(FORTUNE Magazine) – JOHN COSTANZA

COMPANY John Costanza Institute of Technology FIELD Production engineering ACHIEVEMENT Invented demand-flow manufacturing HOBBIES Driving a dragster, flying

Production's New Choreographer

Among the men and women who have helped U.S. manufacturing return to world preeminence, John Costanza, 51, stands out like a lighthouse. He has been on this crusade for 16 years, successfully executing his dream from the John Costanza Institute of Technology (JCIT), headquartered in a big, airy, glass, brick, and concrete building in the Denver suburb of Englewood. Boasting 240 employees in Colorado and at its branch offices in San Jose and in Nice, France, JCIT has trained more than 70,000 representatives of 3,000 companies, most of them American and European. Costanza's 1990 book, The Quantum Leap in Speed to Market, has sold 250,000 copies.

Costanza is the father of demand-flow manufacturing, a concept that stands conventional repetitive manufacturing on its ear. All those famous technologies Americans borrowed from the Japanese--the Toyota production system, just-in-time parts delivery, and similar approaches--are designed for repetitive production of the same item.

The word "flow" was used earlier by Japanese manufacturers but in a different context. Their approach, says Costanza, has been: "Flow it through, fill up the warehouse with the wrong stuff, and go broke. One of their companies recently filled up its warehouses with a VCR nobody wants." Today the world demands many products, particularly computer equipment, made at lightning speed with frequent variations to accommodate and stimulate shifting demand. This calls for a flexibility that repetitive manufacturing lacks.

Demand flow does away with schedules and forecasts, inventories of finished goods, and warehouses in which to store them. For the fabled just-in-time concept, it substitutes raw-in-progress (RIP), in which a cushion of varied materials is kept on hand to meet changing customer demands. Finished goods move directly from assembly lines into waiting trucks. The results are astonishing: American Standard, for example, saved $500 million in working capital during the year it started using demand-flow technology (DFT). Another JCIT client, General Electric, has installed DFT in all its divisions; the technology has earned high praise from CEO Jack Welch for boosting productivity. By shrinking clients' stocks of finished goods, DFT has also enabled them to double and triple their inventory turn rates. This has helped them move toward the goal of zero working capital--the state of ultra-efficiency, proved possible by DFT user Dell Computer, in which inventories are so lean and production so prompt that a company can collect payments from its customers even before the raw materials are on the books.

John Costanza's story is the story of America. An Italian laborer immigrates to toil in the coal mines in Trinidad, Colo., and later in steel mills in the town of Pueblo in the same state. Two generations later his brilliant grandson flies around the world in his own Falcon 2000 jet and owns a $70,000 Federal Mogul dragster, which a sextuple heart bypass operation last year hasn't stopped him from driving. His 18-year-old daughter, Melissa, has her own dragster, in which she has won elimination rounds at the famed Indianapolis Speedway.

Before these luxuries came a decidedly modest childhood in Pueblo, where Costanza's father was a postal worker and his mother a housewife. Growing up, Costanza was an avid reader of science fiction. He was also fascinated by how mechanical devices worked, taking apart and putting together radio-controlled cars and other toys. He helped his father in his small workshop, where they repaired neighbors' motorcycles and other machines.

After graduating from Colorado Southern University with a degree in applied mathematics and physics, Costanza worked at Johnson & Johnson and Hewlett-Packard. Moving from Colorado to Texas and then to California, he gained experience not only in manufacturing but also in product design, materials handling, marketing, and software development. From Texas he ran an HP plant located across the border in Mexico.

In the early 1980s, Costanza's expertise caught the eye of HP co-founder David Packard, who was looking for an engineer to help figure out how U.S. industry could survive the Japanese manufacturing assault. Concerned government officials had created the President's Council on Productivity. To head it, President Reagan had chosen Packard, who turned to Costanza. That he and Packard hailed from Pueblo allowed them to reminisce about their childhood days.

After extensive firsthand studies of Japan's repetitive manufacturing and Germany's highly departmentalized production systems, Costanza decided there must be a better way. Sensing a trend toward global manufacturing that needed to respond quickly to customer demands, he decided to engineer a new manufacturing system. Packard encouraged Costanza to put his ideas into practice.

"I sat down and decided that when you look at manufacturing, it's nothing more than a relationship of work and volume," Costanza says. "The work is performed by people and machines, and the volume is whatever you think your capacity is going to be. I came up with a calculation for that foundation and called it operational cycle time."

He explains: "If, for example, you want to build an automobile and you want to build eight per day, and the plant works eight hours a day; that's one car every hour. If it takes 32 hours to build a car, you just break it into one-hour pieces, staff each step with a person and a machine, and have one piece come out every hour. If you use that as a premise, you're getting to a mathematical foundation. But what if you can't hit that one-hour target? How do you balance this? The auto has four tires. So how do you balance four tires in an hour when you only make one frame in an hour? I came up with a series of formulas to balance it." To avoid sets of tires piling up, the solution was to shift workers who completed their jobs ahead of time to other tasks in the plant.

"Now I could get anything to flow, and I could balance anything mathematically," Costanza says. "The next question was, How do you get material to this process? We don't know what we're going to make, and it's going to be driven by demand, so we have to figure out how to replenish the material that has been used up. I put a math foundation under that as well. And a lot of people thought I was nuts. Others--including Packard--said, 'You may have something, and we should try it.' "

Costanza's demand-flow system, which also substituted semicircular work cells for some linear production lines, performed well when a couple of HP divisions tried it. But it ran into resistance elsewhere in the company. HP was then trying to sell factories on the large-scale use of computers. DFT sharply cuts their use, since it involves no transactions--no work orders and schedules, for example, to make up and follow. In a DFT plant, the role of computers--mostly PCs--is limited to such tasks as calculating total production cycle time and tracking actual vs. planned worker efficiency.

