Originally published in Wine Business, August 2020 by Andy Starr

IN MAINTAINING HIGH-QUALITY STANDARDS, we all try to be more proactive than reactive. It's better to measure and adjust nitrogen levels at crush than to "unstick" a stuck fermentation caused by insufficient nitrogen one month later. But on the bottling line, many aspects are either hard to measure, such as corker jaw compression force or screw capper torque, or to identify, like worn but hidden parts. If we could "see" those hidden places and collect data, then we could make proactive decisions. Instead, we have been forced to wait until things go way out of spec and to pull out our wrenches to fix it.

Winemaking in the Information Age
Wine IoT (Internet of Things) is an ambitious long-term project undertaken by G3 Enterprises in Modesto to use data to avoid some of these common winemaking problems. I interviewed John Cunningham, G3's director of innovation, about its Wine IoT initiative, and about RoboBottle-its first product. Cunningham is a chemical engineer by training with 23 years in winery research and development, who brings a mentality more commonly seen in the pharmaceutical industry to wine. His relatively small team has developed several new products, including a screw cap liner to improve CO, retention and a way to use UV inks in place of wasteful foil stamping on labels.

Cunningham stated that "we are forced to make better and faster decisions today," something many winemakers already instinctively know. Increasingly, data has become part of that decision-making process-but if decisions are going to be data-driven instead of "gut" -driven, then data needs to be collected at every step.

Cunningham explained Wine IoT as an all-encompassing project that revolves around quantitatively understanding "how every aspect of the business affects profitability and the consumer experience." The company wants to collect good data from every part of the chain-from farming to harvest, through production and bottling, and all the way to consumption. The data then can be analyzed to see how much each part of the process affects others. He wants to collect and provide data that enables all winemaking decisions, ultimately to "give every owner of every piece of the supply chain the data they need at the moment they need it in a format to be able to make the best possible decision."

G3 Enterprises is partnering with wineries to expand its Wine IoT program and seeks additional partners. Once commercialized, Wine IoT will be a subscription service for wineries of all sizes.

RoboBottle
G3's RoboBottles are bottle-shaped devices loaded with sensors that collect real-time data about certain aspects of the bottling line. The Setup RoboB­ottle and Line RoboBottle each collect data inside the screwcapper, and the Oxygen RoboBottle collects data about total package oxygen (TPO) during the bottling process.

To understand why RoboBottle could be helpful and why it became the first G3 Enterprises' Wine IoT project, it's good to understand the current methods used to evaluate screwcapper setup and bottling line TPO.

G3 frequently received calls from clients that had issues with screw caps, like broken bridges, lifter problems, cut threads and pilfer-proof region damage. In every case, the damaged screw cap was the symptom of a capper head issue and not the actual problem. It is difficult, and sometimes downright impossible, to send a gauge into a capper head to accurately measure roller heights, spring pressure, and misaligned or worn parts-so to solve the problem a bottling technician will typically tighten or loosen something and hope it works.

Similarly, measuring total package oxygen means pulling a bottle off the line and reading the measurement a few hours later. This means that any identified problem can't be addressed until after many bottles have already been filled and sealed out of spec. In addition, there is no insight as to the source of oxygen pickup. A common but clunky "solution" to this problem is to jack up free SO, at bottling, hoping that it will consume any excess oxygen.

Cunningham thought the time was right to provide customers tools to set up the line properly and measure TPO in real time. Thus, the RoboBottle was created to collect the necessary data that allows a technician to easily understand and then optimize capper setup and TPO.

Setup and Line RoboBottles
The Setup and Line RoboBottles are each equipped with sensors to measure key parameters, such as top load spring pressure, thread roller height, pilfer roller height and misaligned parts. These devices communicate data wire­lessly to BottGuide, G3's RoboBottle dashboard app for laptops and mobile devices. BottGuide will lead the signed-in mechanic through the proper set-up procedure step-by-step. Now that the mechanics have good data, they can adjust their capper heads accurately, with data and settings locked in and saved in the cloud. Capper settings are dialed in and saved, where previously they often couldn't even be measured. It may take some time to initially get a 10 to 32 head capper in spec; but once completed, the Line RoboBottle will quickly check for drift away from the spec midpoint, continuously keeping everything in spec. Already, Cunningham has observed capper head varia­tions reduced by a factor of two.

Oxygen RoboBottle
The Oxygen RoboBottle works similarly, with sensors to provide data to determine how well your line is managing oxygen and where to focus improvement efforts. It uses photo-optic technology in which phosphor sensors are excited in the presence of oxygen, causing a change in light intensity. That change is converted into a quantifiable oxygen concentra­tion measurement. It's an instantaneous response collected by the same BottGuide dashboard, giving both the line manager and the product owner the ability to see TPO in real time and rapidly make adjustments if needed, instead of waiting hours for a lab result that doesn't identify the problem source. It marks the first time we can see TPO changes on a bottling line by machine.

Cunningham showed me a video in which the Oxygen RoboBottle was sent down the line, and TPO was continuously displayed on a dashboard after each step in the bottling process: rinser, nitrogen sparger, filler, capper. The dashboard lit up red when TPO was out of spec and green when in spec. Cunningham noted that "having actionable and timely data capability should lead to huge bottling line and wine quality improvements."

Over time Cunningham expects RoboBottle and Wine IoT will integrate with winery operations software to schedule things, like machine calibration and maintenance. Bottling run records will be automated. Both the co-packer and winery owner will have data about line performance. The battery-pow­ered devices even have their own charging ports, like a cell phone.

Asking the Process What It Needs
"We ask the process what it needs" is Cunningham's philosophy and approach to solving process problems. By that he means that every process can "talk to us" or reveal what it needs to run well. Good experimental design and data mining are how one can "listen" to it. His team designs an experiment, collects data then uses what is learned to design the next experiment. The process is repeated until the desired result is achieved.

The RoboBottle appears to be an impressive first deliverable for G3's IoT Wine team. It's based on actionable data and building quality into the bottling process. Like stuck fermentations and so many other winery processes, it's good to be able to solve a problem but so much better to avoid it entirely.