Lapas attēli
PDF
ePub

during a discussion of assembly planning. I had given an example of alternate assembly sequences for automatic transmissions, indicating that one sequence required mating several gears at once while another mated them one at a time, distinctly easier. The example was intended to show that there is more to DFA than just reducing part count.

Their reactions were several. First, manufacturing engineers work out the assembly process. They critique the detail design as well, pointing out opportunities to use existing tooling by recreating the hole spacing from a past design, or increasing spacing between holes so that all the bolts can be tightened at once. All this is based on past experience and data in people's heads. Second, so many meetings are held that 95% of all the assembly problems are discovered and eliminated. "Drastically speaking, our production engineers are our software," says Kuranaga. This confirms Nissan's view that Toyota can throw people at a problem and solve it. It also supports the view that many Concurrent Engineering activities that we would account for individually have become second nature at Toyota, and no need for specific efforts or additional software is seen.

Toyota uses CAD models to check tool access during assembly. Models of both tools and human hands are available in the database. Also, designers of transmissions and engines take into account the need for chamfers on parts, that a worker has only two hands, that assembly should occur from one direction, and so on. Transmissions are assembled manually. However, engines are assembled by a mix of manual and automatic methods. No formal design methods are used for taking these different assembly techniques into account, as far as I could tell.

As a result of its many meetings and universally experienced engineers, engineers, Toyota has no use for traditional DFA and feels that companies that use it

must have poor communication and experience. The fact that both GM and Ford have made much of their success using DFA only confirms to Toyota that their competition is weak. They are apparently unaware of the innovative uses that companies like Sony and Nippondenso have made of DFA methods, where communication is simply not the issue.

It is possible that Toyota makes too much of communication and, like Nissan during my first visit, felt that I was interested in how computers can support communication. Since few Japanese companies think such support is necessary, I usually hear at first that computers are not essential to the design process. Only later did it emerge at Toyota that Iwase's new CAD project will indeed contain as much process engineering computer support as he can obtain or create, including that for assembly.

I also learned that Toyota is actively investigating new workstation-based CAD software, such as ProEngineer. Negishi is quite impressed with it. However, all CAD systems he has surveyed, including ProEngineer and Ricoh's Design-Base, are unable to represent the complex fillets he wants for modeling connecting rods and other similar parts. While I originally thought CAD vendors' boasts about their filleting capabilities were self-indulgence, this remark by Negishi shows that some weight-critical parts can be essentially all fillets. Thus filleting can be crucial in certain situations.

A final point. My hosts at both Toyota and Nissan were CAD support people responsible for providing computer capability to engineering. The computer people are distinctly ahead of the engineers in their thinking and often propose capabilities or practices that the engineers see no need for. However, the history is that the capabilities are eagerly used as soon as they are made available.

Computer-Aided Engineering in Power Train Design

Apparently the typical CAE applications are in use, including their own software for supporting engine and transmission design. No software is available for setting tolerances, nor is there any support for selecting the style of transmission design. "An experienced person does it." Not enough detail was obtained to make possible an interesting report on this complex and rich topic. Another visit should be arranged.

Tour of CAD/CAM of Stamping Dies at Motomachi Plant

The CAE of dies was covered briefly in the first visit report. The tour covered the CAD facilities and the machining area. Figure 6 of the first Toyota report illustrates the elements of die design and manufacture. Our hosts for CAE and CAM were Mr. Muta and Mr. Amano.

CAE of dies has been forced by the huge growth in diversity of cars and the short design cycle. Many more dies are needed much sooner than before. As long as 10 years ago Toyota foresaw needing many more skilled people and more of their time than could be provided and therefore launched the CAD/ CAE/CAM effort.

The CAD facilities are centered on a large UNISYS mainframe and many terminals. Data on die shape come from body engineering. In this area, die design is completed by the addition of details, clamp surfaces, cutters, and so on. Extensive software written by Toyota then determines the tool paths so that all the die's details can be cut, usually with a 1-inch-diameter ball end mill. Smaller tools are used only when necessary.

CAM consists of two rows of four large NC machines each. The first row was built by Toyoda Machine Tool Co.

in the early 1980s and consists of five axis machines. These yield die accuracy in the ±50 μ range. Toyota decided during the mid-1980s that this was not accurate enough and developed the second row of machines, which were installed in 1988. These are three axis (a surprise) and yield accuracy in the ±20 μ range. Three axis NC has the advantages of being more rigid and of having a smaller tool socket size, reducing interference problems between machine and die. On the new system a die stays on one machine 20 to 40 hours while it receives all the necessary cuts. The shop runs three shifts and runs unmanned over the weekend.

The need for higher accuracy came from identifying quality problems in earlier cars. An example shown is surface waviness near the edge of a door. Accuracy of internal structural parts is just as important as accuracy of the outer panels. (For reference, there are between two and three times as many inner panels as outer panels. Since there are typically 30 outers, there are at least 100 total per car and some require several dies.)

printed on the female die, the tool-
maker can see how much material must
be removed.

