Lapas attēli

approval of issuing partial information 4. Link CAD data to CAM.

labor shortages have forced automain an overlapping job environment and

tion of manufacturing, engineer shorthow to organize feedback and design This is a very pragmatic reduction of ages are forcing automation of design critiques that are based on partial last year's approach and seems driven processes. information.

by the needs of the HYPER 90 proAn essential element in Nakajima's gram. The more intellectual approach YAMAZAKI MAZAK TOKYO approach is to restructure the design outlined last year still seems to be under SALES OFFICE AND process to optimize the flow of infor- consideration but in the background. NAGOYA HEADQUARTERS mation. Last year at the Japan-U.S.A. Also under study or active developManufacturing Research Exchange, ment are

10 and 24 July 1991 Nakajima presented the following fourstep methodology for finding a good • A version of feature-based design Background sequence of design tasks:

that implements generic pieces of

geometry that the designer modi- My host for both visits was Mr. 1. Analyze the steps in the design- fies to suit his needs.

Awane, formerly of Hitachi PERL, now development process, find out how

director of Mazak's new Tsukuba R&D long each step takes and what their • A computer-aided process planning Center. Mazak is a privately held comprecedences are, and determine the system that uses such features and a

pany and no sales figures are available. critical path.

group technology segmentation of It has only 4,200 employees world-wide,

their parts plus knowledge of expert 2,600 in Japan. Of these, about 400 are 2. Subdivide the steps on the critical machinists. This is written in OPS designers, split evenly between elec

path and carefully determine infor- 83 and features easy entry of tabulartronic and mechanical. The main prodmation precedences for the newly process data by the designers them- ucts are NC lathes and milling machines, subdivided task steps.

selves, since IHI has no knowledge machining centers with ability to work

engineers. A major objective is to on several parts in series, plus com3. (Somehow) rearrange the sequence extend their current ability to plete FMS (flexible manufacturing

of these steps in view of the infor- sequence single machining cuts into systems) including large parts stockers mation available upstream that is the ability to plan the sequence of and transfer robots. About 7,000 to needed downstream.

major operating steps such as cut, 8,000 machines are made each year.

measure, heat treat, and so on. Choice The company is known for its 4. Determine the new length of the of the last operation to guarantee

advanced computer-controlled manucritical path and repeat steps 1-3 part quality is an important element facturing systems. Many of the partsuntil the best arrangement is found.

of such plans.

making facilities operate unmanned

overnight and require little tending Nakajima had no plans to use any • Automation of optimized blade during the day, a major reason why computer aids in this activity except for design, including direct data trans- output per employee is so high. Unlike conventional critical path analyses. fer between aerodynamics, stress/ most machine tool builders, especially

As of this year, the methodology thermal analyses, vibration, and in the United States, Mazak is both a presented by Mr. Chikata for use on preparation of process instructions pioneer in FMS and a large scale user the HYPER 90 project consists of and NC programs, with the goal of of them. Mazak is also responsible for

reducing blade design time from a number of important machine tool 1. Find critical path as above.

6 months to 2. Right now 80% of design innovations, such as the integral

the time goes to obtaining, translat- cutter spindle and drive motor. The 2. Analyze information flows as above. ing, and verifying blade shape data integral spindle runs faster and with

from aero design to structural less vibration since there are no gears 3. Begin ordering long lead time items analysis!

or belts between the spindle and its and designing tooling and fixturing

drive motor. The machine's accuracy is before design is finalized, using Stated only obliquely but clearly on thus greatly improved. smarter awareness of when critical their minds is the shortage of engineers Given this background, it is surprisinformation is available and guess- and the need to leverage the experi- ing to learn that Mazak is relatively ing the rest. ence of their senior people. Much as primitive in its use of CAD and CAE.

Design of new machines "takes a long time. This is a problem for our presi- determine the choice since any of the required span of time. Such problems dent." Like other Japanese companies, methods is technically capable of making are typically solved using complex math Mazak relies heavily on the experience the parts.

programming methods (Ref 11). of its senior employees, and some senior However, the choice has never been When the required number of tools managers do not trust computers in easy to make and is a subject of intense could not be made to fit in a machine's design roles. In this sense, Mazak is ongoing research at universities and tool storage racks, “tool management typical of conservative machine tool head-scratching at machine tool makers systems” were proposed. Since a part makers world-wide. Mazak recognizes and users. The FMS easily solved the could need 10 or 20 tools, the logistics the importance of CAE but is not satis- first-order problems faced by the alter- of tools far exceed the logistics of the fied with software currently available natives. NC made it possible to change parts themselves. Tool management from commercial sources.