David Packard loved the logic of DFT, but even he couldn't change HP's computer-oriented culture overnight. Yet DFT became a permanent feature at some parts of the company. Packard also helped Costanza start his institute by investing $10,000 in it. Costanza and his wife, Linda, put in another $10,000, and with the help of numerous relatives, started JCIT in the basement of their Denver home. An advocate of the hands-on approach to learning, Costanza developed a four-day course to be taken in a lab factory that he designed. In JCIT's early years, Costanza and his wife would load the portable factory into a big moving van they had bought for the purpose and drive down, say, to Texas Instruments in Dallas to teach engineers there the new techniques.

JCIT's stationary classrooms now have model factories too. Students first assemble products such as motors in the conventional repetitive assembly style. Then they restructure the assembly lines into semicircular cells and make the same products the demand-flow way. Even skeptical old-timers generally come out of the four-day course believing in DFT.

Costanza's ideas have caught on especially fast among advanced manufacturers such as contract electronics houses, which must quickly accommodate product changes. It takes time to put in a DFT system, and even longer to change a corporate culture, says Costanza. The changes should be coming faster now that JCIT, which formerly made its DFT software available only to clients, is selling it to all comers.

JAMES BRYAN

COMPANY Bryan Associates FIELD Precision manufacturing ACHIEVEMENT Dramatically raised the standards for machine tools HOBBY Sailing

He Wrote The Book on Precision

When James B. Bryan began his pioneering work in precision machining in the mid-1950s, most tool users checked the state of their equipment by making parts and inspecting them for defects. Then they tried to correct the imperfections in the machines, but the procedure was based in large part on guesswork and seldom produced the desired results. Thanks to Bryan's contributions to metrology, the science of measuring the performance of machine tools, the metalworking industry now achieves tolerances that seemed pure fantasy in those days--as little as a millionth of an inch. Growing up in Alameda, Calif., across the bay from San Francisco, the only child of a high school superintendent and his teacher wife, Bryan learned in high school how to make metal objects on a lathe. In 1944, when he was 17, he quit school to join the merchant marine and shipped out as an able-bodied seaman aboard a freighter headed for the Pacific war zone. Six months later he was an assistant in the engine room. After the war he became an officer with the romantic-sounding rank of Second Assistant Engineer Steam Vessels, Any Horsepower, Any Ocean.

As a marine engineer, Bryan was required to be a jack-of-all-trades. He repaired pumps, plumbing, and boilers, and made spare parts in the ship's machine shop. As he did so, he began to question the accuracy tests then prescribed by machine-tool builders, concluding, at age 19, that they made little sense. The recipe for checking spindle accuracy in a lathe, for instance, called for attaching a probe to the cutting-tool holder, which is stationary, and recording the movement of the spindle nose surface as the spindle rotates. But Bryan wondered if other problems, such as defects in the bearings on which the spindle turns, could cause deviations.

In 1951, after going to sea each summer on a merchant ship, Bryan graduated from the University of California at Berkeley with a degree in industrial engineering. He did a stint as a manu facturing engineer at Westinghouse and in 1955 joined Lawrence Livermore National Laboratory in Livermore, Calif. It was there, in the course of a distinguished 31-year career, that Bryan made his wide-ranging contributions to metrology and precision machining.

At the lab, Bryan and his colleagues faced the daunting challenge of making components of unprecedented precision for nuclear weapons. He not only talked to manufacturers around the world but also delved into the old books on metrology. He bemoans today's trend of dismissing almost everything that's old as being of little use. Many of the principles of precision manufacturing, he says, were established by forgotten geniuses who worked centuries ago.

Bryan based his ideas in part on the work of one of his mentors, John Loxham, an engineering professor at Cranfield University in England. It was Loxham, also an entrepreneur, a businessman, and an amateur astronomer, who first suggested that machine tools operate with the same regularity that he observed in the motion of stars and planets. Bryan's approach to metrology expanded on that idea.

His basic idea is that machine tools obey cause-and-effect relationships that can be understood and controlled. According to Bryan, the traditional approach of statistically tracking a machine's performance wrongly assumes that what he calls "non-repeatability," or inconsistent accuracy, is inevitable. There's nothing random or probabilistic about machines' behavior; everything happens for a reason, and random results are the consequence of random procedures.

A better alternative, Bryan argued, is online monitoring and control of variables such as temperature, air pressure, vibration, control-system voltages, and tool conditions. This "deterministic" approach to manufacturing, to use his term, is common sense. Although the major portion of Bryan's work at Livermore remains highly classified, he still managed during that period to publish more than 50 papers with his colleagues and develop an international reputation in metrology.

Bryan began by upgrading standard machine tools such as so-called tracer lathes. These were used to machine contoured parts before computer numerically controlled (CNC) machines replaced them. Bryan was able to boost the lathes' performance to achieve contour accuracies of two microinches, or two-millionths of an inch. By comparison, a human hair has a diameter of about 3,000 microinches.

