Just as input from humans can add context to data and allow better decisions, it can also enable analytics to diversify, and make itself useful in a greater variety of applications.
"We're using advanced analytics in the commercial/contracts domain, live unit monitoring for daily, weekly and monthly reports, and in process monitoring. We're basically trying to shift from legacy tools that had some cumbersome problems to a more modern software platform," says Yugender Chikkula, technology manager for advanced process control optimization at Motiva Enterprises in Houston, which is a fully owned subsidiary of Saudi Aramco, and is North America's largest refiner at 630,000 barrels per day and runs the largest base-oil lubricants plant in the western hemisphere. Chikkula presented at ARC Industry Forum 2020. "We'd already been doing data analytics for a long time, so it isn't new, but most efforts have been at the unit and advanced control level for real-time optimization. The new set of digital technologies can take these condition-making processes and bring them up to solve problems in the higher levels of the organization."
Chikkula reports Motiva uses analytics to optimize its use of industrial gases, such as hydrogen, which is essential for its own oil and gas production. Its hydrogen consumption must be monitored in real time to minimize costs, even though usage decisions are often hampered by a lack of real-time data.
"Previously, we used our old friend, the Excel workhorse, bringing in data every week—or only at the end of the month for custody transfers, which made it very cumbersome to do calculations. Each month required more than 43,000 rows of data in Excel, while opening and closing these files used to take a couple of hours. Lots of time was spent on data wrangling and not much time on analysis," explains Chikkula. "There was no real-time visibility and no forward-visibility, which was causing us an economic penalty on our side."
Motiva uses different analytics tools, but for its hydrogen usage, it adopted Seeq's software because it:
-
Connects to the user's process historian, so there are no daily import hurdles;
-
Allows external data imports, such as contract volumes and business plans;
-
Combines actual-minute historian data with external data from vendors and organizational plan data;
-
Can create predictive models using all instances of data-produced optimized volume balances for the level of operations planned, and splits usage across contact tiers to give them the lowest overall cost;
-
As actual data replaces plan data, displays the effects of any re-balancing in real time with results readily accessible to multiple users; and
-
Makes historian data and external data available to test causations, which lets actual data be compared to plan data to explain any deviations from expected usage.
"This gives us near real-time visibility into what's happening with our hydrogen consumption. We know where we need to be, and we can predict how it's going to evolve over the next few weeks. This gives us handles to adjust takes from different vendors, and keep within limits," adds Chikkula. "We had good success with this solution and it had a big economic impact, so the next thing was to look at other gases and we extend it to modeling constraints on other contracts. We can also use this data for planning for unit changes or a test-run plant, and see how they'll impact contracts, if there will a be a penalty, and if we're better waiting. Previously, this kind of information never went into that kind of consideration."
Chikkula adds that Motiva also used Seeq's causal maps function for operational troubleshooting by feeding loop-cycling oscillations observed in a refinery unit into the software. It had been difficult to pinpoint the source of the oscillations earlier due to the unit's recycle streams, but the software's analytics non-invasively narrowed down candidates and better focused Motive's troubleshooting efforts.
"These causal maps allow us to analyze an issue by looking at its dependent variable and related independent variables, and it indicated causality that lets us prioritze how our investigation should proceed," adds Chikkula. "We talk a lot about ML and artificial intelligence (AI), but software tools that make it simple to connect to multiple data sources, where you can do data conditioning and data mining quick and easy, and build simple correlation models easily can make a big difference. Plus, they can go beyond solving engineering problems, and have a big impact on other parts of an organization.
Michael Risse, VP and CMO at Seeq, adds: "We're really entering the third stage of industrial analytics. The first was pencils, rulers and clipboards. The second was digital networks such as PLCs, DCSs and SCADA/HMIs with sensor data stored in historians and databases. The number one trending tool during this 30-year stage was an Excel spreadsheet running on a desktop PC. It was accessible, could handle calculations, and let engineers do what they wanted, such as cleansing, contextualizing and calculating derivatives and integrals, but a lot of intellectual property was tied up in those spreadsheets. The third stage begins with today's explosion of digital data, and delivers self-service analytics for subject matter experts (SMEs) by leveraging cloud-computing services and continuing software innovations. The cloud is accessible, elastic, scalable and low-priced, while software is integrating big data and accessible ML features, so users can crunch numbers better and find insights faster."
Risse reports the key to prioritizing the data and digitalization explosion involves plant engineers and other SMEs because they can determine which data, context and insights will be the most useful for their applications and organizations. "You can't bring process engineering expertise to big data scientists, so you have to bring data science to the users, who have the expertise to know the individual quirks of their processes and facilities, and can solve problems locally," explains Risse. "They can also ask the right questions to make sure an analytics project doesn't try to 'boil the ocean.' "
Risse adds that Seeq aids analytics projects because it can securely connect to data in dozens of different SCADA and DCS historians and contextual data sources like batch, MES and quality applications, so all the sources feel like they're in one overall system. "This gives engineers and other users a consistent virtual experience with their data," adds Risse. "It also means their data doesn't have to be moved, and is just fine where it is if we can connect to it, and saves users time by doing analytics without a lot of pre-work."