from one part to another, which ordi- systems thus can cost more than they Mazak would like to sell more FMS nary mass production machine systems are intended to save. but finds customers reluctant. Initial cannot do. But stand-alone machine As far back as 1981 Mazak took a investments are large, and senior cus- tools, NC or not, are utilized only about completely different approach toward tomer executives must approve such 5% of the time since the machine bed is FMS for in-house use. It decided to purchases. So the Tokyo office has been used to set up fixtures and cutting tools make partially specialized FMS comset up with plush sitting rooms and a for the next part. The FMS solved this prising only three or four machines. fancy auditorium so that customers' by permitting fixtures and tools to be These machines were chosen to be executives can be wooed in style. A set up in a separate facility equipped identical and were capable of machinshowroom of machines used to be with good measuring tools.

ing a small set of parts, and perhaps enough, since the buyers of lower cost With the first-order problems solved, only a small fraction of the cuts those single machines were engineers who the FMS now faces second-order prob- parts needed. Scheduling and sequencjust drooled and bought.

lems that make the difference between ing problems essentially disappeared.

economically successful and unsuccess- A part entered the system, visited one Design of FMS

ful installations. The major issues are machine where it got all its work done,

keeping all the machines busy when a and left. It then visited another small An FMS is a group of 5 to 15 NC variety of parts is moving through the system and received more cuts. machine tools connected by a parts system and keeping the machines The range of required tools was conveyor. Each machine is equipped provided with all the different cutting limited by the “given tool method.” with many cutting tools and can change tools that such a range of parts needs. A That is, the part designers were given a tools automatically. Parts to be machined poorly designed system will have too stable of tools to use and told to design typically visit several of the machines many machines, some of which are idle, the parts so that set of tools would be in a system in order to receive all the

or too many parts waiting for a machine sufficient. For many FMS, this method required cuts. Each part may have widely to become available. Typical economic eliminated the tool logistics problem. differing cutting requirements, which criteria include cost per part, including For other situations, group technology the machines accommodate by using labor cost and payback of the initial was used to find a group of parts that their NC and tool-changing capabilities. investment and, in Japan, return per used 80% of the defined tool set for a

Mazak is such a heavy user of FMS square foot since land is so expensive. particular FMS, and a tool managethat it is much more sophisticated about The traditional approach in research ment system was used to provide the their proper use than most of its cus- and most FMS makers was to look for rest. Portions of this story appear in tomers. Deciding what range of prod- scheduling methods that would sequence Reference 12. uct types and production volumes is the parts into the system so that work- These two efficiencies have permitted most suitable for FMS has been the loads on the machines were balanced. Mazak to employ FMS very effectively main technical and sales challenge since Another approach was to develop FMS in-house without needing solutions to the idea was born at Cincinnati Milacron design software that would survey can- the long-term scheduling and tool and the University of Stuttgart in the didate sets of parts and decide the right management problems. However, the middle 1960s. The goal was to meet the number and mix of different kinds of discipline required to use the "given needs of diverse manufacturing that machines that should comprise the tool method” cannot be forced onto lies between mass production of almost system. The criteria were that all the customers, so the easy design and highly identical parts and low volume piece- required cuts could be made, there was efficient FMS operation achieved by work production of single individuals. space for all the required tools, and all Mazak is not always available to Economics and efficiency completely the parts could be processed in the customers.


changing, higher rpm, or a new spindle design is checked by Mazak for rigidity design.

and by the motor and bearing manufacThe company relies on CADAM The integral spindle-motor took five turers for heat and bearing stiffness and microCADAM, which runs on about engineers 2 years to design, during which and life. The rest is pretty routine. 72 total terminals, of which a small five prototypes were made. A similar Assembly of such a unit is driven by number are IBM 5080 graphics ter- one can now be turned out in a man- the need to install the rotor on the shaft minals. Most of the CAD work is draft- month (200 hours). The main prob- and achieve the bearing preload. Balancing, preparation of 2D drawings for lems in such units are dissipating the ing the rotor also is important and part manufacture, making shop floor heat from the motor and obtaining the must be done at the right time during instructions and user manuals, and so correct preload on the bearings. Motor assembly. Mazak's engineers did not on. There is essentially no CAE, in heat will cause the spindle rotor to feel that there was much room for flexspite ofobvious potential applications. distort, causing vibration and poor ibility or innovation in this assembly It is up to Awane in his new post to machining accuracy. Incorrect preload sequence. introduce CAE.

either causes poor accuracy or low Cost estimating of new or series bearing life.

designs is important because the marDesign Methodology for

Predictions of heat, distortion, vibra- ket and competitors often set prices. Machine Tools

tion, bearing wear, and so on are obvi- Their main cost estimating technique

ous candidates for CAE. However, is to consider four factors: cost of purThe rhythm of the machine tool Mazak does none of them. Motor heat chased parts, amount of material (usually industry is driven by the occurrence of and its effects are predicted by the motor measured by its weight), machining costs, the major machine tool shows. These manufacturer using its own CAE, and and assembly costs. The first two domoccur every 2 years in Tokyo, Chicago, Mazak merely designs a cooling system inate the final total and are calculated and Hannover, effectively meaning a to take away the heat. Bearing design is very carefully using past data. The other show every year or less. New machines done at Mazak and checked by the two fluctuate too much to be of great cannot be created that fast, and the bearing manufacturer, who has exten- use. CAD does not play much of a role typical cycle is about 2 years. Totally sive CAE for this purpose. This method in cost estimating except that CADAM new machines or technological inno- of farming out the hard parts is used can calculate volumes of parts easily. vations take longer. The shows are used often in Japan.

