Polyethylene likes the cold, which is part of the reason NOVA Chemicals set up two resin plants in Joffre, Alberta, Canada. But recently, product output was not reaching its full potential in one of those plants, regardless of how low temperatures dipped. Given that market pressure, a global economy, and new product demands mandate that production processes consistently run at maximum capacity, NOVA Chemicals had a major challenge on its hands.
Conventional wisdom suggested that to solve such a problem, the plant manager should collect data and analyze it to find the source of the capacity constraint. And that's exactly what an engineer at the NOVA Chemicals plant did -- downloading data from a historian and the distributed control system on a second-by-second basis for a full week. Gathering all of the information was easy. Making heads or tails of it, however, was taxing.
"They knew there were cooling constraints," notes Alan Schrob, NOVA Chemicals' director of manufacturing excellence. "But it was so onerous to go through the volumes of data generated to figure out what was happening."
Moreover, even if the company could identify the root cause of the bottleneck, it would all be historical information. The need for an innovative production optimization process was clear.
For years, process industries -- and, increasingly, discrete manufacturers -- have been tapping into real-time data to manage production. They use the data to create key performance indicators, build historical models, test procedures, control processes, and survey plant performance. They even conduct offline simulations to train plant personnel. But to date, it's all been information that provides a peek at the past. Now vendors are offering tools that can give manufacturers a glimpse into the future.
"What's happening now is that more predictive models and intelligence has been added on to that data," says Tom Fiske, senior analyst at ARC Advisory Group (Dedham, MA). Those predictive models, coupled with visualization technology, enable not just current analysis but also what-if scenarios, Fiske says.
For example, tying in real-time production data with financial systems allows manufacturers to gain insight into actual business scenarios. When a new order comes in, the model can provide up-front answers to questions like, How do we process it? Do we have the capability to make it? Even, Is it worth making? If it is worth pursuing, the model can address the question of how to adjust the plant so that it is still operating optimally, according to Fiske.
The concept is what ARC has dubbed real-time process optimization (RPO) and training. Today, with Web-based standards helping to integrate enterprise data with plant data, process automation vendors adding analytical tools to control systems, and manufacturing intelligence applications becoming granular enough to find an area of constraint within a production line, companies like NOVA Chemicals are moving some of the corporate decisions directly to the plant floor.
"It allows the plant managers to make capital decisions on where to invest to de-bottleneck based on actual real-time information coming from the plant," Schrob says.
To get to this point, NOVA Chemicals enlisted the help of SAP Americas (Newtown Square, PA), using its Lighthammer manufacturing intelligence technology, and Pavilion Technologies' (Austin, TX) Pavilion8, a visualization tool with an embedded analytical modeling engine. By connecting the Honeywell TDC 3000 DCS system and Aspen Technology Inc.'s InfoPlus.21 data historian, Pavilion8 can predict the performance of the line based on current operating conditions. It can also model theoretical maximum capacity.
"If we know the theoretical max and track it against the current capacity utilization of the plant, we can impact behavioral changes in the control room and get the plant closer to the theoretical predictive number," Schrob says, noting that operators now have a baseline against which they can measure. They can also find the constraint, resolve it, and push the line to peak efficiency. That, he says, is a powerful capability.
More importantly, real-time analytical engines can make an immediate impact on production processes. "Putting predictive modeling in place makes everything proactive versus reactive," says Matt Tormollen, Pavilion's chief marketing officer.
Waste Not, Want Not
Identifying whether a plant has the capacity to make a particular product and finding a point of constraint within existing operations that may be hampering productivity are different problems. Both, however, can benefit from predictive modeling techniques using real-time data.
At NOVA Chemicals, for instance, there was a concern that not producing polyethylene at the maximum capacity was costing the company revenue. At Fonterra Canpac, a maker of cans and can components for a variety of consumer goods including milk powders, nutritional products, and infant formulas, the diversity of the products was matched by the diversity of reporting and management tools on the production lines.
Fonterra Canpac takes orders on demand, meaning the company waits for an order and then makes it. Considering the variety of ingredients needed to make the company's products and the detail of the packaging, each order is a logistics challenge in the fast-moving consumer goods market.
Some lines are operating at 160 units per minute, says Paul Hogben, maintenance services manager at Fonterra Canpac in Hamilton, New Zealand. It can be downright chaotic.
