There was another problem, which was that the low-temperature alloy had to be removed from the blade after the grinding operation. Several sophisticated steps were then required to ensure that the alloy was truly removed. (Even microscopic amounts of the alloy would cause hot spots and rapid failure of the blade once in the engine.) These involved X-rays and an atomic absorption process using caustic chemicals to test for trace elements of alloy. This last step created a serious environmental problem of radioactive acids as well. The system as it was installed is shown in Figure 8.4 .
F IGURE 8.4: A UTOMATED B LADE G RINDING C ENTER
Yet another problem was the changeover times needed for the Blohm grinders to convert from one family of parts to another. Because of the need to move layer after layer of automation away from the grinding tool in order to change it, eight hours were needed for every changeover. The planners of the system apparently believed that extremely long runs of parts would be possible—permitting completely automated mass production—but in practice Pratt needed to make small numbers of a wide variety of blades. The long changeover times prevented this and required the production of large batches of each part type instead.
Finally, many of the direct and indirect hourly workers had to be replaced by skilled technicians who debugged the elaborate computer system controlling the entire process (with two thousand parameters). In the fall of 1993, when Ed Northern arrived, there were twenty-two technicians tending to the needs of the Blohms, a number not much smaller than the number of direct workers needed for the old manual system.
In the end, eight of the nine processing steps involved in the new system, plus the AGVs and the ASRS, added no value whatsoever. What was more, the three minutes of grinding time were accompanied by ten days of batch-and-queue time to get from the beginning of the encapsulation process to the end of the deencapsulation process. And the complex machinery was temperamental. Even at the end of a lengthy learning curve, it was difficult to get past about an 80 percent yield. A disappointing result from an $80 million investment.
We mention the Blohm grinders because they exemplify a whole way of thinking which is now obsolete. The twin objectives of speeding up the actual grinding—what you might think of as a “point velocity” within a lengthy process 36 —and the desire to remove all hourly workers because of their “high” cost per hour both miss the fundamental point. What counts is the average velocity (plus the length of the value stream) and how much value each employee creates in a typical hour. (We’ll return to this point in the next chapter when we discuss German “technik.”)
Initially, North Haven tried to work around the Blohms, placing their step in the turbine blade fabrication process behind a “curtain wall” so it would not interfere with single-piece flow in the rest of the process. But this was difficult. The great majority of the cost in the total process was caused by the Blohms, and their erratic performance thwarted attempts to achieve smooth flow in the rest of the process. They needed to retire.
F IGURE 8.5: L EAN B LADE G RINDING S YSTEM
Cell with eight 3-axis grinding machines and two electrostatic discharge (EDM) machines (drawn to larger scale than Figure 8.4 )
By late 1994 the process mapping team at North Haven had the answer. They proposed to replace each Blohm machining center with eight simple three-axis grinding machines utilizing ingenious quick-change fixtures to hold the blades firmly in the machines without the need for encapsulation. 37 Each cell would have one worker to move parts from one machine to the next by hand, standardize his or her own work, gauge parts to check quality, change over each machine for the next part type in less than two minutes (with the help of a roving changeover assistant), and make only what was needed when it was needed.
By increasing actual processing time from three minutes to seventy-five minutes, the total time through the process could be reduced from ten days to seventy-five minutes. Downtime for changeovers could be reduced by more than 99 percent (as each of the nine machines was changed over just-in-time for the new part coming through). The number of parts in the process would fall from about 1,640 to 15 (one in each machine plus one waiting to start and one blade just completed). The amount of space needed could be reduced by 60 percent. Total manufacturing cost could be cut by more than half for a capital investment of less than $1.7 million for each new cell. No encapsulation; no AGVs; no automated storage warehouse; no deencapsulation with its environmental hazards; no computer control room with its army of technicians. Lean thinking at its best, as summarized in Table 8.1 .
T ABLE 8.1: L EAN VERSUS M ONUMENTAL M ACHINING
When the first of the new cells—called chaku-chaku, meaning “load-load” in Japanese—went into operation at the beginning of 1996, North Haven was on its way to a cost and quality position, using a high-wage, high-seniority workforce with “simple” machines in a World War II vintage (but immaculate) building, that no one in the world could match.
The latter fact led to the final step of Ed Northern’s strategy. He knew that lean thinking would continually free up more workers and resources. Unless he proposed to continuously hand out termination notices and explain to his work teams why they should continue to put their hearts into working for a company with no apparent interest in protecting their jobs, he needed to find more and more work and find it quickly. (Ed calls this “keeping hope alive.”)
One method was to take work back from suppliers, particularly when incorporating it into North Haven’s activities permitted more continuous-flow production. (It’s important to understand that this is a one-way process. A firm cannot take work in to suit its needs, then subcontract it again later to suit new needs. The suppliers won’t be there.) A second approach was to enter the turbine blade repair business in collaboration with other units of Pratt, taking on engine overhaul work, another world of batch-and-queue thinking awaiting a lean awakening. Both concepts were well along in planning in 1995.
