January 1, 2004
Supply & Demand Chain Executive
By Andrew K. Reese
Real-time performance management promises to deliver greater information to high-level executives, but one chemicals manufacturer is building its performance management platform by starting on the plant floor
The phase separator at the ATOFINA Chemicals plant in Calvert City, Ky., did not appear to be properly forming a separation layer during the production process, and it looked like the quarter-million-dollar unit would have to be replaced. Elsewhere at the plant, a production unit was chewing through pumps at an alarming rate, generating a half-million-dollar bill for repair parts every year. All too common problems, all too frequent expenses.
In days past, lacking the tools to gain insights into the root causes of such problems, ATOFINA's management might have had to go on signing checks to cover these expenses. But now, with the help of a performance management solution, the company is finding ways to avoid previously inevitable costs and to glean new value from the equipment and machinery that keeps the plant humming. For companies looking to use real-time performance management (RTPM) to better run their supply chains, ATOFINA's experience demonstrates that RTPM often begins on the plant floor.
Squeezing Data from Closed Systems
Philadelphia-based ATOFINA Chemicals, a $1.4 billion subsidiary of ATOFINA (itself the chemicals branch of the oil group Total), has 17 manufacturing locations in North America and 16 worldwide. The Calvert City plant, situated along the banks of the Tennessee River in southwest Kentucky, began operations in 1948, when it was part of the Pennsylvania Salt Manufacturing Co. Today, about 270 people work at the plant producing Kynar polyvinylidene fluoride (PVDF) plastic resins used in manufacturing and construction, as well as the company's Forane line of CFC replacements used as coolants in air conditioners, refrigeration equipment and insulation foam blowing.
For some time the Calvert City plant has been using distributed control systems (DCSes) to manage its various processes and product lines, but Dwight Stoffel, principle plant electrical instrumentation engineer at the plant, says that those systems were essentially closed, making it difficult and time-consuming to extract data from them. "We did four or five projects where we went through the pain of getting the data from the DCSes," Stoffel says. "Those projects proved that being able to access that data would be very valuable. We just needed a better way to do it."
Data integrity presented another problem with the DCSes, which would store data for a maximum of seven days. "Once the seven days pass, it's gone and you can't get back to it," says Stoffel, explaining that the disappearing data could be an issue if, for instance, an engineer or manager at the plant wanted to study a piece of equipment and needed historical data for comparison. The only way to do that was to start from scratch and move forward, delaying any possible benefits that could be derived from the study.
Setting a Standard
Stoffel says that the various data-extraction projects that he undertook demonstrated the data not only could be useful for troubleshooting problems with equipment, but also held the promise of allowing the plant to lower its costs by improving its processes. With that business case in hand, he was able to get funding for a project to implement a solution that would make it easier to gain access to, and exploit, data from the plant's system.
At the time Stoffel was working on his pilot projects, unbeknownst to him, engineers in other parts of ATOFINA already were attacking their data issues by implementing a solution from a company called OSIsoft. Founded in 1980, San Leandro, Calif.-based OSIsoft got its start developing software for the petroleum industry ("OSI" originally standing for "Oil Systems Inc.") to collect, manage, archive and analyze data related to manufacturing processes, a solution that came to be known as the PI (for "Plant Information") System. Back in those early days, OSIsoft's solution ran on pricey VAX computers, which made it too expensive an option for a facility such as Stoffel's plant. But as PCs became more affordable and powerful, and as Windows became a widely adopted operating system standard, OSIsoft developed a new generation of its flagship software products to run on Windows NT, making it possible for a broader set of companies and facilities to adopt the software.
Stoffel looked at PI, among other solutions, and decided that OSIsoft best met the Calvert City plant's needs. He says he particularly liked PI's method of data compression and the fidelity of archived data, as well as the set of client tools that the solution offered for accessing data from the system. Meanwhile, at about the same time, ATOFINA as a company made the decision to move toward a corporate-wide standard for performance management as a way of facilitating collaboration between different plants and research and development sites around the world. Representatives from the company's European and North American plant operations met with ATOFINA's information technology group and, after reviewing the options, settled on OSIsoft's PI as the standard for the corporation.
Overcoming the Collaboration Challenge
Before ATOFINA could deploy the PI System in Calvert City, the chemicals company first had to put in a certain amount of hardware infrastructure to support the software. They had a half-dozen production units at the plant, and each had its own DCS, all from the same vendor, but the company had not tied the control systems together. ATOFINA acquired a server, a workstation that serves as the interface between the DCSes and the plant's local area network (LAN), and some additional DCS equipment (including a computer interface unit that hooked up to the interface server and the DCS LAN) to get the systems tied into the plants computing network, a process that took about six months. Once the necessary infrastructure was in place, though, Stoffel says it took the company all of about three days to get the PI software installed and running. Altogether Stoffel estimates that the company has spent about $200,000 on the project since its inception.
Training staff to use the system was not a big challenge, according to Stoffel. "It's a complex system, but when it comes down to actually using it and maintaining it, it's not difficult at all," he says. He adds that the plant extended the PI System to end users in virtually every department at the facility, including personnel in maintenance, logistics, engineering, lab, accounting, process technology and operations. "They all have access to it, and they all use it one way or another to do their jobs better," Stoffel says.
In fact, the biggest challenge in the project, Stoffel says, perhaps was cultural in nature. "You'll notice it says 'engineer' in my title," he quips, noting that he has been in process control at the plant for 25 years and that, for much of that time, a certain distance has prevailed between the engineering staff and the company's IT organization. Ensuring the success of the PI project meant overcoming that distance and collaborating more closely with IT. "We don't always agree with the way they do things," Stoffel admits. "But once I got involved in this project, it became apparent that we were going to have to work together to make it a success. So just learning to work with them and learning about their world and local area networks and how things communicate was my particular challenge."
