« iepriekšējāTurpināt »
Figure 11. Parameter model and algorithm for the self-maintenance copier. From Ref 19; reprinted with
permission from ASME.
The company has 1,700 employees and of part names, machine models, part
makes about 3,000 machines per year. size and material, and year the part was Qualitative models appear to have Unlike Yamazaki or any of the designed. It is a memory-jogger system many advantages, among them being “big 5” in Japanese machine tools, H-S that behaves like a library card catalog. the ability to represent important behav- makes special transfer lines for the auto Any way the designer can think of to ior patterns with relatively simple industry. These are totally different in help him remember a part is available models. However, it remains to be seen technology and structure from the NC to help him narrow the search. The whether they can ever attain the sophis- lathes and machining centers which output is a part number. He must then tication and complexity of well- comprise 80% of its sales. Like go to the CADAM system and punch in developed quantitative models, such Yamazaki, H-S has trouble selling FMS. the drawing number in order to get a as those exhibited by the automobile Customers want simpler machines and picture of the part. companies for simulating crashes, skids, systems such as the six-pallet storage They have brought in outside and air flow. Tomiyama feels that QP machine. The pallets can be loaded in consultants in the hope of making their not only can coexist with quantitative and the machine left overnight to work factory floor more efficient. Right now models but has the potential to model the parts. H-S has not yet resorted to a few hours of actual cutting time takes types of phenomena that cannot be high-pressure sales tactics for FMS the 10 days. All they got from the consultants represented in traditional models. My way Yamazaki has. Yamazaki even over 2 years was advice to group all the feeling is that augmenting quantitative makes its own six-pallet machines look needed machines around the parts, which models with logical state variables could bad in economic comparisons in the is a variety of flow shop. Such an increase their generality and a meeting hope of selling more FMS.
approach is bound to be costly and ground in the middle might be found. H-S's main business problem is inappropriate for a job shop environ
managing fluctuations in demand. Five ment. Analytical techniques exist to HITACHI-SEIKI (H-S),
years ago the transfer line business fell address this problem, but it is a tough ABIKO WORKS
apart and many people had to be one. The FMS was invented in the first relocated. Today they have 50% more place to address just such situations,
, 21 August 1991
design business than they can handle. and nothing much better has come along
"It's a big panic." Their CAD does not since. Only Yamazaki has learned how Background
help them much. They work nights and to design its parts to make FMS
weekends and take on marginally capa- particularly efficient. H-S's president Our hosts were Mr. Kubota, R&D ble subcontractors whose work they has challenged his people to become Center Director, Mr. Ishihara, Con- must often redo. H-S is the only com- more standardized in their design trols Design Manager, and Mr. pany to admit to me that this type of methods. They are still in the process Takeshita, General Manager of the overwork hurts product quality, includ- of interpreting this challenge and may Technology Development Center. Other ing design quality.
again utilize outside professional coninteresting participants were Mr. Otani, However, you can't say no to a sale, sultants. Right now their only response General Manager of the Design Depart- especially from a car company. They is to consider more reuse of existing ment, and Mr. Katoh, Manager of the know when to buy, that is, when the designs and more modularization of Technical Research Department. suppliers are hungry. H-S notes, in a their subassemblies. They do not have
H-S makes machine tools very sim- fascinating aside, that they used to be a strong tradition of the "series method" ilar to Yamazaki's, but the company is able to count on a lot of design help described by Yamazaki. In general, they about one-fourth Yamazaki's size. It is from the car companies’own engineers do not have a clear view of the imporin a similar state regarding CAD and is but that's not true any more. Those tance of design methods, designed less modern in its manufacturing facil- engineers now have only 2 or 3 years' characteristics of their machines, and ities. However, H-S uses its own FMS experience and H-S ends up teaching design technology to improving their for making almost every part it uses. them!
company. No other kind of machine tool or sys- A major reason why H-S is having tem was observed. In fact, H-S uses not trouble is that they are behind in CAD Use of CAD/CAM/CAE only FMS but also several temperature- and cannot utilize many sophisticated controlled rooms for machining and concepts such as data archiving that H-S relies on CADAM. Ninety-five assembly for the purposes of improv- would permit them to use past designs mechanical engineers access 28 tering quality and of impressing customers. more quickly. All they have is a home- minals and about 20 drafting boards.