The Continuous-Flow Engine
Meanwhile, in the final assembly hall, Bob Weiner energetically introduced lean principles from the moment of his appointment in July of 1994. As Ed Northern’s former deputy at GE Aircraft Engines, it’s not surprising that the steps he took were exactly the same: Cut headcount at the outset to a level which can be sustained for the long term, replace managers who couldn’t adjust to the new system, standardize work, and deal with quality problems so work could flow continuously. Then introduce continuous flow.
As Weiner and his team studied the situation, they realized that Chihiro Nakao’s goal of a three-day engine was achievable, but would require a substantial investment to combine the assembly hall with the test cells 38 in another building. However, simply by introducting modular assembly—what Nakao called a “head of the fish” system, where major components flowed in fully built up and ready to snap together from the Product Centers representing the bones of the fish—they found that by mid-1996 they could cut the time through the process from thirty to ten days and substantially reduce assembly effort. The key was to place the engines on an imperceptibly moving track and to eliminate all backflows and rework caused by upstream quality and delivery problems. The new system brought the component modules and tools to the assemblers in kits so they didn’t waste time on “treasure hunts” and provided the assemblers with a simple PC-based system beside the assembly line to display assembly diagrams and instructions relevant to a each step.
A Concurrent Quality Crisis
The final problem to be overcome was a concurrent quality crisis. In 1993, Pratt was besieged with customer complaints about the rate of in-flight shutdowns of its engines, the primary measure of quality in the aircraft engine industry. Indeed, several airlines were threatening to cancel future orders or even to go to court to claim damages, and it appeared that Pratt’s in-flight shutdown rate on some engines was running at seven times the level of GE’s and Rolls’s.
In one way, this seemed impossible. In 1992, Pratt’s Quality Assurance Department had 2
,300 employees and everything that could be checked was being checked. But on another level, it was clear that the quality movement of the 1980s had gone badly wrong. Quality Assurance had become the classic corporate superego or nagging nanny, checking up on production employees to make sure they hadn’t taken shortcuts on quality in order to meet production targets. This, of course, created a very negative, reactive reputation for Quality Assurance.
It also meant that production managers cheerfully referred any alleged quality problem to a series of Material Review Boards (MRBs), which decided long after the problem was first noticed whether parts rejected by Quality Assurance were acceptable to ship. In the early 1990s, Pratt was conducting 66,000 MRBs per year. But 90 percent of the time the part was finally accepted for shipment “as is” because the variation from the formal specification was deemed to be insignificant. This was after lengthy delays and hours of meetings to assess the problem.
One solution to this problem was to completely reorganize the Quality Assurance Department under a new head, Roger Chericoni, a longtime Pratt product engineer with no quality background and no baggage. Only 150 employees were retained in this function, with the rest being assigned to business units on the plant floor to directly resolve quality issues as they arose.
The other solution depended once more on George David, who had in fact been enlightened about lean thinking twice, the first time several years before meeting Art Byrne. In the 1980s, in his position as head of Otis Elevator, he had also been chairman of the Nippon Otis joint venture with Matsushita in Japan. In 1990, David had faced a crisis when Matsushita announced that it felt it could no longer put its National brand name on the joint venture’s products.
“The head of Matsushita called me to point out that our product had for years been breaking down four to five times more frequently than competitor products from Hitachi and Mitsubishi. Given our history of product performance in Japan, I knew our relationship was moving toward a breach. I also knew that if Otis couldn’t compete with Japanese firms in Japan, we would eventually lose to them elsewhere.”
Fortunately, Matsushita offered help in the person of Yuzuru Ito, Matsushita Electric’s corporate quality wizard who was dispatched to help fix the quality problems at Nippon-Otis. “We needed his help because we were determined to make our products the best, but we just didn’t know how. It was that simple.”
With Ito advising an Otis task force on its quality problems, “call back rates” (elevator industry-speak for the number of times per year an emergency call is required to fix a malfunctioning elevator) began to plummet and eventually fell below those of Hitachi and Mitsubishi. David notes, “There’s no doubt Ito-san single-handedly saved our relationship with Matsushita and made it possible for an American firm to succeed in Japan against the best Japanese competitors.”
When Ito retired from Matshushita shortly after this episode, George David pleaded with him to help Otis full-time, and when David moved up to president of United Technologies in 1992, he expanded his mandate to all UTC companies. Eventually, he even convinced Ito to move his home from Japan to a site near UTC corporate headquarters in Hartford.
When Ito started to help with manufacturing operations at UTC, it turned out that his techniques were based on flow thinking. He used “turn-back rate charts” to see how many times mistakes interrupted the flow of production. He always found that continuous flow and perfect quality were achieved together, after rigorous root-cause analysis and corrective action.
“When the Pratt quality problems at the customer level began to reach a crisis in 1993, I knew there was a perfect interlock between Ito’s quality philosophy and Shingijutsu’s flow philosophy. I realized that together they were an unbeatable combination, so I told Ito to devote all of his time to helping Roger Chericoni at Pratt.”
After targeting the root cause of the in-flight shutdowns, Ito turned his attention to the general problem of backflows in the Pratt production system. For example, at North Haven, the 10 percent of product getting through a typical manufacturing process the first time was soon raised to practically 100 percent.