Bryan went on to develop numerically controlled diamond turning machines, which use diamond cutting tools to make parts to unheard-of accuracies. He showed that, as he puts it, "by giving a machine a proper home," even a 50-year-old lathe's precision could be improved from 2,000 microinches to one microinch. "He did this through a combination of detailed mechanical understanding and temperature control" of the machine tools, says Robert J. Hocken, director of the center for precision metrology at the University of North Carolina at Charlotte, who has known Bryan for more than 25 years. Says Hocken of Bryan: "He has the unique ability to examine a machine and a process, realize the variables, and control the variables to produce parts of the highest quality."

Bryan's contributions to the field have been so numerous that his most significant one, according to his former Livermore colleague Ken Blaedel, "varies depending whom you ask." Blaedel believes that "his most significant contributions apply at the highest levels of dimensional precision. In this domain, Jim's determinism shows us how to wring more precision out of a manufacturing process."

In the late l980s, Bryan's work at Livermore culminated when his group designed, built, and ran the largest diamond turning machine in the world--a gigantic $8 million lathe that took eight years to put together. The machine weighs 120 tons, is cooled by an oil shower with a flow rate of 400 gallons a minute, and can work on parts eight feet in diameter, ten times larger than what typical machines of this type handle. Equipped with a gem diamond cutting tool, the machine turns out complex parts for the mammoth laser that Livermore is building in its latest attempt to harness nuclear fusion power. The advances Bryan pioneered in diamond turning have spilled over into the commercial world, where diamond machines are now widely used to make optical components such as lenses for reading glasses, hard disks for computers, supermarket scanners, laser-printer drums, and other products.

Some techniques developed by Bryan became instant national and international standards. One example is the American benchmark for specifying and testing axes of spindle rotation. Bryan has also done important work on using oscilloscopes--the engineer's counterpart of EKG machines--to measure motion errors in production machines. This has led to the development of PC-based instrumentation.

Bryan's work on the effects on part accuracy of internal overheating of machine tools has become the basis for international standards. After determining that such "thermal effects" are the largest single source of error in close-tolerance machining, he proposed that tool builders enclose each machine in a box that has an air or liquid shower with temperature control sufficient to meet the required tolerances. Precision steppers, the machine tools of the chip-fabrication industry, now use this approach, and machine-tool builders are beginning to accept the idea.

Where Bryan has not been a prime innovator, he has improved on the brainchildren of others. Thus, he added a telescoping arm to the fixed ball bar, an old British invention, turning it into an instrument now used throughout the world to test machine-tool performance quickly. The U.S. Department of Energy received a patent for Bryan's telescoping magnetic ball bar in 1984--and paid the inventor $1 for his trouble. In 1986, Renishaw Inc. of Schaumburg, Ill., took out a license from the Energy Department to build and sell a slightly different version of the device; it has become the leading make.

Bryan lives in a modest apartment in Pleasanton, Calif., and at 73 continues to sail his 30-foot boat in San Francisco Bay. He goes right on consulting at the Livermore lab and with manufacturing companies, astounding his friends with his perpetual enthusiasm. And he's never too busy to explain the world of precision to anyone who wants to know, from a CEO to the youngest machinist.

GENE KIRILA II

COMPANY Pyramid Operating Systems FIELD Production technology ACHIEVEMENT Developed a portable cell for molding composites HOBBIES Boating, hunting, raising beef cattle on his farm

His Brainchild: Instant Mini-Factories

A cerebral 35-year-old college dropout is fanning the flames of an ongoing revolution on the plant floor: coupling digital information with manufacturing processes. Gene Kirila II has dreamed up a novel portable cell in which a computer runs the complex chemistry of making products from composites. Because the cell can be controlled from afar, it's also an example of Internet-assisted manufacturing.

Kirila's cell, expandable from its basic dimensions of 24 by 24 by 18 feet, can be used for the automated production of composites in low to medium volume, up to about 100,000 parts a year. Tooling, or making a mold for the cell, requires an investment of only $25,000, vs. some $1 million to $2 million for an existing technology that is its closest rival in quality of the finished product.

The cell uses a thermosetting technique, one of the two that dominate the making of composites. In thermosetting, liquid chemicals are mixed and allowed to solidify with the aid of liquid catalysts. The other technique is thermoforming, in which plastic pellets are melted and linked together with fiberglass and other reinforcing materials under high heat and pressure. Both those technologies, when used in the conventional manner, involve a lot of labor as well as capital. Unlike Kirila's cell, they are also dangerously polluting and require expensive equipment to control their exhausts.

Kirila, in short, has created the manufacturing counterpart of a McDonald's hamburger outlet. It serves up boats, truck parts, and even railroad cars with quality and precision, at low costs never before attained. A Kirila cell can be dropped into China or Minnesota, or any place that has electricity and compressed air. In three days the customer has a humming little factory.

The cell's one or two workers don't have to be technicians; a computer program monitors and controls the process, and they are guided step by step by graphic displays on computer screens. The system doesn't let the workers proceed unless they have properly executed the previous production step. If a worker encounters a problem he or she can't solve, a Solutions Center, manned around the clock by experts, is available on the Internet for help. Kirila's idea succeeds where somewhat similar Japanese "lights out" manufacturing failed because it relied on robots rather than people. It also had no communications link to knowledgeable engineers able to remedy problems as they occurred.

Kirila grew up in an industrial setting in the Shenonda Valley, which spreads across the Pennsylvania-Ohio border 60 miles northwest of Pittsburgh. His father is co-owner of a heavy-duty construction company; early on, Kirila was exposed to steel mills and other industrial enterprises. But in high school and college, football was Kirila's true passion: Not only did he play offensive lineman, but he also, in working in an industrial-arts class during his high school years, built exercise machines for the football team. While at Youngstown State University, he started a company to build exercise machines, Pyramid Fitness of Sharpsburg, Pa., and dropped out to run it. Ten years later its annual sales reached $44 million.

