Defeat undetected downtime
Manufacturing intelligence systems improve productivity, ensure stated goals deliver value
By Malcolm Wheatley, senior contributing editor -- Manufacturing Business Technology, 5/1/2006 12:00:00 AM
In 2002, Australia's Coca-Cola Amatil—the Australasian region's Coca-Cola bottling franchise—recognized that the bottling and canning lines at its New South Wales facility were critically close to their capacity limits. Yet it wasn't a propitious time for Amatil to engage in a major upgrade to the plant, one of six that it operates.
Instead, as part of a continuous-improvement program under the banner of Manufacturing Excellence, the impasse was resolved with a manufacturing intelligence solution from Activplant. Rather than buy new equipment, the focus would be on delivering the hard data and operational insight necessary to boost the capacity utilization of the Northmead plant's three high-speed filling lines—portions of which dated back some 30 years.
The Activplant solution was implemented first on a 500 bottles-per-minute plastics line, explains Coca-Cola Amatil Manufacturing Analyst Wayne Luxford. Detailed data from various points along the line was routed to a dedicated database server for analysis with Activplant tools—captured using the plant's existing PLC network. Substantial productivity increases followed within weeks, says Luxford, and in early 2004 the project was extended to the other two lines in the plant—with similar success.
It proved possible, he explains, to achieve a 97-percent decrease in the number of jams on a particularly problematic section of conveyor, while a 14,000 case-per-week improvement in output was achieved by rectifying some in-feed problems. And on the bottling line, says Luxford, gross machine efficiency—roughly equivalent to overall equipment effectiveness (OEE)—increased 15 percent. In short, unplanned downtime halved.
"Because of the granularity of the data you're working with, you get to see—and resolve—issues that you'd never have spotted otherwise," notes Luxford. It turned out, for example, that the combination of a troublesome valve and a programming problem with a PLC was momentarily stopping a filling line 300 times per week. The Activplant solution highlighted the issue, which hadn't attracted much attention before, because the duration of each stoppage was so short. Collectively, says Luxford, the resolution of such problems contributes to the 10 additional hours of production time that the line achieves per week. "And at 2,000 cans per minute, that's quite an impact," he says.
A handful of ISVs
To the small but growing body of vendors selling manufacturing intelligence solutions, stories like Coca-Cola Amatil's are business-as-usual.
"Give people real-time data in a common format, then provide the tools for them to analyze it, and they'll start to see things they could not see before," says Gregg LeBlanc, director of product marketing at OSIsoft. "It's very genuine low-hanging fruit. Suddenly, they have all the data they need to ask the right questions, and figure out what's really going on."
This clarity of vision extends to operations, as well as to providing data for continuous-improvement initiatives. Manufacturing intelligence, notes John Lynch, senior systems specialist at the Northumberland, U.K., manufacturing facility of pharmaceutical company Merck Sharp & Dohme—which uses Parsec Automation's TrakSYS application—"drives performance improvements because it delivers the information with which to make individuals accountable for that performance."
And if "low-hanging fruit" is a common theme when talking to manufacturing intelligence vendors, another familiar refrain is the magnitude of the financial return. "Compared to the ROI, the cost of the license fee is a rounding error," says LeBlanc. "You'll see a $100,000 implementation payback in three to six months."
Indeed, notes John Hall, director of enterprise lean systems at Elkhart, Ind.-based NIBCO, a 3,000-employee, 12-plant manufacturer of flow-control plumbing products, "The trick is convincing senior executives of the ROI projections."
When Hall went with a manufacturing intelligence system from Informance International, his first two ROI calculations—based on estimated improvements in OEE—showed a payback time of a month or less. Faced with management skepticism, Hall's third stab at the ROI showed a much more conservative—and conveniently more plausible—payback time of six to eight months, based on a reduction in inventory levels. Submitted, it got approved.
The July 2004 pilot implementation of the Informance solution actually conformed to Hall's earlier expectations. The implementation, in a capacity-constrained cell for 't'-plumbing fittings—which operated on a 24/7 basis, and on which the company was anticipating spending $750,000 for another forming press—delivered in spades.
