New Tool: ProSource
Checkout our packaging and processing solutions finder, ProSource.

Deschutes Brewery identifies flexible production solution for the changing market

Looking into the future and understanding consumer behavior in the food and beverage industry are hearty challenges for companies as dramatic changes drive enterprises to find flexible production solutions. One compelling consumer behavior change within the craft beer sector is brand loyalty. Nowadays craft beer consumers keep moving from one product to another, continuously searching for the “next new beer” to share with friends. 

For craft brewers like Deschutes Brewery and its two brewhouses, these shifts in consumer behavior are driving the need for brewmasters to test new beer varieties and find production time to experiment with new product batches. Deschutes is the seventh largest craft brewer in the United States and decided to optimize beer production via OSIsoft’s PI System, a data-modeling solution for processing and manufacturing operations. 

The choice allows Deschutes to invest in software as a service (SaaS) and avoid the large capital-expenditure dent at its Oregon facility, while the company plans to open a new production facility in Virginia.  

The brewing process at Deschutes includes many phases and transitions, such as filling, fermentation, free rise, cooling, diacetyl rest and maturing in tanks, to name a few. For the initial rollout, the craft beer producer decided to bring in the data modeling for just one stage of the process: the fermentation to free rise transition. During this stage, operators conduct, prepare and perform analysis on samples to understand when the fermented beer can move to the free rise phase. This step requires operators to manually measure the apparent degree of fermentation (ADF) — the density of a batch. 

The OSIsoft data-modeling and visualization solution retrieves data from processing equipment and production lines, stores critical information, and produces analysis of time-series/event-based data in real time. This scalable solution can also store millions of data points and interface with more than 450 off-the-shelf proprietary interfaces. 

“We wanted to start small and focused on just one brand and brought ADF data into the data-modeling solution,” said Tim Alexander, assistant brewmaster at Deschutes Brewery, during the recent OSIsoft users conference in Germany. “The team had a specific time frame of the data and applied a very simple linear regression over the top of the information.” 

“The question was can we use all this data — manual and sensors — that we've been collecting to predict when these transitions are going to occur?” said Brian Faivre, brewmaster at Deschutes Brewery, at the OSIsoft user conference. “If so, we needed a sophisticated model that was going to adapt and learn over time.”

During the initial data-modeling demonstration, Deschutes decided that it would undertake this data-modeling process with each specific beer brand to keep the variables limited.  

“The brand was the key indicator,” said Faivre. “If you looked back at fermentations for the same brand, before modeling, we were really close and had a good idea. Then, we ran the data-modeling curve, and it would say where this fermentation point is going to be.” 

For this process, Deschutes entered all of the processing data — past operator manual measurements and fermentor framework — into the modeling platform and produced data context via the PI Integrator for Microsoft Azure and a SQL data warehouse. The integrator tool for Microsoft Azure is a business-intelligence solution that runs multiple analyses based on Cortana Intelligence and identifies operational patterns, dependencies and correlations. 

Deschutes Brewery now estimates a loss of 72 hours of production capacity per fermentation if they didn't accurately identify this critical transition, not to mention quality issues. 

“Brand creation is getting bigger, and all the trends in the craft industry point to consumers drinking what’s new," said Faivre. “Now we'll be able to have a data curve for an established brand and say, ‘Apply that curve to this new brand,’ and quickly utilize this functionality without a lot of data behind it.”