Frustration drove Kirila to design his cell. Composites, which can be made flexible in one part of a product and rigid in another part, are an attractive material for building exercise machines. But a lack of reliable production methods discouraged Pyramid from using the materials.

It took four years to develop the software that runs the cell, which is called virtual engineered composites, or VEC. It is an outgrowth of Kirila's longtime belief that an operating system, similar to the one that runs computers, can be designed for a manufacturing process.

By experts' accounts, VEC changes the game in low-volume composite manufacturing. Not only does it sharply reduce the cost of making thermoset composite products, but it also makes sturdier and better-looking boat hulls and other products than traditional technologies can. It's speedier too. VEC can make a 17-foot boat hull, for example, in 70 minutes. The older way can take days.

With computer controls throughout the process--supervising as many as 280 different manufacturing operations--Kirila's cell slashes production time. "The formula for success," as Kirila calls it, is relentless adjustment of the capricious and changing chemistry of the liquid components as they solidify. VEC links resin storage, injection pump, mold, and process controls into an integrated system. Temperature, viscosity, and other variables are under constant supervision and control; the data are displayed graphically for operators to see in real time. Automatic on-the-spot verification that each previous step was properly executed accounts for the system's reliability.

Inside the cell is another clever invention, a "floating" mold. In the types of composites manufacturing that VEC competes against, a new set of molds, costing as much as $2 million, has to be made for each product. The halves of the mold, which close like the halves of a walnut shell, are chiseled out of steel or aluminum in a painstaking process that can take up to a year. Molds of that type still make sense for large production volumes. But for smaller runs, the floating mold is a better way to go.

Kirila and Robert McCollum, director of engineering at Pyramid, devised a system in which two tough composite laminated skins, each cut to accommodate the shape of a new product, are attached to a universal metallic mold frame. "Universal" means that the frame is reusable and stays unchanged. The space between each skin and its metallic support is filled with water, and the air is pumped out. Since water is noncompressible in the pressure ranges used in VEC, the skins become a rigid hydraulic system--like two firm waterbeds facing each other across a cavity. The mold halves are closed, and composite material is injected into the cavity.

The floating mold allows better control of the mold surface, resulting in faster production and a better-looking surface on the product. The mold skins cost only $1,000, compared with a minimum of $42,000 for conventional molds. There's another advantage. It takes only ten minutes to change the skins, thus creating a new mold. This introduces a new degree of flexibility to the manufacture of composite parts and products. "Out of that same work cell, I could be making bathtubs in the morning, boats in the afternoon, and utility-truck and trailer parts the next day," says Kirila. "The cell would be perfect for Third World countries where they don't have the volume to make expensive new molds."

What really forced Kirila and McCollum to come up with the floating mold was a multimillion-dollar contract that Pyramid could not have carried out with conventional molds; it didn't have the money to buy them. Kirila and McCollum hit upon the bright idea of substituting laughably cheap water for costly steel and aluminum after long discussions of such alternatives as sand, glass beads, and even Ping-Pong balls. If ever there was a triumph of ingenuity over convention, the floating mold is it. It opens the way to making composites competitive not only with aluminum and other metals but also with plain wood.

In 1995, Kirila built a spacious factory on a 15-acre plot in an industrial park in Greenville, Pa., rimmed by cornfields--a site that testifies that Silicon Valley has no monopoly on ingenuity. Kirila and his 65 employees used their new method to make composite parts for Ingersoll-Rand and Acuma Machine Tools, a Japanese company, and boats for sports enthusiasts. With accurate dimensions achieved on the first try, boats built with VEC have been shown in industry-sponsored tests to ride faster and better.

Kirila's original intention was to spread the VEC gospel and franchise his cells around the world. He and some friends invested $12 million in the project. But when venture capitalists failed to come through with an additional $50 million that Kirila felt he needed, he sold the technology and the company last year to Genmar Holdings in Minneapolis, the second-largest automated builder of boats in the U.S. Pyramid now operates as Genmar's R&D subsidiary.

A cell is already making boats at Genmar, and a $12 million VEC plant containing four cells will be completed this summer. The troubleshooting Solutions Center, located at Pyramid in Pennsylvania, now connects via the Internet with the cell in Minnesota. Genmar has big plans to use VEC to build other things besides boats: transportation equipment, recreational vehicles, doors, window frames, and walls and roofs for small buildings. The company will both license VEC and work with others in joint ventures.

Says Steven J. Kubisen, senior vice president at Genmar, who joined the company last year because he was excited about VEC: "We estimate that VEC is applicable to more than $10 billion of the total $25 billion annual market in composite parts." Not a bad prospect for former offensive lineman Kirila, whose college professors thought that all he could do was play football.

RICHARD MORLEY

COMPANY R. Morley Inc. FIELD Inventor ACHIEVEMENT Invented the programmable logic controller (PLC) and other computer solutions HOBBIES Skiing, motorcycling, working his "fake" farm

A Leading Visionary Of Manufacturing

When he was a young engineer of 35, Richard E. Morley put a lasting imprint on manufacturing by inventing the programmable logic controller (PLC), the still ubiquitous factory computer that's produced by a $6-billion-a-year industry and that runs production lines and machines around the world. Another Morley invention, the parallel inference machine, a kind of super-PLC coupled with highly advanced software that he developed, helps keep Japan's 130 bullet trains on their 165-mph schedules. And in what he considers a preview of manufacturing's future, Morley has used chaos theory to devise money-saving solutions in such complex tasks as painting trucks in a General Motors plant.

