Prior to the early 1950s, information transfer on the factory floor meant
one thing: paper. Nothing could be built without paper instructions and paper
blueprints--all interpreted, more often than not, by a silver-haired machinist.
Based upon vacuum-tube logic, the computer of that day was much too expensive and too difficult to operate for direct application to the factory's problems. The solution: numerical control. In 1952, the Servomechanism Laboratory of the Massachusetts Institute of Technology demonstrated a milling machine with three axes controlled by digital technology. And at the 1955 Machine Tool Show in Chicago, companies introduced a variety of commercial NC systems.
Because of NC, information could move about a factory on paper tapes, which workers delivered directly to numerically controlled machines.
At the 1970 Machine Tool Show, a General Electric Co. mainframe computer at Chicago's Navy Pier demonstrated digital NC, which eliminatedthe need for paper tapes. And bythe mid-seventies, Computer Numerical Control (CNC) systems had moved onto the factory floor to stay.
Please note the time frame. Engineers at MIT developed the first NC controllers in the early fifties; machine-tool companies put such systems on the market by the mid-fifties. Yet nearly 20 years passed before NC and CNC became widespread. "In the 1950 time frame," says Charles Carter, vice president, technology, at the Association for Manufacturing Technology (AMT), Washington, DC, "people believed that within ten years, 90% of all machine tools would be DNC. It never happened."
Factory of the future. Tomorrow's factory will depend upon information transfer. In fact, some engineers believe that information transfer will completely change the way manufacturing companies operate.
T. Chris Miyazaki, manager, Mechatronics Div. of THK America Inc., Schaumburg, IL, sees centralization of computing power in the future. "The trend I see is either a hard-wire line or a fiber-optic link that starts at the host, goes around the plant, and goes back into the host," says Miyazaki. "Currently, that's the state of the art. The machine systems designer--in terms of software--and the control platform designers are going to be looking at situations where a central processor, a host, can control every single event and can also monitor every single event that happens on the line. I'm including human inputs to interfaces located throughout the line." In this approach, the need for universally acceptable protocols becomes critical (see "Take the right bus," Design News, 2/19/96, p. 59).
The vision of a powerful central computer gathering information and running the factory floor flies somewhat against the present trend to distributed intelligence. Why not use smarter PLCs or other local intelligence? After all, older machines, and machines now coming on line, often are controlled by PLCs. "You could have multiple CPUs and thousands of I/O points," says Miyazaki. "And the control platform will be the scanning type of PLC using ladder logic, which becomes slower and slower as you load it up with I/Os. The throughput of the machine and the events occurring in the machine, either synchronously or asynchronously to each other, must be controlled differently for the machine of the future."
He sees software, rather than hardware, as the key to organizing tomorrow's factories. "The processor's already there," he points out. "The software needs to be written for the specific industrial application."
Smart machines.The other side of the discussion about tomorrow's factories looks to distributed intelligence. David Johnson of Rockwell Automation's Allen-Bradley Div. in Mayfield Heights, OH, sees customers demanding open networks from suppliers over the next five years. "We have a three-layer model," says Johnson. "DeviceNet is our open device network and generally connects low-level components such as sensors and actuators. That's complemented by a higher-speed network for more complex devices called ControlNet. Then the third layer of the network is Ethernet, which we refer to as the information layer."
Near-term, Johnson sees copper, whether twisted pair or coax, as the means of moving information through a factory. But in certain critical applications, fiber optic is already valuable. "Wireless is an interesting area we are already researching," Johnson remarks. Cultural questions arise in the traditional marketplace, he says, when a system goes wireless. And those questions may prove more challenging for wireless than technical issues.
When it comes to the issue of distributed intelligence versus centralized computing, Johnson takes the side of distributed intelligence. "I expect distributed control to continue beyond the cell lower and lower into the control hierarchy." Will this lead to the lights-out factory? Probably not. "There's a better understanding today of the human resource as a productive asset in manufacturing," says Johnson.
And 10 to 30 years from now: "You'll continue to have large, highly automated production facilities that rely on human beings, not to run the machines, but to improve the process. And you'll have more and more 'boutique' manufacturing in small lots." Collections of controllers on open networks will make this type of manufacturing possible, he believes. Companies of the future may also use intranet and Internet technology to help run and be informed on manufacturing processes.
