Lapas attēli
PDF
ePub
[blocks in formation]

The cycle from sketch to solid prototype part is 3 days, most of it taken up Background by 24 hour per day operation of the SOUP machine itself.

Our hosts were Mr. Okuda, General Manager of the Production EngiOther Uses for CAD/CAE/CAM neering Headquarters (HQ), and Mr.

These are fairly routine, compared to the rapid prototyping. Typical FEM is done using NISA 2, commercial software from the United States. Vibration, deformation, and stress analyses are done on a Fujitsu workstation. Fujitsu also uses FEM5, an in-house product, and ABAQUS. The preprocessor is called Concept Station, made by Unigraphics. CADAM is used for ordinary engineering drafting.

Another in-house program is coincidentally called ICAD, but has nothing to do with the U.S. company of that name. This in-house ICAD is for mechanical design. They did not show it to me and said that it needs improvement. They are more proud of ICAD's circuit design capability, but did not show that either.

Other Tour Sites

Mr. Uchiyama hosted us on a tour of Fujitsu Laboratory's robotics activities. These included a simulator for zero-gravity robot operation, robot off line programming, precision class 10 cleanroom robots for disk drive assembly, and a space mechatronics laboratory. The latter is making experimental modules for Space Shuttle or Space Lab activities in the late 1990s. Fujitsu has about 100 of its FAROT precision PUMA-style robots in use, mostly for precision assembly and circuit board probing. These robots have about 30 micron repeatability. Uchiyama was the designer of the first prototype for this robot over 10 years ago.

Okamoto, the Senior Assistant Manager of the CAD-CAM [computer-aided design/manufacturing] Center, a very design/manufacturing] Center, a very astute person who may be ahead of the astute person who may be ahead of the rest of the company's thinking on many issues.

Nikon is a diversified optical company with 7,200 employees and $2.2B annual sales (at ¥150/S). The main products are cameras (42% of sales), instruments (microscopes, theodolites, surveyors' telescopes - 7%), industrial equipment (mostly for making semiconductors and recently "very busy" 32%), and eyeglasses (8.5%). Smaller divisions have less than 4% each and make electronic imaging cameras, magneto-optical storage disks, and magneto-optical storage disks, and bioengineering products, such as glasscoated metal dental implants.

The Production Engineering HQ is a corporate-level activity that makes production equipment for all the divisions. While equipment design is done at Nishi-Ohi, some of the equipment manufacture is done at the plants where it is used. Curiously, Okuda does not report directly to Corporate HQ but instead to the Executive Director (VP) of the Industrial Equipment Division, who reports to the President. There is one other Executive Director, for Consumer Products, the only other main division. Okuda says that reporting in this asymmetric way causes no problems. In any case, Nikon is generally typical of Japanese companies in making its own manufacturing equipment in a corporate-level department.

Okamoto stressed several times that Nikon is a conservative company that serves a conservative and highly professional clientele.

Product Design Methodology for Cameras

Camera technology is changing rapidly, with microelectronics being the main driver. From a mechanical viewpoint, the main change is to styling the exterior with arbitrary sculptured surfaces. This has thrown the camera companies right into the car design arena without the car companies' years of gradual buildup of experience, attitudes, and software development

resources.

Altogether Nikon has about 100 camera designers, split 70:30 mechanical:electronic. (Since the optics are designed elsewhere I do not have a good view of that process. See the Ricoh report for an optics story.) There are so many design projects going on and so little time that the staff is stretched thin.

A top-of-the-line camera has about 1,000 mechanical parts. The F4 took 3 years to design and occupied 10 to 15 designers full time: 5 mechanical, 5 electrical, 1 or 2 optical, and 1 software. Lower level cameras may have 600 parts. Totally new cameras are designed every 2 to 3 years, with small changes all the time on 1-year cycles.

Another important driver of camera design is the almost vain attempt to keep the weight from growing as optics and other features are added. The main casualty of this effort has been the die cast aluminum body, which has been replaced by a set of precision molded plastic parts, usually 20% glass-filled acrylonitrile butadiene styrene (ABS). (The same material is the prime one for printer parts. See Seiko-Epson report.) Stress analysis would thus seem to be a prime concern, but neither Nikon nor Ricoh has enough solid modeling capability to use such techniques on a regular basis.