made alphanumeric database consisting They would buy more terminals if they were not so costly (average Y400,000/ As far as I could tell, there were no the aim of obtaining a final accuracy month including the host IBM 4341). computerized methods in use for deter- requirement. This is much more sophisThey think workstations will be their mining line balance or job assignment ticated than their current method of next move.
in the design of these transfer machines. just using higher accuracy components. All the facilities of CADAM are One experienced person accomplished Similarly, cost analysis is based on used, including generation of NC data. these essential conceptual steps in about past data, especially since 80% of the This is sent via fiber optic cable to the 3 months.
parts are identical to existing ones or factory control room.
To support all of this design activ- nearly so. New parts are hard to preEight months ago they got CAEDS, ity, H-S has 95 mechanical engineers dict. The main components of cost are the IBM version of SDRC's I-DEAS and 35 electrical/electronic engineers. materials, vendor costs, and machining solid modeler, and they are just learn- Another 20 engineers are in R&D. time, expressed in standard hours. ing how to use it for stress and defor- About half of the engineers are in mation analysis of machine tool beds. production and manufacturing. They R&D Activities
Even if they get a terminal for every comment extensively on the product designer they will not abandon paper designs but apparently do not take a H-S has originated a number of because the CAD screen is too small. large part in product design itself. About interesting concepts. From their own Other companies have said the same half the product designers are skilled in experience using their products, they thing.
production techniques already. Even realize that setup time is a major head
so, H-S agrees that there is too much ache for customers. So they have develProduct Design Methodology delay and correction to the design draw- oped novel techniques for easy setup
ings; about half are returned for changes and programming of their machines H-S has two product lines, medium or questions.
and easy in-process measuring of parts. size NC lathes and machining centers Engineers gain their breadth of This is based on programmable logic and transfer machines. The NC machines experience from a short 4 months ini- controllers (PLCs) that they build themtake about 2 years to design: 1 year for tial training plus continuous rotation selves and link up to Fanuc controllers. concept design, 6 months for detailed until age 45 or so. After that, an expe- The PLCs do the work while the Fanuc design, and 6 months for preproduc- rienced designer stays put. Those that controllers provide a familiar name and tion tryouts and performance docu- do not understand production methods an easy user interface. mentation. This leaves out fabrication and new technology are "a headache." Other R&D activities include direct of the prototypes, but I assume it is H-S is having trouble attracting new drive spindles (like Mazak's, I presume) included in the first 18 months. employees. North of Tokyo is not con- and two-component casting of machine
The transfer lines present a totally sidered “in” territory. Other Hitachi ways. These castings are cast iron underdifferent kind of design task, since they companies located in the area said the neath plus a layer an inch or so thick of are made up of a linear string of stan- same thing.
steel on top, where the ways are dard modules. Each module drills a set Otani, the Design Manager, oper- machined. My hosts could not explain of holes, machines a flat, taps a hole, or ates with a number of rules of thumb. how these parts are cast. does some other simple operation on For example, to do cost estimating, his New products include laser heat an engine block or other similar item. first test is to count the parts and mul- treating and machines that grind ceramFiguring out what each module should tiply by ¥10,000. To determine if his ics, especially for bearings. Kyocera is a do and wiring up the controls are the designers will be over- or under-loaded, big customer. Their product has half major challenges. Such a machine takes he multiplies the predicted sales by the market because it is priced aggresabout a year to design and build. A set 10% and compares to his shop's salaries. sively, the result of its being just a of four almost ready for delivery to He can support Y6,000M of design effort modified NC machining center with Nissan were scheduled for customer per year, so annual sales over Y60,000M some novel tool attachments. buyoff the day we visited. Our hosts will force him to go to outside contracand I together estimated that about tors or overtime.