Cincinnati Milacron, Cincinnati, OH, now passes digital information to manufacturing, purchasing, and support groups to review designs, eliminating a great deal of paper. After approval, digital data release takes place to process planning and NC programming people, explains Ray Littleman, manufacturing engineering manager. Next a DNC network takes the information and stores it in a network of NC programs that cell controllers and stand-alone machine controls can call down to the factory floor.
How will the shop floor change? Doug Carter, marketing services coordinator at Cincinnati Milacron, points out that the company still has manual machines on the floor. "But over the last five years, the number of NC machines has greatly increased," Carter says. "There's a dramatic change. Now CNC is much more affordable." And Littleman agrees: "We're just starting to leverage CNC with these other systems. In a lot of factories, this change to CNC has happened over the last five years." (It's worth noting, once again, that numerical control is now 44 years old.)
At Giddings & Lewis, Fond du Lac, WI, Bob Elliott, manager of sales and marketing for automation controls, sees movement of information--not data--via in-tranet and Internet as a major force in tomorrow's factory. Intelligent devices on the shop floor can now directly communicate with other parts of the plant. Actual usage of devices can be monitored, and predictive maintenance can prevent shutdowns. "We've just scratched the surface as far as being able to use and apply intelligent devices, and communicate that information to people who can apply it," says Elliott. "With the intranet, everyone in the company should be able to have access to that information."
Intelligence will be distributed across the machines on the factory floor, according to Elliott and Product Manager Bob Kollmeyer. Communications on the floor will employ a mix of wireless, fiber-optic, and copper techniques. "We see everything from the fiber-optic FDDI links with branches to machines to wireless spread-spectrum systems," remarks Kollmeyer. And at different times, tomorrow's factory will use both RF and infrared wireless to achieve maximum flexibility.
Open systems are important to customers, according to Kollmeyer and Elliott. "But there are degrees of openness," says Kollmeyer. "Just about everyone you would query would bring a different set of criteria to the table." Open systems now play a role in all the projects Giddings & Lewis gets involved with. Openness has significance from communication through connectivity to application programming. To deal with the last, G&L has used the IEC 1131-3 language set for automation controls since 1990. But, Kollmeyer points out, even within that set, degrees of openness exist. "It's out there in the future somewhere," he observes.
Other views. How do large users of automation equipment in traditional industries see automation and the factory of the future? Chuck Jones, V.P. for technology planning at Dana Corp., Ottawa Lake, MI, be-lieves in new technology, to a point. "At one time the factory of the future meant throw out all the old equipment, replace it with new. It's automated, it's CNC, it's lights-out. Then reality set in," he remarks. "Throw out all the old equipment, and you can't afford the depreciation expense and expect to sell products."
Engineers at Dana have learned to apply PCs to the shop floor, rather than install totally new automated equipment. An existing plant, he points out, can install a local-area network, put PCs at operators' stations, and eliminate much paperwork while improving throughput.
Gradually, the company can then replace older machines with CNC equipment, and the computer becomes a device understood by operators, rather than something frightening. Ten years from now, Dana intends to be capable of going "from art to part" for its products. As for the issue of distributed intelligence and centralized intelligence, Jones sees a role for both. "Our company has certainly subscribed to the idea of distributed intelligence. But we're finding that the nodes on that network are becoming more and more powerful, to the point where you could almost say each one becomes a mainframe computer in its power. But they're still networked." In situations where a huge amount of data must be crunched, Jones believes a mainframe will still run things.
In speaking of the factory of the future, Bob Null, V.P. of manufacturing at Baldor Electric, Fort Smith, AR, remarks: "You can take the time estimates on the factory of the future most of the technical wizards come up with and double them. Neither the technical issues nor the people issues will come nearly as fast as they expect." He warns that interconnectivity issues present more serious barriers to implementing automation than many observers realize.
"People must be able to comfortably interact with an increasingly electronic information system on the factory floor," Null says. Baldor has made a pioneering effort in electronic communication by installing a powerful information system in its newest factory. At each workstation in the plant, a PC calls all information the operator needs--such as drawings--to do his work. Training refreshers and safety warnings are included.