Morley's original PLC has been retired to a display case at the Smithsonian Institution. But its inventor, now 67 and operating out of a rebuilt, computer-crammed red barn near Nashua, N.H., has no thought of quitting. Into a busy life that would leave a 30-year-old breathless, Morley has squeezed a range of activities that go far beyond inventing. Investor. Entrepreneur. CEO. Director of companies. Author. Consultant. Lecturer. Motorcyclist. Bulldozer operator. Handyman. Foster father to 25 kids. Gadfly. Jokester. One of the country's foremost experts on manufacturing. And more.

Morley is one of those geniuses who see things from what his wife, Shirley, calls a "Martian" viewpoint. In his own self-deprecating words, he calls it looking at the world through a "twisted mirror." He grew up on a farm in Clinton, Mass., 50 miles from his present farm. At age 6, he was already driving a tractor. In financial difficulties because of the Depression, Morley's father became a machinist and took his family to New York City and later to New Jersey.

Morley was interested from childhood in mechanical things. He hung around factories where his father worked, picking up machining skills. He worked summers as a machinist while in high school and at MIT, from which he dropped out after four years because he was suffering from severe migraines and had to support a growing family.

A friend later remarked that a secret of Morley's success was that "he actually built his stuff." He invented the PLC on New Year's Day in 1968 because he had gotten tired of designing individual devices to direct machine tools in manufacturing plants. Production machines were then controlled by clunky electromechanical relays in refrigerator-high boxes up to 50 feet long, which manufacturers were eager to replace with computerized controls.

Fearing the onset of a hangover from the previous evening's merrymaking, Morley worked feverishly that New Year's morning. He came up with a 30-pound box containing three computer boards and no cooling fans or other moving parts, which was less bulky than the smallest computer then available. Morley designed the PLC to be robust. He and an associate inadvertently dropped an early one in the doorway of a client company, but it hummed along fine when plugged in. Some PLCs run for 20 years without breaking down. With $200,000 from Harlan Anderson, co-founder of Digital Equipment Corp., Morley and his friends in 1969 started a company called Modicon to make the controllers. "We were just techies," he recalls. So when stronger competitors such as GE and Allen Bradley entered the business, he and his friends sold out.

Morley's other important invention, the parallel inference machine (PIM), is a supercomputer that employs up to 16,000 logic chips in parallel for faster and more sophisticated data processing. A PIM operated by Yaskawa Electric, a major Japanese maker of motors and robots, enables Japan Railways to run devilishly complicated bullet-train schedule simulations and helps train dispatchers cope with emergencies.

Yaskawa is a licensee of Flavors Technology, a small company Morley started in Manchester, N.H., to supply PIM computers and their special software. The software has been successfully used by a number of companies. One of them, International Plastics, uses a PIM to improve the complex scheduling of pellet making at a big plant in Germany. In all, Morley holds 22 patents, including some for ideas that have been used in the design of handheld computer terminals and floppy disks.

PIM software is based on a concept called complex adaptive systems (CAS), a generic term that encompasses new approaches to coping with complexity. An outgrowth of so-called chaos theory, CAS was developed and refined at the Santa Fe Institute, a think tank where Morley rubs shoulders with Nobel laureates. For inspiration, practitioners of CAS look at nature. The activity at an anthill or the movement of a flock of birds, they believe, is so complex and sophisticated that no central intelligence could ever direct it. But by following relatively simple rules, individual ants create a viable community. The flock of birds in flight, too, is governed by such simple principles as staying within three feet of your neighbor, slowing down in turns, and so on.

Morley devised a CAS-based system for painting trucks at a GM plant in Fort Wayne, the first factory where the concept was put to use. In effect, he replaced the software equivalent of a Soviet-style planning bureaucracy with a "free enterprise" system in which individual machines behaving in their individual best interests paradoxically benefit the whole operation.

Instead of the normal "central" scheduling based on the assumption that trucks will move into paint booths in a prearranged order--which they seldom manage to do--Flavors Technology created a software regime in which the plant's ten paint booths "bid" for jobs. If the next arriving truck was to be painted white, for instance, the booth just finishing a white job put in the highest bid because it could easily go on painting white. The previous system, in which a booth sometimes had to switch colors to accommodate the next truck that happened to appear, caused delays and wasted paint, then running as high as $160 a gallon.

In Morley's setup, each booth became like an ant interacting with others to create an anthill. "We were not programming a paint shop," says Morley. "We set up behavior in the booths that automatically turned it into an efficient system." Morley's approach worked faultlessly at Fort Wayne for six years, saving $1 million a year in paint alone. Yet GM replaced it with a more conventional system--an inferior one, in Morley's view--when the time came to change from hydraulic painting robots to electric ones. Morley's disciple Gregg Ekberg, who had been GM's project manager in the CAS-based paint shop, fought in vain to have the Morley system adopted throughout the corporation. He left to found a company that provides CAS solutions (see next profile).

Many experts are convinced that CAS will be the factory-control technology of the new millennium. To promote the concept, Morley, through his consulting firm R. Morley Inc., sponsors an annual symposium in Santa Fe on how to deal with chaos in manufacturing. In his recent book, The Technology Machine, Morley predicts that manufacturing's future will include adaptive software, but with humans still in control.