A third driver is price competition. Few Japanese companies can assemble cameras domestically. Assembly is basically manual and labor costs are

high. Nikon, at the top of the price chain, can assemble in Japan, but it can (and must) fabricate only the critical plastic parts in-house. For the others, it makes use of (to us) a dismaying array of mom and pop shops scattered around town and country, working in poor conditions and subject to dismissal at any time. For this reason, among others, the influence of production engineers in the design process is not typical for Japanese companies, since they are not representing in-house people and it is difficult to predict the capabilities of such suppliers. Incoming inspection is a big effort.

The last driver is the design of a very complex and oddly shaped flexible printed circuit that holds several custom integrated circuits (ICs) and lots of little components. This must snake around to link several locations on several of the functional elements inside the camera (film drive, range finder, autofocus, autoexposure, liquid crystal display (LCD), flash, and so on.) Design of this flex circuit is tedious and not well supported by CAD.

The camera design process (same at Ricoh) begins with "industrial design" of the exterior. As many as 10 outer shell prototypes are made by hand from wood or styrofoam. Each such prototype takes a week to carve and finish. Design is difficult because handfeel is important and peoples' hands are such different sizes. Especially U.S. and Japanese people are quite different in body size. No solid modeling is used for these explorations because it is “too slow." At least, the car company engineers tell Okamoto that! He also worries that no solid modeler could hold all the data for a camera's interior. Again, this mirrors car company opinion.

Stereolithography (SL) is not used. They have heard of SOUP (see Fujitsu report). For them, SOUP's 3 days is also too slow. (Yet they wait a week now??) But other SL machines give results in 2 to 3 hours. I was told this at Ricoh and shown the parts. They have

rougher outsides than SOUP's, which is to be expected. Perhaps camera and phone designers must have the fine detail and smooth finish that take 3 days to create.

An important issue is weight and balance, which must be judged when outer shape is designed. Current and past data are both lacking on internal part weight and especially location of centers of gravity, a major shortcoming according to Okamoto. But the decision must be made anyway, usually by top management on a date set in advance on the master schedule.

When the outer shell is approved, a solid (actually surface) model is made and all subsequent engineering and preproduction prototypes are designed using CAD. "We pay more attention to feature lines than the car companies do."

The interior of a camera is divided into "blocks." We would call these subassemblies although each likely has several subassemblies of its own. The main blocks are the mirror box (for single lens reflex (SLR) cameras), the shutter block, the rewind spool and fork, and the takeup mechanism. Autofocus motor and gears are inside the detachable lenses of Nikon cameras. The size and shape of these blocks drive the shape of the interior of the camera. Dividing up the detailed functions of the camera among these blocks is "difficult" and no systematic procedure seems to exist. However, there is not too much room for change, and much of the arrangement of past cameras is copied in new ones. Each block is usually the responsibility of one designer, who rotates to another block on the next design cycle. Thus there are no "shutter gurus" and so on.

The main factors driving new block arrangements and shapes are elimination of the aluminum frame and radically new camera shapes, such as ones with vertical formats. (Videocameras in horizontal format have appeared as well.)

It is surprising to see little or no solid modeling in use to attack these problems or those of checking that all the parts will fit. Okamoto sees the potential very clearly but says that the camera companies are debating how to best approach the problem. In the meantime they make use of the ability of two-dimensional (2D) CAD to generate lots of cross section drawings, which they compare by eye with automatically generated sections of the solid model of the shell.

Once the outer shell is approved, the interior is designed and one or two engineering prototypes are built. All such design is on 2D CAD. About halfway through this cycle a co-located production tryout team joins the process. They will build preproduction prototypes before outside contractors and assemblers are launched. The designers visit the factory where the current models are made to learn about any problems they should avoid. Production people attend design reviews near the end of the design process but their comments are usually applied to the next camera. New designs are coming along all the time with so much overlap and so little major change that a phase shift of one design does not matter in applying the factory comments, or so says Mr. Sasagaki, the assistant manager of the camera design department. Major disasters are obviously prevented, and the designers learn to anticipate most other problems. Whether this approach would work at Canon with its more radical changes and faster design cycle is not clear.

They have heard of Concurrent Engineering but associate it with solid modeling, not with a different kind of design process.