First, the solution identified that the existing forming press wasn't the constraint at all: instead, it was another machine altogether—a loader. Within two months, management boosted the percentage utilization of the loader from mid-40 percent to above 60 percent. Over the same two-month period, downtime on the loader fell by 25 percent—and continued to trend downward.
"We were able to increase the productivity of the cell to produce in four and a half days the number of 't'-fittings that we'd previously been producing in seven days," says Hall. Not only was the additional capacity helpful in meeting growth projections, he adds, but the company escaped the need to spend $750,000 on an additional forming press.
And, echoing the Coca-Cola Amatil experience, Hall affirms the value of the solution in dealing with sources of downtime that would otherwise go undetected. "With manual systems of recording, you just don't see 60 percent of your downtime—because it's a minute here, and a minute there. Manual systems essentially only 'see' interruptions of five to ten minutes, not a few seconds," he says. "In retrospect, the issues are obvious—but so obvious that you walk right past them on the shop floor, every day."
A question of priorities
With productivity-sapping issues so endemic, where does it make sense to focus a manufacturing intelligence application to exert the greatest leverage?
In one sense, the answer is obvious, says Eddy Azad, president of Parsec Automation. "The place to start is the constraint, or bottleneck. Improving the performance of a non-bottleneck, or something upstream of a bottleneck, doesn't matter," he says. "It's the difference between making an improvement, and making a difference."
According to Dennis Cocco, chief product strategist at Activplant, "Collecting tons of data from all over the plant is very wasteful, because you're collecting information on pieces of the plant that just don't matter. Look at product flow from a systems perspective, identify the constraints, and then target the improvement actions on the most chronic constraints."
Even so, adds Richard Hopkins, melt shop electrical supervisor at Nucor Steel's mill in Berkeley, S.C., manufacturing intelligence can still generate useful savings in energy consumption—even if the resource in question isn't a constraint. A manufacturing intelligence capability built up from combining Wonderware's Industrial SQL Server historian, Industrial Application Server, and InTouch HMI applications, he says, enables more precise control of energy consumption than the mill's PLCs could achieve on their own.
"The operatives are looking at data, and inferring that they need to do something before it gets to the point where a PLC would respond," explains Hopkins. "There's an 'acceptable' tolerance band, and the operator can see a trend forming within the band and take action before the point that a PLC would activate." Typically, he says, the mill's 1000-volt, 150,000-amp direct-current furnaces operate at voltages 5 percent to 10 percent lower than would otherwise be the case—not an insignificant saving, given today's high cost of energy.
Never forget focus
Focusing on the correct key performance indicators is vital if manufacturing intelligence is to deliver. There's both a strategic dimension to get right, and a tactical dimension. From the strategic point of view, explains Claus Abildgren, program manager for production and performance management software solutions at Wonderware, "the challenge is aligning the metrics used on the plant floor with those used to run the business. If they aren't aligned, efforts to improve operational metrics will be suboptimal from the enterprise point of view."
For example, Abildgren continues, "Instead of manufacturing more efficiently a product that you don't make much money on in the first place, you're better off focusing on another product where you make more money—even if the potential efficiency improvement is less."
And at the tactical level, it's also important to choose appropriate measures of efficiency, adds Activplant's Cocco. "Blindly targeting OEE and downtime is very popular—but incorrect," he asserts. "OEE and downtime are useful on stand-alone pieces of equipment, but for linked machines, with buffers and interconnects, you need to look at flow. Not every resource needs to run at 100 percent—attempt it and you'll generate inventory."
Finally, don't forget the human dimension. "A manufacturing intelligence application can tell you what went wrong, and when—but it can't necessarily tell you why. At the sensor level, the real reason for the fault might not be obvious," warns Sunil Singh, chief executive of Informance. Most customers, he says, come to recognize that operator buy-in will be required at the solutions stage, so it's best to involve them in problem definition.
"Involving the operator shows that you are serious about your continuous-improvement program, and that you value the positive role that operators may play in its success," adds Parsec's Azad. "Many manufacturers have realized significant percentage-point improvements for this very reason."
Perfecting the intelligent enterprise
07/01/2006