"We wanted to accurately measure uptime, job changeover, and machine efficiency," Hogben says. The company implemented Pavilion as a common production monitoring system. "There are many ad hoc methods of reporting, as each department has its own system, some more sophisticated than others. But there was a significant opportunity in terms of reduction of waste," Hogben says.
Fonterra configured the Pavilion system, which was pulling data from the manufacturing resource planning (MRP) application, to include a trend line, a graphical display that shows on a minute-by-minute basis how the line was running against a production target.
"Operators can see the line trending and see if they are behind or above the line," Hogben notes. "They are more in tune with the operation," he says.
Indeed, ARC's Fiske notes that the technology available today is creating what he calls "on-demand knowledge." Predictive and analytical applications that tap into real-time data can not only help operators make decisions about the production line, they can also alleviate some of the loss of knowledge that is occurring as a result of the aging workforce.
Fiske notes that "in this real-time environment, decisions are being pushed down to the closest point in the operation." Sometimes the greatest impact is made right on the plant floor. "Are they making the right decisions? Do you find out a month later or do you want to know right now?" he asks. "The way to [find out now] is with these types of tools."
The Real Deal
Many of the next-generation process control systems have predictive modeling capabilities built in. For example, the Honeywell Process Solutions Experion PKS system includes abnormal-situation management indicators that target specific factors that may cause disturbances in a line. It also includes a feature called Profit Loop, a patented algorithm that can predict the effect of past, present, and future control moves. Similarly, Invensys's Avantis and SimSci-Esscor business units add real-time condition monitoring and simulation for process optimization, respectively, to the company's next-generation enterprise control system called InFusion. And Emerson Process Management's PlantWeb includes predictive intelligence technology that can sense abnormalities in field devices.
But there is also a new generation of innovative, general-purpose bottleneck-busters that has evolved out of manufacturing intelligence applications. Companies like Activplant (London, ON), Aspen Technology (Cambridge, MA), Acumence Inc. (Chicago), Informance, Inc. (Northbrook, IL), SlipStream Software (Atlanta), and Umetrics Inc. (Kinnelon, NJ) are storming onto the scene with software that adds real-time analytics or simulation to uncover the cause of a constraint on a manufacturing line.
They all tackle the problem slightly differently, but the goal of making adjustments in real time is common to them all. "If you fix it immediately it only costs $100; if you fix it after the fact it costs $1,000; and if you fix it once the product has gone out the door to the customer it costs $10,000," says Gary Hopkins, president and CEO of SlipStream. The company says its Real-Time PM product, released last month, can pinpoint what caused a slowdown.
Informance's Manufacturing Strategist, to be released later this year, aggregates data collection with a real-time transaction layer that can be synchronized with MES or ERP applications, an analytical layer that applies intelligent drill-downs and diagnostics, and a decision support layer to apply what-if scenarios.
"You can start with a high-level business strategy and filter it down to what it means at the plant level," says John Oskin, Informance's executive vice president and founder. "The innovative element to this is tying in the corporate business strategy with the plant business strategy."
Jeff Herrell, director of continuous improvement at TRW Automotive's Occupant Safety Systems (OSS) division, describes his company's use of Activplant's Throughput Analyzer as "an octopus reaching out with its tentacles to touch every PLC and measure the bottleneck."
TRW encountered a situation in which a laser welding cell with an optimal capacity of 1,000 pieces per hour was netting about 650 pieces. The company had worked on the machine for over a year, addressing all of the electrical, mechanical, and downtime issues, but it made no significant difference. The problem, Herrell realized, was that they were concentrating on the machine's downtime instead of characterizing the loss. He shifted the focus to where the other 350 pieces were going.
By applying the Activplant technology, Herrell says, "We found out that half of the 350 was [due to] a legitimate breakdown item, like a sensor failed or a robot messed up. The other half was intermittent block and starve, which translates into the operator not loading and unloading at the rate that is needed. And the magnitude of that was never understood or appreciated," he says. "We couldn't see it, but it was eating away at us a little bit at a time."
For companies like NOVA Chemicals, the technology is a must-have. "In North America there is no new capacity" for chemicals production, Schrob says, "so one of the ways we can better compete is [to] get more innovative."