Why is this technique needed when
20μ accuracy is obtained? Apparently
this was an embarrassing question. The
answer was not entirely satisfactory:
there are not enough high accuracy
machines in the shop to make all the
dies to 20 μ. Either they are made on
less accurate machines, or they are made
by outside contractors to lesser accu-
racy, or the time is not devoted to a
final cut at a fine pitch. More hand
finishing is needed, adding error. Thus
some dies must be hand-fitted.

Unlike any other shop I have visited that attempts work this accurate, this shop was not temperature-controlled or air conditioned in any way. The 20 μ is clearly meant as a relative error limit, as we could tell by noting how the CMM was working. That is, it is only necessary to be accurate relative to the hard points on the two dies that bottom on each other at the end of the stroke. Most dies are too shallow for temperature excursions to make large changes in height relative to the hard points. Length and width were not being held to such accuracies.

The dies are machined using 5-mm pitch cuts at first, then finished with about 0.5- to 0.7-mm pitch. Very high feed rates (as high as 4 m/min, they claim) are used on the final cuts. The resulting dies look like they have been sandblasted, with only a hint of linear tool marks about 15 μ high. Before finishing, the dies are checked in a coordinate measuring machine (CMM) to be sure that the 20 μ is obtained. About 15 to 20 hours of hand stoning and emery cloth polishing by three people then converts the die to a smooth, Visit to Teiho Plant almost mirror finish.

It takes about 22 days (three shifts) to make a die, according to Muta, of which about 3 to 5 are for tryout. These figures must be averages, since tryout of difficult dies often can take much longer. I was not able to find out how long it takes in such difficult cases, but Figure 9 of the first Toyota report shows tryout times in hundreds of hours, which is many days at 24 hours per day.

Mating die pairs are checked with yellow transfer ink. Where a space equal to sheet metal thickness must remain, sheets of rubber are laid on the male die and then inked. These sheets have raised patterns of different heights in fractions of a mm with a different pattern for each height. From the pattern

At this plant, Toyota designs and makes automation equipment, some of which it sells as well as uses internally. Mr. Takano showed us local area network (LAN) equipment and associated controllers, cables, and connectors all designed to Toyota's own standard. These controllers were attached

to a variety of Japanese computers, controllers, and robots. The data rate is 1.25 MB/s.

We also saw a video illustrating several pieces of factory automation designed and installed by this division. These included beam-transfer handling systems for merging car bodies, chassis, engines, and axles; vision-aided robots for installing dashboards; and a system of three robots that (in a rather complex way) installed wheels on cars and then installed and tightened the nuts. The body-engine-axle merging system looks somewhat like what VW installed in Hall 54 in the early 1980s.

Kuranaga noted that the dashboard being installed was empty, that workers later must lie down on their backs in the car and install the instruments and wires. "We do not install complete tested cockpits. In fact our weak point is that only 5% of our final assembly is automated." VW claimed 25% in Hall 54, and GM has been installing complete cockpits in some cars for several years.

A tour of this division's workshop turned up two cooperative robots programmed by mutual timing to open a beer bottle and pour beer, some AGVs being tested, plus several conventional assembly and machining lines under construction. The beer opening robot contained no sensors and operated in a very conventional way available to any commercial robot. I was told that in a factory, a similar pair is at work loading balls into constant velocity joints.

The other equipment was also conventional. One machine consisted of robots and transfer equipment intended to weld together three parts for the tube and brackets of MacPherson struts. These are simple parts and such struts are undoubtedly available from any number of vendors. Given the outsider's impression that Toyota buys conventional components, I was surprised to see this machine. No explanation was available.

However, I have heard separately that Toyota is beginning to bring as

much as 10% of routinely procured parts in-house so that some idea of a fair purchase price can be determined internally. This includes not only simple parts like the struts but complex electronic items.

Concluding Remarks

These two visits obviously barely scratched the surface of one of the world's leading manufacturing companies. It is clear that use of computers in the design process is growing rapidly and that it is company policy to encourage this process. The priorities are focused on designmanufacturing integration of efforts that currently take a lot of time, for which accuracy is required, or for which human experience is needed and ways exist to augment this experience.

IBM FUJISAWA PLANT 9 September 1991

Background

My hosts were Mr. Sawada from the Tokyo Research Laboratory (TRL), Mr. Tsunoda of Manufacturing Engineering, and Dr. Koda also from TRL. This visit was mostly a plant tour to see manufacture of hard disk drives (HDD) for PCs.

Product and Factory Design

A disk drive is designed by a team of about 30 designers. This total includes all support people, such as draftsmen, of whom there are few or none. Design is done at another plant in Yokohama about 10 miles away. The HDD assembly system is almost totally automatic, comprising many class 100 clean rooms, robots, conveyors, stacker cranes, and test equipment. IBM and its vendors took 9 months to design, build, and install this large system. In the future, Tsunoda notes, they must reduce the time to 6 months.