The One-week Payback
Once the system was up and running, the immediate impact was to give the plant's staff and management direct access to data on the performance of the facilities production equipment. That resulted in immediate time savings. "Without the PI System, if you wanted to see what was going on with the DCS, you had to go over to a control room, shove an operator out of the way, take over a console and start looking at things," Stoffel explains, tongue only slight in cheek. "But PI has provided a way for everyone to have that information on their desktop."
Moreover, the plant's personnel gained a greater ability to manipulate the data to glean whatever information they might require. Previously, if they wanted to acquire information not already provided by the control system, they would have to put in a request with the DCS specialist and then wait while their request worked its way up the specialist's task list. With the new solution in place, essentially anyone at the plant could configure the software to pull in the required data and make the necessary calculations to get the information they needed. As important, the system archived all the data it collected from the very first day, providing plant staff with access to the kind of historical data that the DCSes alone could not provide.
The utility of having this type of tool in place quickly became apparent. In the first week following the implementation, the plant experienced a process problem with a phase separator. Initially the plant's staff believed that the separator was functioning properly, forming the separation layer, but that the instrumentation on the unit simply was not recording the process properly. "Well, it was my instrument, and I wanted to prove that it was right," Stoffel says now. "But not being around 24 hours a day, I couldn't catch it at times when it showed that it was working." After working with PI, within a week Stoffel proved that, in fact, the instrument was functioning properly, prompting the plant's staff to look into replacing the separator itself with a larger unit, a $250,000 proposition. However, a process engineer subsequently used the solution to prove that the fault lay elsewhere, requiring a much less expensive fix and a significant cost-avoidance. "So probably starting off in the first week, we paid for the PI System," says Stoffel.
In another instance, one of the plant's production units was chewing through pumps, resulting in an annual bill for repair parts in excess of $500,000. A mechanical engineer at the plant used the PI System to determine that the pump currents in the unit, normally fairly flat, would periodically start "getting nervous," indicating that the pump was cavitating for one reason or another and resulting, in short order, in damage to the pump.
With that knowledge in hand, the engineer put together an application using a PI add-in to Excel that pulls data every six minutes from the monitoring solution into the spreadsheet application to show the status of some 100 pump currents. He combined this bit of handiwork with a dynamic link library (DLL) file that he found on the Internet to create a system that pages him whenever one of the pump currents starts getting "noisy." The engineer is able to access the PI System from work or his home computer to check on the troublesome pump. In addition, he used IP pop-up software he found on the Web to create hot links to troubleshooting sheets specific to whichever pump is causing the problem.
With that information in hand, the engineer can respond quickly to any anomaly by contacting an operator at the plant to take action before the pump dies. Moreover, over time, the engineer was able to use the collected information to identify designed-in problems in the equipment that now have been corrected. The result: a $363,000 year-over-year reduction in pump repair parts for the unit in question.
From "Trust Me" to Hard Savings
Elsewhere, users in the plant's purchasing department are able to assess some of the major raw material inventories from their desktop for use in scheduling orders. The environmental and safety departments use PI to monitor aspects of compliance and for incident investigations, and the plant has plans to interface the solution with an environmental management system and online release modeling software.
In discussing the return on investment (ROI) on ATOFINA's investment in the PI System at the Calvert City plant, Stoffel says that when the initiative began, "It started off as one of those 'trust me' projects. But before I got it put in, they were doing the five-year plan, and somebody decided that in five years — and I was kind of concerned about this — that this thing would be saving $500,000 a year." In fact, Stoffel says that he can probably claim more than $1.8 million in one-time savings and recurring annual savings of $590,000.
In addition, the solution actually has provided a platform for measuring ROI on other investments, too. "When I first started trying to get distributed control systems put in," Stoffel recalls, "they said, 'Justify it. Prove to me that it's going to make a difference.' Well, I didn't have any information, any good data to show what we were doing then and what we could do in the future. So besides returning investment itself, PI also provides a way to prove return on investment in other projects."
Too Much Information?
Do PI and similar solutions carry the risk of overwhelming end users with all the data that this type of system can churn out? "Yes, that's possible," admits Stoffel. But the engineer — who lives by the axiom that "data is not useful to anyone, information is useful to some, and knowledge is useful to all" — says that the key is to learn what data is important to which person and to help that person turn the data into useful knowledge. At the same time, Stoffel believes in archiving more rather than less data. Currently the Calvert City's system is archiving 16,000 individual data points, or tags, in the PI solution. "Maybe some of them you really don't need, but you never know," he says. "The key is to throw it all in there, and then someday, if you have a safety incident that you have to investigate, you can use all the available data to try to get back to a root cause and solve that situation."
Looking ahead, Stoffel says he is exploring new ways to increase the utility of the PI System to the business side of the company. He's started with a small project to merge information from the plant's SAP enterprise resource planning (ERP) system (such as daily production numbers and yields) with PI information (specifically, data on energy use) to generate daily reports on how much energy it takes to generate a pound of a certain product. Plant managers can monitor those reports and use the information to find ways to improve the plant's efficiencies, with the potential for much greater returns for ATOFINA. The plant also has long-range plans to take advantage of OSIsoft's certified interfaces to SAP and to use PI to supply greater information to the company's supply chain.
Of course, beyond the hard and soft savings that the Calvert City plant has achieved using the OSIsoft products, the PI System has had one additional benefit that, for Stoffel at least, may be priceless. "This is the one piece of software," the veteran engineer says, "that I don't hear people complain about."