Use of Computers in the Design Process

Most of the factories and other design divisions use CADAM or CATIA, running off a total of six large IBM

mainframes (3090, 4341, or 9370). All the mainframes are linked by Ethernet and all the terminals in a factory or design center are on some kind of local area network (LAN).

Nikon uses Computervision (CV) for most of its camera CAD and CAM. Newer software such as a few seats with SDRC's I-DEAS level V operate on DEC or HP workstations. "Solid modeling" of the exterior parts is actually surface modeling with NURBS (nonuniform rational B-splines). Realistic shaded images are possible, though not as good as Toyota's.

These sculptured shapes are made by injection molding but, although Moldflow software is available in-house, it is not in use. Okamoto hears that it is hard to use. (Like many people in many companies, he could use better information than he can get.) The data for molded parts are converted from CV to CATIA in order that NC can be used to make the molds. The recently established Mito plant does all the mold design, fabrication, and part production for the company. CNC and direct numerical control (DNC) are used in the mold machining process.

CV's 2D drafting is used to design all the other parts, but the mom and pop shops do not have NC and usually employ hand tools.

Part of Okamoto's job is to modify commercial CAD software for internal use. This takes three forms: improving the user interfaces, creating data translation code, and integrating in-house special computer-aided engineering (CAE) software. An example of the latter is code written by Sasagaki to permit shutter kinematic analysis. It is not graphic but merely the engineering calculations. No animation is supported. They tried using IGES to transfer files but too much useless detail came along and the result was too big.

The flex circuits are designed using CV software. Okamoto says that CV is especially good at this aspect, but he

admits that it does not support auto-
admits that it does not support auto-
matic routing or checking of node lists.
These are pretty basic steps in complex
circuit design that are supported fully
or partly by other commercial software,
but here the designer does it all by eye.
All the software does is smooth the Other CAD/CAM
lines and check design rules like line
width and spacing. The example circuit
he showed had two layers and a very
complex curved perimeter shape, plus
about 100 parts including 6 ICs. It took
"a very expert person" 2 months to
design it but “visitors usually guess it
took a year."

engineers do, and complain that the
designers don't. It is a manual method,
not connected to CAD, though he would
like to make this connection. “It will
save enormous time and effort."

The CV software then computes a good route for the NC drill software for putting the holes in the circuit. For 332 holes, this step took about 5 minutes. Lens design is done in another department using Nikon's own ray tracing software operating on a UNISYS computer. In Okamoto's department, a simple geometric ray tracer is on hand to discover design problems like hoods that interfere with the image. He is equipped to handle customers' inquiries on such things by phone in real time.

Okamoto has some software research and development (R&D) responsibilities as well and would like to try many things that the company is reluctant to support or which he fears the designers would not use. Among these are use of solid modeling to check interferences and to support various kinds of design for manufacture (DFM) and design for assembly (DFA). Assembly planning is especially interesting but no one else seems to support him. He also wants to try artificial intelligence (AI) for helping the circuit board routing process and for converting 2D drawings into solid models. He is dissatisfied with current Al research on these topics since none of the methods appear fast enough.

The company has its own assembleability evaluation method (AEM), similar to Hitachi's, but the camera designers do not use it. The production

In the software R&D laboratory I was shown

• CAT [computer-aided testing; when mechanical engineers in Japan use this term, they mean computer-aided surface inspection of mechanical parts using a coordinate measuring machine (CMM)]

• CAD of eyeglass frames

• Conversion of 2D drawings to threedimensional (3D) models

• Reverse engineering of arbitrary shapes (scan part and create computer model)

CAT has been improved by providing offline programming of the CMMs that Nikon sells. The goal was to create a computer screen imitation of the CMM's control console plus an easy graphical user interface. CAD data about the part to be measured are used to drive the software. Animation of the machine's actions and pictorial images of available measuring probes are also available. He hopes that this software will enhance sales of CMMs, which are slow.

CAD of eyeglass frames is done on a color Silicon Graphics console using a drawing program similar to many now available on the Macintosh. A very skillful woman demonstrated this. It permits lens shapes and frame shapes to be designed. Both can be colored realistically, and nice advertising-style drawings can be made. NC data for cutting molds for plastic frames are also generated. Several features of the process are dimension driven, but it

was not clear if any constraints can be designers. Overall, he has a broad view imposed. of the possibilities for integrating the product realization process.