Morley and his wife live in a 17th-century farmhouse they rebuilt with their own hands, embellishing the grounds with big stone walls. On the 200-acre property are tractors, bulldozers for clearing snow, and a backhoe that can be remotely operated with a handheld PLC. The ground floor of the barn is full of tools, and Morley does all the electrical and mechanical work. "The ability to use my hands is as important to me as using my mind," he says. (Appropriately, Morley's e-mail address is morley @barn.org. His Website, www.barn.org, displays a picture of the big barn.)

Visitors to the farm can expect surprises, such as the answers Morley sometimes gives. If you ask what trees grow on his property, he replies, "I really don't know. I'm a silicon life guy." For members of the National Center for Manufacturing Planning, an industry consortium that meets at the barn once a year, Morley once flew a four-foot-long remotely controlled blimp. Slipping in manure from the cows and horses that Morley and his wife then kept, the startled Gucci-shod attendees kept their eyes glued to the silvery model airship. It was part of a project Morley is still working on, in which blimps would fly in formation, behaving like a flock of birds, in a test of CAS ideas.

Morley's varied activities have not made him rich. He did get some stock and a 30-year consulting contract from the buyer of Modicon, now part of French-owned Schneider Electric. But it was hardly a major windfall. "I guess I think of myself as an artist, not a manipulator of money," he says. "Money mostly annoys me. It's an irritant--I don't even like writing checks." Says Shirley: "He'll never be rich. He feels there's always someone who needs the money more than he does."

Over the years, Morley and his wife have taken in 25 homeless children, with as many as seven living with them at one time. Morley drives a 1995 Chevy Impala SS and a 1980 Harley-Davidson Sturgis motorcycle. "My passion is work," he says. "I don't have any ego. I want to move on, not stop and take credit for anything."

GREGG EKBERG

COMPANY Highline Controls FIELD Simplifying factory controls ACHIEVEMENT Improved factory efficiency with novel software HOBBIES Soccer, freestyle skiing, golf

He Prunes Away Factory Complexity

A bespectacled visionary at 37 and a student for whom Richard Morley (see previous profile) served as mentor, Gregg O. Ekberg operates where few have dared to tread. Four years ago he founded a company to apply the principles of complex adaptive systems (CAS), a smorgasbord of new approaches for simplifying the operation of factories. Ekberg's Highline Controls of Fraser, Mich., a Detroit suburb, has successfully solved sticky real-life problems on the manufacturing floor, in the process boosting the productivity and product quality of small and medium-sized manufacturers.

In the factory, Ekberg's CAS approach is to model the behavior of system components and then improve the operation of each component, one by one. Out of this comes software to control a machine, a process, or a whole factory floor. Highline's software is much simpler than conventional software because it doesn't try to boss all the processes in one all-embracing system. It accomplishes in a few lines of code what would otherwise take hundreds.

When Hirsh Industries, a Des Moines maker of metal filing cabinets, ran into a serious painting problem, Highline Controls came to the rescue. The plant's two paint booths were having trouble painting three sides, the top, and some drawers of the cabinets. In a $100-million-a-year operation turning out 260 cabinets an hour, 20% of the finished cabinets were having to be repainted, at major cost.

One problem was the rigidity of the paint operation's overly elaborate computer and software system. As many as 80 cabinets could be lined up in a queue awaiting paint. If a key switch failed to trigger the paint guns when a cabinet went by or switched on twice instead of once, the whole queue could be thrown off and the remaining cabinets would not be painted properly. Highline substituted a simpler setup in which each approaching cabinet separately "informed" the paint guns where it was, and each gun knew when to start spraying. This made it impossible for the mispainting of one cabinet to affect all the others behind it.

"If we stop writing software that tries to be an all-knowing master controller, we can avoid a lot of the cost and downtime," Ekberg says. In Hirsh's painting line, "we eliminated the master controller and focused, say, on what the guns needed to know, which is when to turn on and off."

Highline was also able to eliminate needless safety checks. Hirsh's existing software, Ekberg and his colleagues found, prescribed redundant steps such as checking the air pressure before turning on the paint gun, making certain there was no emergency shutdown in effect, and verifying that high-voltage electricity was on.

In writing new software, Ekberg did away with these steps. There was no need to check for sufficient air pressure, because without it the paint gun doesn't turn on anyway. There was no point in checking for an emergency stop, because if one was in effect there would be no air pressure. And because paint guns won't work without high voltage, that check could be eliminated too.

Some precautions were still needed in the software, to be sure, but Highline handled them differently. One concern was making sure the high voltage was not on when the paint guns were off. The solution was to take "responsibility" for controlling the voltage from the paint guns to the "job," or the cabinet approaching the paint booth.

On balance, Ekberg removed the clutter of safety and failure checks. Sixteen hours after Highline's changes in the paint operation were installed, and enlarged to control five paint booths instead of just two, the plant was in full production and turning out perfectly painted cabinets.

Highline Controls found another simple solution for Shane Steel Processing, also located in Fraser. Shane Steel shears, straightens, and grinds steel bars for the auto industry. In finished form, the bars wind up as springs in cars and trucks and also as structural components. The company's 60 employees process more than 10,000 pounds of steel a month.

When Ekberg was brought in, Shane was handling bars 0.4 to 1.5 inches in diameter but wanted to expand into the market for bars up to four inches in diameter. A conventional consultant most likely would have recommended the purchase of a new $200,000 machine. Ekberg offered a simpler and much less expensive solution. He had a bar feeder and a finished-bar remover attached to the existing grinder, which he found could handle thicker bars. He also supplemented those attachments with "takeaway" arms that handle finished bars and laser gauges that check bar dimensions. Total investment in new equipment: $75,000.