The Bottom Line in Physical Production
By mid-1995 Pratt had totally revamped its entire physical production system. The mass-production, batch-and-queue, “tinker till we get it right” philosophy built up over nearly 140 years was gone and the company was completely converted to a flow organization stressing first-time quality with no backflows.
The MRP system formerly driving the movement of every part had been reassigned to the task of long-term capacity planning and long-lead-time delivery of parts from suppliers not yet lean, while flow through each module center and into final assembly was regulated by a simple pull system.
The eighty business units, one for each major part family within a component module, were reconfigured both organizationally and physically. Business unit heads were given a much simpler “scoreboard” with a much smaller fraction of allocated costs (in a system similar to the Wiremold approach we saw in Chapter 7 ), and told to manage costs down through kaizen activities. Production engineers and quality experts were physically reassigned—that is, their desks were moved—from “upstairs” in plant offices or at engineering headquarters to space on the shop floor within or immediately adjacent to the work cells.
In the end, all seven thousand of Pratt’s machines were moved (some many times), and by the end of 1995, every production process in the entire Pratt & Whitney Company had been kaikaku ed and kaizen ed at least once, with the objective of creating a continuous-flow cell for each part with substantially zero in-process inventory within the cell. At the same time, a host of improvements in quality thinking spurred by Ito were leading toward “certification” of every process—that is, redesign of activities and adjustment of tools—so that first-time quality with no backflows for rework could be absolutely assured.
As a result, throughput time fell from eighteen to six months (with a near-term target of four); inventories of raw materials, work-in-process, and finished goods on hand fell by 70 percent and are still falling; the massive central warehouse which formerly stored all parts moving between production steps was closed; referral of quality issues to MRBs declined by more than half (with a goal of eliminating MRBs by the end of 1996); and unit costs for a typical part have fallen 20 percent in real dollars even as production volume has fallen by 50 percent. This last measure is perhaps the most important because in the old days of mass production, Pratt’s unit costs would have gone up by 30 percent or more in this circumstance and the company would probably have been forced to merge or exit the industry.
The original goal of a 35 percent cost reduction, set at the beginning of the crisis in 1991, remains in place but the collapse in demand, which only began to reverse in mid-1996, means it’s going to take a bit longer to get there. In addition, while Pratt’s own costs have fallen dramatically, the supply base, which now accounts for more than half of Pratt’s total production costs, must now be kaikaku ed and kaizen ed to the same extent as Pratt. In many cases this will involve rethinking whole industries, in the fashion of the glass example in Chapter 5 , in order to introduce time and cost savings and quality improvements back through casting and forging all the way to base metals.
The Point of No Return
The critical moment for the lean transformation at Pratt occurred in the spring of 1994. Although upstream aspects of production were steadily improving, problems in delivering engines to customers meant that nothing was visible to the external world. The unwillingness of the old management to adopt to the new system and errors in implementation at various points upstream caused Pratt to deliver only 10 percent of its engines on time, a historic low.
As Mark Coran remarked later, “I kept wondering that spring why I still had a job in a situation where the results of our efforts weren’t yet showing up. But in retrospect, the secret was simple: George David and Karl Krapek, unlike most senior executives in American firms, actually understo
od what I was doing. They realized we were going to have steps backward along with our steps forward and that the trick was to hold an absolutely steady course.”
As soon as the new management was in place in final assembly in the summer of 1994, and once Ito’s quality initiatives began to show results and a pull system from final assembly began to replace MRP across the company, everything came around very quickly. What was more, the new general managers began to clamor for more time and help from Bob D’Amore’s beefed-up Continuous Improvement Office, and Pratt was able to sustain the gains made in the weekly improvement blitzes. However, it had taken more than three years of hard work to reach a point where turning back became unthinkable.
The Next Leap
In 1995, Karl Krapek began to turn his attention to the rest of Pratt, where the slow-moving, inward-looking product development and engineering system had changed only modestly. The organization chart at this point, with the new Product Centers fully in place, looked unsettlingly like a Rubik’s Cube (as shown in Figure 8.6 ). Any new product program involved an elaborate matrix of divided responsibilities and loyalties between the product development teams (called Propulsion Centers), Pratt’s core technologies in seven Component Centers, and the detailed engineering and manufacturing in eight Product Centers.
In simplest terms, developing a new product meant defining the whole (thrust, weight, fuel consumption, product cost) in a Propulsion Center, engineering and producing each major component in Component Centers, and then engineering the individual parts making up each component in the Product Centers. The project was essentially handed off twice between three massive organizations reporting separately to the president. Confusion and high costs were the predictable results.
The solution, which was announced at the beginning of 1996 but which will take all of 1996 to implement, is to create much stronger Propulsion Center product teams, including dedicated component design engineers. The design engineers remaining in the Component Centers will be relocated either to a small engineering function charged with developing new design methods and technologies, as well as maintaining design standards and engineering systems, or to one of the new Module Centers created out of the current Product Centers to give a “lean organization” as shown in Figure 8.7 .
Lean Thinking Page 22