Factory Floor Tour 112 man-months of design effort were There are few systematic design involved in this system, which had over techniques in use. Tolerances are deter- This factory makes more specials 75 modules. Half these man-months mined by reference to previous “expe- and fewer ordinary machines. The “mass were obtained from subcontractors, as rienced parts.” Only recently have they production” work is at another factory were many of the machines' more stan- begun to adopt the idea of the “error where five machines a day come out dard elements like conveyors.
budget" for allocating tolerances with "right on schedule." Assembly line repetitive working conditions drive many priority in this unstructured environ- been slowing for years) because mainof the employees away. They can easily ment and of determining the overall frames are used in large communicaget jobs nearby since the whole area's behavior of the system by choosing which tion systems and commercial systems manufacturing economy is growing. parts to load into it. This system resem- like those for banks, automatic teller
Regardless of the kind of work, the bles in both physical layout and human machines (ATMs), and airlines. Abiko plant is fully automated in involvement a system build by Sunstrand Fujitsu's introductory video stressed machining and partly so in material for Ingersoll-Rand almost 20 years ago. use of computers to do things “the way handling. No automatic guided vehi- It was very efficient and easy to schedule. people do.” Applications under study cles (AGVs) are visible as at Yamazaki,
include neural net machines for machine however. There are several FMS lines, Future CAD Needs
translation and robot controls. Other most about 8 years old. Rescheduling
work is on x-ray lithography at 0.2 micron them is a problem, since it takes halfan Igot in English only part of the long line width, coherent light digital optihour. It appeared that the larger sys- list Otani gave in Japanese in response cal communication, optical memories, tems were not very efficient and not to this question. He sees the potential ceramic circuit boards with approxiwell utilized. At least one suffered of solid modeling as an input to CAE mately 50 layers (probably for use with from an old design error, namely, a such as static and dynamic analysis. multichip modules although that tersingle straight line parts transfer vehicle The main use for this is to shorten the
The main use for this is to shorten the minology was not used), and real time with one-part carrying capacity and no 6 months of performance validation image processing. buffer positions at the machines. Moving that is currently required of new designs. Fujii's people mostly do electronic parts around is like solving the canni. He also wants to be able to design one- design, so most of their CAD/CAM is bals and missionaries problem. This of-a-kind machines, such as transfer in that area. He showed only mild intererror was discovered at Caterpillar in lines, more efficiently than he can now. est in our group's work in feature-based the mid-1970s. The more successful These are not money-makers in the design and modeling of mechanical H-S FMS have more buffers at each current environment and use up a dis- assembly processes. He may have felt machine and a continuous flow con- proportionate share of his engineers. that I was proposing to replace the veyor system with parts circulating past More extensive use of standard ele- designer with a computer program, the machines all the time. The advan- ments would be a likely outcome. He whereas I was proposing a tool to help tage of this setup was also recognized calls this “automated design.”
the designer sort through many possiin the mid-1970s.
bilities. His final judgment was diploSignificantly, H-S plans to junk the FUJITSU TECHNICAL
matic: when solid modelers become oldest and least efficient FMS and CENTER AND FUJITSU
easier to use, then our work will be of replace them with stand-alone com- LABORATORIES, KAWASAKI great interest. However, he stressed puterized numerical control (CNC)
that design is very complex and the machines. It may arrange some of these 22 August 1991
designer considers comprehensively all into part-type-specific cells. One
the factors at once. Such a process cannot L-shaped cell for spindles has already Background
be given to a computer. been tried successfully.
The most interesting system would Our hosts were Mr. Fujii, General Design for Assembly (DFA) interest Prof. Okino of Kyoto Univer- Manager of Corporate Manufacturing sity. It is called Holonic and won a prize Systems Development, and his assis- Mr. Miyazawa gave a short presenfor H-S in 1988. This system consists of tant, Mr. Asada.