The feeder and the bar-removal systems have eight different components. But in a typical example of Ekberg's pursuit of simplicity, each component is individually controlled, and there are no connections between them. If a component fails, it can be fixed while the rest of the system continues to run. Specific bits of software code called autonomous agents control the feed and removal fixtures and the other components.

In situations like these, says Ekberg, 50% of the gains come from modifying hardware, so there is no reason to write endless lines of software code covering all the things that could go wrong. In some situations where safety is a concern, Ekberg recommends adding redundant sensors, not redundant software. Highline wrote and installed Shane's software in a mere two hours.

Until recently Ekberg and his colleagues wrote customized software only for clients. But a few months ago, in association with Flow-Works, a software company in Naperville, Ill., Highline began offering a package called Flow-Ware, designed to model factory operations reliably, head off bottlenecks and foul-ups, and determine which product flows and staffing are best. Flow-Ware does all this by breaking manufacturing operations into segments. It then uses a subset of CAS called genetic algorithms, in which a succession of randomly generated schedules are "valued" in a computer by how well they measure up against a previous entry in achieving the desired purpose. The schedule that keeps coming out on top in this evolution-style competition is the one chosen for the plant.

Ekberg grew up on Chicago's South Side in a family predisposed toward manufacturing. His father first ran a blast furnace in a steelworks and went on to become president of Birmingham Steel in Alabama. In high school, Ekberg says, he was "good in science, horrible in English." An ardent soccer player, he graduated from Michigan Technological University as a mechanical engineer under a cooperative program while working at GM. The automaker then sent him to Carnegie Mellon University to get a master's degree in manufacturing engineering that focused on controls as well as on computer science.

It was during his 14 years as an operations and maintenance engineer at GM that Ekberg learned the frustrations of dealing with manufacturing complexity. Frequent breakdowns led him to conclude that "you can't put huge integrated control systems on the manufacturing floor--they are not maintainable. You've got to get the complexity out of it."

Ekberg served as GM's project manager on an imaginative conversion of a paint shop at its Fort Wayne truck plant to smoother operation under a complex adaptive system. Automotive factory paint shops are notoriously fickle, and breakdowns often upset the sequence of painting car and truck bodies. Along came fellow Hero of Manufacturing Richard Morley. His company, Flavors Technology, put in a clever CAS to run the plant's ten paint booths. After failing to persuade GM to install the paint system throughout the corporation, Ekberg left in 1996 to set up Highline Controls.

Sometimes Ekberg's solutions go so far as to eliminate some computer operations. He is part owner of Falcon Cold Forming, also located in Fraser, a supplier of torsion bars, anchors, brake pistons, and other small steel parts for the automotive industry. Noisy hydraulic presses at Falcon "take hockey pucks and make coffee cups," says Ekberg, alluding to the compression of raw steel rounds into finished products with a precision of three-thousandths of an inch, higher than can be attained by hot forging.

At Falcon, Ekberg has automated and changed production processes with his CAS techniques, converting a once sickly outfit into a profitable tier-one supplier to GM, Ford, and other companies. Before Ekberg took over, Falcon had 95 employees working six ten-hour days a week to turn out 300,000 parts a month. Now it turns out the same number with 60 people, working five eight-hour days. Ekberg also made a startling suggestion: Remove computer systems that record transactions such as sales, and keep records by hand on paper forms. In a computer-worshipping age, a practitioner of the supposedly exotic doctrine of complex systems has brought some things back to where they started.

MICHAEL MARKS

COMPANY Flextronics International FIELD Contract electronics manufacturing ACHIEVEMENT Building his company into a multibillion-dollar giant HOBBIES Skiing, bicycling, piano

Boss Of A New Electronics Giant

He cheerily admits to having flunked the only manufacturing course he took at Harvard Business School, 23 years ago. Today Michael E. Marks, 49, says he loves manufacturing and "wouldn't trade it for anything else in the world." Well he should love it. Marks is CEO of one of the fastest-growing manufacturing companies in the world, whose annual revenues have been leaping ahead an average of 60% for the past six years.

When Marks took over Flextronics International in 1993 it was an ailing $93-million-a-year weakling on the verge of bankruptcy. In its fiscal year that begins April 1, it expects to hit $8.5 billion in revenues, with a substantial chunk of that coming from a recent acquisition. But it is primarily internal growth that has propelled Flextronics from the No. 22 spot among America's so-called contract electronics manufacturers, or CEMs, to being a contender for No. 2.

CEMs are the unsung heroes of the infotech and telecom boom. They make a plethora of electronics products, from servers to cell phones, that bear the brands of such famous OEMs (original equipment manufacturers) as Hewlett-Packard and IBM. Contract manufacturers are also the unseen force behind burgeoning electronics product sales over the Internet. "We're the speck in the dot in the dot-com," says the boss of one CEM.

The CEMs do more than make things. Flextronics, for example, makes routers and other products exclusively for Cisco Systems at a new $30 million plant in San Jose. It typically gets orders from Cisco electronically, makes the products, ships them directly to customers, and notifies Cisco that the goods have been delivered. CEMs also provide product design and testing services, and run the whole supply chain for OEMs that request it. As more and more OEMs sell their production plants to CEMs, the latter are growing so fast that Marks expects a number of them to be in the $20-billion- to $30-billion-a-year league before long. He plans for Flextronics to be one of them.

