OSISoft: old-school process-control company on IoT cutting edge
- 03 May, 2018 02:52
While the eyes of the tech world are on the usual suspects like Google and IBM, as well as high-profile operational tech firms like GE and Siemens, an almost 40-year-old company called OSISoft has quietly leveraged its expertise in process monitoring and management into an impressive list of prominent clients. These include Aramco, the national oil and gas company of Saudi Arabia, which is arguably the single most valuable and profitable company in the world, Chevron, Pacific Gas and Electric, Heineken, Tyson Foods, and Lawrence Livermore National Labs, among many others.
OSISoft has a lot more in common with the GEs of the world than with some of the IT-based powerhouses that are making a lot of noise about IoT these days. It’s a company with a long history of experience in real-time data collection, making the transition into an IoT-enabled world a smooth one.
IoT automates the science of beer
The PI System is the foundation of OSIsoft’s offerings, a structured data lake for real-time information that can let businesses hunt for trends and potential cost savings down in the weeds of machine-generated data. It’s designed to digest information from a huge array of potential sources – anything from medical devices to oil rigs – and get it into a usable format.
It’s helped Tim Alexander, brewery operations technology manager at Deschutes, trim production time and improve quality control at the company’s main production facility in Bend, Oregon, where about 350,000 barrels of beer are brewed each year. (Deschutes is the 10th biggest craft brewer in the country.)
Like OSISoft itself, Alexander’s brewery operations department traces its lineage to a line of business unit, not the IT shop. It’s not just managing the brewing equipment, according to Alexander – the group is directly involved in producing the beer itself, maintaining databases of ingredients and recipes and monitoring numerous variables present in the complex brewing process.
Deschutes was doing this very much by hand in the past, writing fermentation data into an Excel spreadsheet and trying to make predictions based on this. But it was too much work to perform for each and every batch of beer that went through the system, so lots of data was lost.
The main thing that OSIsoft lets the brewery do is get access to a lot of data quickly, Alexander said.
“We solved a problem with our thousand-barrel fermenters right away [after implementing PI], which cut a full day off of fermentation time,” he said. “That enabled us to produce a lot more beer in a year.”
The work on implementing OSISoft’s PI system – with an assist from some consultants as well as Microsoft, whose Azure back-end is used by Deschutes – essentially expanded the brewery’s asset framework beyond the cellar tanks to include a much higher proportion of the company’s brewing apparatus.
At first, according to Alexander, the team wasn’t sure how specific it needed the fermentation model – created from the growing stock of available brewing data – to be.
“Working with OSISoft and the data scientists at Microsoft, we eventually realized that really all we needed was the brands and specific gravity measurements in order to produce models with the accuracy that we were looking for,” he said. “But we’re still sending all of those tags up to the cloud, because we’re thinking we may use them for future predictions on future fermentation steps.”
Process engineers peek inside the box with IoT
OSISoft, historically, has been a process-control tool, according to long-time customer Larry Shutzberg, who is the CIO and executive vice president at GEC Packaging Technologies, a conglomeration of three large packaging firms.
About a decade ago, OSISoft started to turn process control into something more proactive, able to detect trends and potential problems before they occur, Shutzberg told Network World.
“My job is to turn data into information into knowledge into action into dollars,” he said. “OSI’s my cash register.”
When Shutzberg started at GEC, all the machine-generated information was in the form of 24-character, cryptic codes, “one long list in an Excel spreadsheet, and if you didn’t work there for 40 years, you had no idea what any of them meant.”
GEC collects 120,000 points of data in real time at its Pine Bluff, Arkansas, facility, but Shutzberg said that’s just the start of the possibilities.
“We can actually contextualize and dimensionalize that data by augmenting it with ERP data, with Internet data, and then that allows us to start asking questions like ‘what’s the true consumption of energy?’” he said.
There’s a cultural change aspect to the whole thing, he noted, given the tension between older process engineers who have tried-and-true methods of accomplishing tasks and younger ones who want to reinvent those methods.
But it's tough to argue that the payoffs haven't been worth it. Shutzberg gave the example of digging into the specific operational details of three separate wood digesters that were performing at different levels, and being able to identify why one machine was outperforming the others. “Sure enough, we’re now making $1 million more a year in pulp.”