A believer in distributed intelligence in the factory, Null says the variables in the manufacturing process are more complex than most people on the software side appreciate. As an example, he cites an attempt by Baldor to use commercial software to generate an optimized tool cut program for a specific lathe and download it to the machine, without doing programming on the floor. "I've had several engineers working for six or eight months getting four or five lathes in one plant and three machining centers in another plant onto that system," says Null. "It looks like it ought to be straightforward. But it's not."
What will the factory look like? Null expects tremendous changes in just the next 10 years. "We will see more and more paperless operations; we'll see more and more small cells and work groups, with more automation in them. You'll see automation that eliminates or minimizes setups and changeovers."
Fifty years from now, Null expects a more sterile environment in manufacturing facilities, with fewer people, less noise, fewer chips on floors, and more exotic materials being handled. People will run smaller groups of machines, monitoring operations and handling special products. Also, information technology may make it possible to reduce the size of manufacturing plants and put them near markets. "Now they're all in one big site, for economies of scale and ease of management," says Null. Information transfer can make both of those factors less important, and allow small-scale manufacturing to replace giant factories.
This concept has other champions. "We believe that someday the factory will be as big as a six-car garage," says Richard Morley, designer of the first programmable controller. Internet-connected throughout the world, the future factory will be able to make products like watches, coffee cups, and tape recorders within a distributed system. "They're calling that approach teleoperations and telepresence," says Morley. "We'll move the factory to the point of consumption."
Rapid manufacturing. Information flow lies at the heart of another approach to the factory of the future. Rapid prototyping already provides "art-to-part" production of prototypes. Research now underway at Argonne National Laboratory in Argonne, IL, seeks to use rapid-prototyping technology to produce functional parts for turbine engines.
Team leader William A. Ellingson, senior engineer, Instrumentation and Nondestructive Evaluation Section, at Argonne's Energy Technology Division, sees rapid prototyping growing into rapid manufacturing. "As rapid prototyping evolves, you can do limited production runs," he asserts.
This year, Ellingson and his colleagues are working on ceramic components for stationary gas turbines. Using a fused deposition machine made by Stratasys, they intend to produce fully functional ceramic blades and nozzles made from silicon nitride. Orthopedic implants also may be made using the equipment now in place and operating at Argonne.
To fine-tune their designs, the Argonne team employs computer tomography (CT) scans of components under load. The high-resolution CT scan reveals areas of high stress within the part. Engineers can then geometrically adjust an FEA mesh to enhance the component's ability to handle stress. This process amounts to reverse engineering based on the part's internal reaction to loads.
Ellingson and his colleagues are also seeking to make dense ceramics from materials that use polymeric binders. In this work, after a rapid-prototyping machine creates a part, it undergoes five to seven days of binder burnout. Parts emerge approximately 90% dense, according to Ellingson, without the need for hot isostatic pressing. Technical issues now being addressed by the research group include part isotropy and dimensional stability.
As he looks ahead, AMT's Charles Carter sees possibilities and problems. "People dream of moving from a drawing to automatically having a part program and a part made, without human interference," he remarks. "Fifty years from now, engineers will complete a solid model drawing, then push a button. Information goes into a machine and the part's made. Never any human interference. And for that to happen 50 years from now there has to be some step-input in our predictive ability."
Today tremendous technical capabilities exist. They offer immense promise, and could change the factory floor as much as did numerical control.
But it's worth pointing out that NC, CNC, and solid-state electronics were developed during the Cold War, spurred on by the Department of Defense and NASA. Yet despite that pressure--and generous funding--NC and CNC took hold rather deliberately. Will the forces of competition suffice to spur the adoption of new, information-based automation systems? Or will these technologies remain on the horizon for many years, like the peaks of mountains seen from far away?
PREDICTION FROM THE PAST
In the 1950 time frame, people believed that within ten years, 90% of all machine tools would be DNC.
--Charles Carter, Association forManufacturing Technology
I've had several engineers working for six or eight months getting four or five lathes in one plant and three machining centers in another plant onto that system. It looks like it ought to be straightforward. But it's not.
--Bob Null, Baldor
THE ENGINEER'S ROLE
Fifty years from now, engineers will complete a solid model drawing and push a button. Information goes into a machine and the part's made. Never any human interference.
--Charles Carter, Association forManufacturing Technology
COMPONENTS TO WATCH
Universal communication protocols