Conversion of 2D models to 3D is done interactively. The user can apply

some primitive feature-based design RICOH TOKYO OFFICES while selecting which portions of a

drawing to convert. The process carries 29 August 1991
along some extra lines which the user

removes with mouse clicks. This would Background
be a useful companion to the CAT
software but it is still under development.

Reverse engineering is done by using the CMM to scan a part, such as a clay model for a camera. (Ricoh uses a laser.) They have no confidence that any future CAD will permit direct camera design

on a screen.

Future Needs and Problems

Okamoto is worried that Japan is falling behind. "The U.S. is 10 years ahead in use of solid models," he says. But he knows that faster and larger product variations are coming, and the design cycle must be shortened. The designers do not see the wave about to break over them. Just introducing solid modelers will not be the answer, because the entire design process must be thought through and restructured. There must be more standardization, reuse of past designs and data, and more systematic methods. These changes cannot be imposed from above because Japan is too bottom-up oriented, meaning that the rank and file must get behind the effort first.

As if to underscore the problem, Sasagaki says that standardization will hurt designers' creativity.

Specific future needs Okamoto sees are use of solid modeling and featurebased design (FBD) to drive a true link between CAD and production engineering, plus better ways to engineer tolerances. He comes out of the CMM world and is thus highly sensitive to tolerances and inspection. His main hope for FBD is that it will create an engineer's interface to solid modelers and make them more acceptable to

My hosts were Mr. Ageishi, Director of the IMS Department, Dr. Toriya, a CAD expert, Mr. Yazawa, a lens a CAD expert, Mr. Yazawa, a lens designer, and Mr. Watanabe of the Camera Product Planning Department.

Ricoh, like Nikon, is a diversified optically oriented company with the great difference that it leans toward computing and communication applications such as copiers and fax machines, with only a small presence in the camera market. It has 34,000 employees, of whom only about 13,000 are in Japan, and sales of about $5.5B (at ¥132). Office automation, including integrated document preparation, is the main product line. Ricoh has 20% of the product line. Ricoh has 20% of the domestic fax market and 38% of copiers. However, my visit followed the theme of cameras, making comparisons with Nikon interesting. Copiers (see below) were the subject of a brief discussion.

Ricoh also makes laser printers and the laser engines that go into them. Ricoh also has a laser printer page description language software called Ricoh-Script.

A video showed off the many technologies that Ricoh supports, including R&D on neuro-chips (one chip with 256 small neuro-chips on it), voice recognition as a user interface (any person, limited vocabulary, one word at a time), digital signal processing chips sold to Nintendo for voice synthesis, magneto-optical storage disks, conductive polymers (described by a lady chemist), and so on. The video also hinted at some problems which my hosts confirmed: the company was a pioneer in offshore camera manufacture but has had difficulty finding qualified

employees and parts suppliers overseas. Ten years ago Taiwan could not assemble complex cameras for them; now it can but only after lots of on-site training and education. Also, a "restructuring" program has been underway for about a year. They are trying to improve their design methodology by overlapping design tasks, improving communication between designers and production engineers, and increasing the use of assembly automation. But total reorganization is needed first, and the project is still young.

Ricoh has about 4,000 engineers among its Japanese employees. The company worries about hiring new ones. Each of the last 2 years 300 were located, but next year looks difficult.

The visit had five parts: a general discussion of Concurrent Engineering and how copiers are designed and made, an explanation of their own solid modeling product, industrial design of camera exteriors, design of optical trains, and mechanical design of camera insides. The most interesting aspect was their first attempt to write software that will analyze tolerances in multi-element lenses.

Concurrent Engineering and Copiers

Concurrent Engineering (CE) to them means intense communication between product designers and production engineers, including some overlapping of their tasks. Ricoh has done this for years without calling it that. The main method is meetings and design reviews. In view of this, the question arises: what is the restructuring mentioned in the video? Apparently the company is aiming to reduce development time by another 50% and cost by 30% by 1993. Copiers will be attacked first. A joint team of 400 people is working this problem. The methods are cost analyses, planning of design processes, technology improvements, and more overlapping of product and

production engineering. A major problem is that each division of the company has a different culture in its design methods, driven by the different production volumes and rates of product change in each. Only 15% of employees change division each year, so some differences have built up.