The decrease in time to produce the factory coincides with the faster pace of new product introductions. Only parallel development of product and processes will be capable of sustaining this cesses will be capable of sustaining this development. This is Concurrent Engineering, according to Tsunoda, who did most of the talking during this discussion and appeared not able to answer my questions easily.

designed, then to a third group at the second facility where the equipment is designed or specifications are written, and finally giving vendors the specifications for fabrication and installation of the equipment. For example, no IBM robots are among the 10 used for HDD assembly, test, or material handling.

This method is very much in the American style. This style prevents the typical Japanese method of bottom-up automation from occurring. The technology belongs to the vendors. Each department is a specialist. "We Japanese learn by trying, not by buying. The Japanese maker's treasure is his factory." "So IBM Japan is an American company with Japanese employees?" "Yes."

A possible reason emerged later in
discussions with younger engineers. I
was shown a promotional videotape
that first described IBM's software vision
of CIM, then showed how it has been
implemented at Fujisawa. The icon of
this video is a set of three concentric
rings. Each ring is segmented into activ-
ities: on the outside are the factory's
activities (engineering, marketing, etc.),
then inside are the common support Plant Tour
elements like computers and displays,
and on the innermost ring are the archi-
tectural elements like communication,
databases, and presentation software.

This icon has been associated with
IBM's CIM publicity for about 10 years
and it gives the impression that if one
has the right software one has CIM.
The video confirms this by listing all
the IBM products or VAR items that
make up the system (CATIA, EDCS,
VALISYS, COPICS, MRP, DB2, SNA,
RIC, AS, QMF, SMART, and AD/
Cycle), omitting explanations of what
they stand for.

It is clear, however, that behind the acronyms there are problems because in fact the communication channels between people are not as open as they should be. HDDs are very hard to design, especially high performance ones (not shown to me). IBM is not the only company I know where the design is passed on to the manufacturing system people without their being able to comment on it, much less get a head start on their system design.

The method consists of product designers at one facility making the design, passing it on to another facility where process specifications are

This is a truly integrated factory for HDD assembly. (Procuring and installing it in 9 months was clearly quite an achievement.) Incoming parts are stocked in an automated warehouse in boxes or bins. They are called out by the assembly equipment as the need arises. This is called "Auto Pull Mode," an automated version of the Kanban system. Parts are first sent by automatic conveyor to a cleaning machine which presently soaks them in a swirling Freon bath. This bath will soon be replaced by one that uses just water. Then they travel by conveyor to the assembly area.

The mechanical portion of the product consists of about 16 parts plus 4 vibration isolators and 5 screws. Unidirectional assembly is used. A line of 10 robots builds it up, starting from a kit of parts (base-motor-spindle, head package, set of platter spacers) manually loaded into the pallet. The robots take the platters from a feeder magazine, where they sit upright, and put them gently on the drive spindle, then insert a spacer, then another platter, and so on. Finally a cap and screws are installed. Each drive then gets servo marks written and a 10-minute test. If it passes

this test, an elastic seal gasket and the lid are installed using more screws, and the drive is leak tested. If it passes, it leaves the clean room for more electrical tests, addition of the logic board, still more tests, and finally packaging. Some of the equipment in this second line, especially the screw feeding and installing parts, seemed very elaborate and complex. The tour was so rushed that I could not ask about any of the details.

Only a few of these steps require human intervention. One of these is the place where TRL's force-controlled robot is supposed to plug the test connector into the drive. A second feeder robot was broken so the whole station is down. This robot uses a low insertion force connector invented by IBM. It reduces the insertion force by about half by recessing half the contacts part way in the insertion direction. Half the males and females mate early in the insertion process while the other half mate later. The insertion force history, therefore, has two small peaks rather than one large one.

REFERENCES

1. J.P. Womack, D.T. Jones, and D. Roos, The Machine That Changed the World (New York: Harper Collins, 1990), Chapter 4.

2. R.E. Gomory and R.W. Schmitt, "Science and product," Science 240, 1131-2, 1203-4 (27 May 1988).

3. K.B. Clark and T. Fujimoto, Product Development Performance: Strategy, Management, and Organization in the World Auto Industry (Boston: Harvard Business School Press, in press).

4. J.L. Nevins and D.E. Whitney, Concurrent Design of Products and Processes (New York: McGraw-Hill, 1989).

[blocks in formation]

Daniel E. Whitney received his B.S., M.S., and Ph.D. degrees in mechanical engineering from MIT. He has been with the Charles Stark Draper Laboratory (CSDL) since 1974 and is currently in the Robotics and Assembly Systems Division. Before coming to CSDL he was an associate professor of mechanical engineering at MIT. At CSDL Dr. Whitney's research centers on assembly automation: part mating and assembly systems analysis, application of control theory to robot operations, supervision of robot assembly machine design and fabrication, tradcoff analysis of automation systems, and producibility analysis of products. He also docs industrial consulting on complex products and systems design and operation, including shipyards. Dr. Whitney holds a number of patents and is a Fellow of ASME and a Senior Member of IEEE.

« iepriekšējāTurpināt »