries. ALPLA employs 20,900 employees at 181 locations across 46 countries.

As the complexity of production machinery increases, so does the need for highly trained specialists in each factory. This creates the following challenges:

Higher personnel costs in every plant
Harder to recruit experienced talent at each location
Personnel turnover becomes more costly
Less experienced operators run the machines sub-optimally, impacting resource consumption and overall equipment effectiveness (OEE).

Inform "Mission Control" with real-time factory floor data

“We are running visual inspection systems in about every line of our plants. Having people constantly monitor these systems on the factory floor and reacting to any error is almost impossible,” said ALPLA General Manager, Jodok Schaeffler.

In 2016, ALPLA used manufacturing machinery data to help solve these problems. Each factory had 900 different types of embedded sensors.

Data from the sensors was used automate the machines in real-time, telling them when to eject defective products and informing nearby operators on the factory floor about equipment performance, which they could then tune and optimize.

The big idea at ALPLA was to integrate sensor data from multiple factories and production lines into a central “mission control,” staffed by experts who monitor production via interactive dashboards. Those centralized experts could give direction to less experienced, lower-cost factory workers.

A challenging use case for traditional SQL

ALPLA initially chose SQL Server as the data store for the sensor data, but it was unable to cope with their data requirements. They needed 900 different tables, one for each of the 900 different sensor data structures. Each production line generated thousands of readings per second.

Trying to query streams of that volume of sensor data in real time, and running complex machine learning analytics on terabytes of historic data, was outside of SQL Server’s sweet spot. Charts in the mission control dashboards were taking 3 to 5 minutes each to update, which was prohibitively slow since each dashboard contained a dozen or more charts.

From insight to action via tablets and HoloLenses

With CrateOM providing real-time operational insights in the mission control center, experts there could monitor, troubleshoot and optimize factory equipment efficiency remotely. Operating adjustments and other instructions were pushed to factory floor workers via tablets, and they could even use virtual reality equipment to connect with the factory floor.

“By collecting and analyzing the information from all of the sensors and cameras in these vision systems we can actually direct people on the floor to the 'hot spots' and with that, improve the waste rate and efficiency of production.“