They (like many Japanese companies) wonder how to accomplish CE more efficiently. They think the phone and meetings are just fine and wonder how they could afford to give every engineer a PC. Communication is first, then technology. Moreover, they spoke of having to further reduce the conservatism of managers and interdepartmental restrictions. If these institutional barriers can be reduced, then new design technologies like feature-based design (FBD), CAD with product and process analyses, and rapid prototyping (like SOUP) will fit right in.

Copier Design

A new copier takes from 1.5 to as much as 5 years to design, depending on the complexity of the new functions and copying technology being attempted. Apparently 2.5 years is typical. On big projects the specification keeps changing because the market changes rapidly at the high end. New toners and new scanning engines are designed in parallel. There are about 550 design engineers supported by about 55 production engineers. A new copier will occupy 30 designers, while a modification of an existing design may take 10.

These numbers are typical of those obtained from most Japanese companies I visited except car makers: design teams are small and manageable, ranging from 10 to 30 (versus 500 at car companies).

Design-Base - Ricoh's Solid
Modeler Product

Toriya explained this product, which he apparently spearheaded. He works for the Software Division, whose director

is another lady, Dr. Kunii, whose Ph.D.
is in database management from the
University of Texas. She has pushed
University of Texas. She has pushed
new database management products
and Toriya hopes to combine them with
Design-Base (D-B).

Unfortunately, D-B is not widely used inside Ricoh. I was told that it is so recent that there would be difficulty dislodging the considerable investment in commercial CAD, which was purchased for its surface modeling ability before D-B was available. Also, a lot of camera-specific CAE has been added to the existing CAD and they claimed it would be too hard to convert it all. Now, however, Ricoh understands the importance of solid modeling technology, and D-B is expected to be used more and more.

Design-Base is a UNIX-compatible
system written in C that is similar in
many respects to ACIS from the U.S.
company Spatial Technologies. Meth-
company Spatial Technologies. Meth-
ods and marketing are similar: its archi-
tecture is open and its engine is avail-
able for licensing to companies that
want to use it as part of their own CAD/
CAM work. “ACIS is our main com-
petitor in America." D-B is the only Industrial Design Center
UNIX-based solid modeler available

in Japan, which says something. It has
all the usual features: rendering, boolean
operations, free surface modeling using
Gregory patches, sections, filleting, and
so on.

It also has several novel and inter-
esting user interfaces that permit very
rapid construction of shapes from simple
primitives. These were demonstrated
on a Sony NEWS workstation, where
performance was impressively fast. (As
at every company visited, the CAD
demonstrator was a very skillful lady.)
APC version exists called D-B Jr. The
new user interface (UI) was written in
Motif, so it is easily ported to other
machines. It includes ability to call
forth standard primitive shapes and to
detach surfaces from them and slide
them easily along XYZ coordinate
directions. Edges can be shifted around
and can be broken up and made into
curves. Such operations are entirely
artistic in the sense that no "surveying,"
coordinate values, or dimensions are
involved. So, while it is easy, intuitive,
and fast, it does not really support
engineering. Toriya agrees that feature-
based design would not only help but
might be easy to add to his existing
methods.

Thirty researchers and engineers support development and enhancement of D-B, while 30 more support database management system (DBMS) work.

D-B was demonstrated here showing aesthetic design of a new copier and a children's educational toy. The final colors of the toy were chosen by a consultant who used color printouts from solid models made on D-B. However, most of the work is done with other commercial software whose name I did not catch.

Most of the discussion here centered on camera external design. This is the first step in making a new camera. Like car design, the first step is a series of hand color sketches, followed by several clay models. Shape is the issue, not size or weight. The final clay is digitized by a laser, a CAD drawing is made, and dimensions are added to create scale and main radii. For this last, a plastics engineer adds his comments, but no molding analysis software is ever used in the entire process. From this drawing a very realistic mockup is hand made and the designers and marketing people critique it.

In a recent case shown to me (RZ 800) the market was ladies, and a major change in the shape was made, eliminating a sharp edge in front and making the whole camera softer looking. The above process (sketch, clay, digitize...) is then repeated. Data from the second version are sent to the mechanical engineers so that they can begin trying to fit their parts inside.

« iepriekšējāTurpināt »