tation about Fujitsu's home-grown DFA a number of machine cells and contin- Fujitsu is a large diversified com- methodology. It employs a PC to take uously circulating parts. There is no pany focussing mostly on computers in data from the designer about the overall schedule. Instead, parts are and electronics. Its major products are assembly, including part descriptions, marked with bar codes that permit mainframe and supercomputers, as well part mate types, and so on. I remarked each machine to recognize the part as laptops, high performance disk drives, that all of this data would be in the and its required remaining work. A array processors, low noise super tran
feature-based solid modelers of the part can go to the next machine that is sistors for use in space telescopes, and future. capable of doing the work it needs next. soon. Unlike IBM and Hitachi, Fujitsu
The DFA method has four characThe operators have “human involve- has not suffered a drop in mainframe teristics: to provide a quantitative estiment” consistingof giving certain parts computer sales (although sales have mate of ease of assembly, to give a rough estimate of (manual) assembly his opinion is merely to make the prod- computer tools. The ones in use at time, to improve the design, and to be uct easier to make, shorten develop- Fujitsu are 3D structural design using easy to use. Its objectives are to reduce ment time, reduce costs, and reduce SDRC's I-DEAS solid modeler, mold the cost of the product, to shorten the the number of prototypes needed. He flow software, and stereolithography. development time, and to provide an specifically feels that CE does not The latter is implemented on a machine easy way to communicate to factory (cannot?) address product quality and called SOUP (Solid Object U-V Laser personnel about assembly problems. function. The design for assembly Plotter) made by Mitsubishi. Input to
In these ways the method remark- manufacturing (DFA/DFM) aspect of this machine is via I-DEAS. ably resembles Hitachi's, which they CE makes it hard enough already for (Stereolithography is a technique know about but have not bought. My the designers, and he worries that they for making rapid prototypes of solid report is sketchy since, like Hitachi, cannot remember everything. So he is parts. It functions by forming a part they do not want to give out all the in favor of computer tools to help them. layer by layer, either from a liquid or a details.
powder. Laser scanning is commonly To use the method, the designer Aside
used to form each layer. In the SOUP makes an assembly flow chart (pre
and similar machines, the laser scans sumably an assembly sequence) and Several companies have their own the top free surface of an ultraviolet types in data about the parts and mates. DFA. Each shares basic properties but (UV) curing liquid plastic. When the The computer scores 30 different char- each has interesting additional elements layer has formed, the object is lowered acteristics, such as direction of assem- that reflect characteristics of the prod- about half a mm and the process is bly motion, part size in relation to direc- ucts or of the underlying philosophy of repeated. Parts with re-entrant features tion, the method of attachment, whether what DFA ought to be able to accom- can be made as long as a hole is left to the part is flexible, how many moves or plish. Nippondenso is the broadest, but remove the uncured plastic. The layeractions are required to affix the part, Fujitsu is the most careful in identify- ing gives the part a stepped outer surand so on.
ing different kinds of parts and think- face finish just like "jaggies" on comUnlike Hitachi, Fujitsu's method ing that they should be scored differ- puter drawings. These are usually includes some interesting twists on part ently. No basis is given by any company handsanded off. characteristics. Parts are designated as for the different emphases; they just Stereolithography is capable of "main,” “subsidiary,” and “fastener," appear obvious to the developers. creating the shape of a part, and methamong others. Fastening methods are
ods exist for using it to make metal designated as single action, snap action, New Method for Mold Design molds from which parts of the correct and so on.
Linked to Rapid Prototyping material and strength can also be made. The scoring is done separately for
The method permits people to quickly each class of part and a profile is made The brief presentation was in "get their hands on the part,” including up. A perfect score would be a product Japanese with interpretation. The communicating to factory or subconwith one side, one direction, single action overheads were also in Japanese, so I tractor personnel in ways that drawassembly of rigid, little parts. The score could not follow too well. The labora- ings cannot.) is presented either as a spider chart or tory tour afterwards was better.
Fujitsu uses the SOUP output to a profile and is accompanied by the The old method of mold design was make a mold from silicone rubber. scores ofrecent similar products. From a series of steps, each with an approval Accurately made silicone inserts are these charts, obvious ways the product before the next one could start. The used to create some mating surfaces, is worse stick out clearly: too many steps are, approximately: initial part with the result that mating pairs of awkward fastening methods, too much design, design verification, mold design, molded ABS parts assemble with sureffort to put in subsidiary parts, not and mold feature evaluation. Each step prisingly close-fitting mates. enough simple actions, and so on. Using was conducted by an expert. Too many A companion rapid prototyping this method, the estimated manual final people had veto power, too many steps facility is a Fanuc NC machine, which is assembly time of a laptop computer were done sequentially, etc.
programmed from a ProEngineer was reduced from 6 minutes 20 sec- The new method uses the same system. (ProEngineer was chosen onds to 5 minutes 10 seconds.
experts but launches them simultane- specifically for its easy user interface, After this presentation, Fujii ously. Significantly, it was stated that suitable for the technician they have remarked that DFA is part of Concur- merely doing this does not save any assigned to operate this facility.) This rent Engineering (CE), whose job in time. In addition, the team needs some machine makes smooth profile molds