Marks' emphasis on manufacturing excellence has been a major force behind Flextronics' growth. When he was assigned to run a Flextronics plant in 1989, he discovered how little he knew about factory production. A business school education wouldn't have been much help back then, even if he had passed that course at Harvard. So Marks hit the books and became one of the first disciples of demand-flow manufacturing, a concept that calls for making products only after orders come in instead of producing ahead of time for inventory. Many of Flextronics' managers and workers were trained by experts from the San Jose office of the John Costanza Institute of Technology (see John Costanza profile).

Taking a visitor on a walking tour of the big San Jose plant that produces solely for Cisco, Marks proudly notes the absence of clutter such as component parts on racks. He points to an area marked RIP in the back of the plant. RIP stands not for "rest in peace" but for "raw-in-process," a demand-flow substitute for just-in-time (JIT) delivery of specified components, which originated in Japan. Unlike JIT, RIP provides for a variety of components to be on-hand to meet unexpected demand.

Cisco isn't the only company impressed with Flextronics' manufacturing quality. Microsoft was so taken that it asked Flextronics to add its name to Microsoft's own on computer mice that Flextronics makes for the software giant in a plant in China. CEM names don't normally appear on OEM products. Flextronics also makes inkjet printers for HP, Pilot electronic organizers for Palm, cordless phones for Ericsson, portable glucose-monitoring systems for Johnson & Johnson, reusable cameras for Kodak, and DVD drives for Philips, among other products.

Marks' second recipe for breathtaking growth has been the development of industrial parks. Flextronics moved in force into low-labor-cost areas near major world markets in China, Mexico, and Hungary while most of its competitors concentrated on the U.S. Flextronics' industrial parks, which it calls campuses, are different. The company buys extra land next to its production facilities and then offers some of it to suppliers and distributors to set up facilities. This makes things easier for the suppliers, which often are smallish U.S. companies. The land already has water, electricity, and other services, and Flextronics has taken care of governmental red tape. For some suppliers, Flextronics even puts up the buildings, which they then lease. Others build their own.

In Guadalajara, Mexico, a fast-growing manufacturing center, Flextronics has ten suppliers on its campus. They include a sheet-metal maker, a plastics manufacturer, a packaging company, and a distributor. Explains Marks: "We wanted our suppliers to be within the range of a forklift truck instead of trucking in supplies from far away." All this naturally cuts costs and speeds finished products to market. Unlike Japanese auto-parts suppliers clustered around Japanese automakers, Flextronics' suppliers are free to work for other manufacturers, including other CEMs, but they wind up doing most of their work for Flextronics. CEMs without such arrangements are at a disadvantage and are beginning to copy Flextronics.

The fast move into new areas has entrenched Flextronics as the dominant CEM in Central Europe and other places. Marks says his company does five times as much business from its Hungarian plants as the other CEMs collectively do in all of Central Europe. Flextronics has two campuses in Hungary, at Sarvar and Zalaegerszeg. The Hungarian factories are an hour and a half from Vienna. "You can be at the factory in the morning and hear Mozart at night," says Marks, a lover of piano music. Flextronics also has a campus in Duomen, China, and is building one near Sao Paulo. Worldwide, the company employs 22,000 people.

In countries with low labor costs, Flextronics turns out inexpensive consumer products such as cell phones and TV-set controllers, which it mostly ships to developed countries. But Marks cautions that "it's a great simplification--and a lot of people fall into this trap--to say that all manufacturing is going to get done in Mexico, Hungary, and China. It's not. Consumer products will be made there. But the infrastructure products--technically complex value-added products--are easy to manufacture in developed countries. That's why we also have big operations in the U.S., Germany, France, and Sweden, where you have high capabilities in engineering. The OEMs like us to be everywhere." Flextronics is growing in the U.S., Marks makes clear: "We're expanding in Dallas, North Carolina, and Boston. We can't hire people fast enough."

The company doesn't just manufacture. It runs 11 so-called product introduction centers, where its engineers help OEMs design, build, and test prototypes and launch new products. Flextronics engineers helped with the design of the Palm electronic organizer, for example. Aiming to cover all bets in e-commerce product development, the company has been investing heavily in e-business companies that provide portals on the Web. Marks foresees an increase in joint product development by designers in different locales who communicate by the Internet.

Marks gained some of his business acumen during his childhood in St. Louis, where his father ran an air-conditioning and heating-equipment distributor. He worked in the business alongside his four brothers. He was interested in math and science and holds a master's degree from Oberlin in experimental psychology, which makes heavy use of statistics. After earning an MBA at Harvard, Marks ran a $10-million-a-year computer maker in St. Louis. Lured by the excitement of digital technology, he then set off for Silicon Valley, arriving there with his wife and two children and no job.

Soon he was working for Flextronics, but a year later left to work for other companies in the area. He ran one, a maker of chip-processing equipment named Metcal. When Marks rejoined Flextronics in 1993, the company couldn't afford the costly changeover to a new method for assembling computer boards. Marks became CEO after he organized a group of famous venture capital firms, including Kleiner Perkins and Sequoia Capital, to buy control of the company for $12 million.

After pulling Flextronics out of its nosedive, he began to execute a growth strategy that included improved productivity. Marks recalls that when he ran a Flextronics plant in 1989, it took 13 days to go from raw material to product. Today, he says, "in all our facilities around the world, with the exception of products that have to be burned in"--tested in heat chambers--"nothing takes longer than a day to build. Today, the whole world is about speed. Just build it, and move it right out."

Stories from FORTUNE's Industrial Management & Technology section can be found at www.fortune.com/imt. Write to GENE BYLINSKY at gbylinsky@fortunemail.com.