Cost of new entries in a series is deterto get customer input, look over the Advanced bearings such as magnetic mined by altering the data from the competition, and show your latest. In or ceramic are used sparingly or not at parent machine, which has been in addition, at Mazak, most designers have all. Magnetic bearings are the subject production for a while. visas up to date for most countries of some Mazak-sponsored university where the company sells and are ready research, while ceramic balls are used Future Computer to fly the moment a customer needs with steel races in some high speed Design Needs something. Customer input and minor applications. changes are thus the main forces gov- Integral motor-spindle design is now Mazak is unclear about what it needs erning typical design cycles.

so well understood that a routine has in the future. It has a mild anticomA major design strategy at Mazak, been established. The overall diameter puter frame of mind in the design departand probably most other companies, is of the rotor is decided by the size of ment and relies on its experienced “'series design,” meaning a series of chuck it will hold and the size of any people, some of whom have been with machines based on one principal design hole inside the spindle. Spindle speed, the company decades. One identified with many variations such as number motor horsepower, and required rigidity need is to calculate machine bed and of tool storage places, size of bed, and also contribute to sizing the shaft diam- column deflections under cutting loads so on. New designs are apparently not eter. The front bearings and their pre- and (at my suggestion) under thermal too hard to create within a series family. load are chosen almost straight from distortion. Other companies sell Information about existing machines, the handbook to support cutting loads. machines with built-in temperature their drawings and their performance, Required horsepower determines motor sensors and heaters that deliberately is kept in computer files. Five mechan- diameter, which determines the over- introduce compensating distortions to ical and one electrical engineer can all size and length of the spindle. The keep the machine accurate. Such comturn out a new lathe design in 2 years. rear bearing is chosen to hold up the panies must be ahead of Mazak in Improvements may include faster tool shaft, nothing else. This preliminary applying computers to design.

Mazak also would like an automatic modelon looks out for the conditions inserting, and so on. The peg and hole design system that would take in a set under which it could run; if they are modelons respond. I could not tell how of specifications for a machine tool satisfied, the modelon runs and depos- the assembled state was established and spit out a complete design. Such its its results in the common memory it for the purpose of identifying the free statements are heard at other companies is attached to. Modelons are also like faces. Perhaps the user constructed it and are not totally whimsical. How- methods in object-oriented program- as a way of posing the problem. ever, they are totally out of step with ming in the sense that they can send There is no agenda structure in this the state of the art. Prof. Kimura notes messages to each other, activate each search. Watabe had not heard the terms that many companies do not have clear other, etc.

“forward chaining” and “backward plans or coherent explanations of what A major feature of this structure is, chaining” before. (These are common they want. Visiting them every month as said above, that each modelon acts techniques in expert systems for solvfor a year is often not sufficient to autonomously. What the convergence ing the kind of problem he is working figure out what they are thinking, even problems of such a structure might be on.) Such searches commonly are not when the company is Toyota, whose Okino does not say.

very efficient since there is no gradient thinking would seem to be fairly Several topics are being studied under to follow and no metric to score how systematic.

this structure. I was shown software close one is to finding a solution.

demonstrations of two: a robot modelon PROF. NORIO OKINO,

that interacts with prismatic peg and 2. Hidden Line Removal: In this demo, KYOTO UNIVERSITY

hole modelons to determine where to three prismatic blocks intersect each

grasp the peg for the purposes of put other. Before the hidden lines are 26 July 1991

ting it in the hole and a hidden line removed, one cannot see what is what,

removal algorithm in which modelons even with three views, because several Background

representing each of three prismatic of the edges appear parallel and the

parts inform each other of the loca- front-back optical illusion interferes. In the 1960s and 1970s Prof. Okino tions of their vertices. We discussed a These facts make the demo more interdeveloped one of the first solid modelers third for which there was no demon- esting but of course have nothing to do based on constructive solid geometry, stration:a shaft modelon made of three with the intrinsic difficulty of the probcalled TIPS. He still works on CAD/ shape features, each of which is repre- lem. Each block has its own hidden line CAM and solid modeling but his main sented by generic feature modelons. removal method and seeks informainterest recently has been what he calls All the work is being done on Apollo tion from the other blocks concerning "bionic manufacturing.” Domain 10000 workstations.

where its lines enter or leave their Bionic manufacturing involves two

boundaries. Information is passed elements: self-governing behavior and 1. Robot Grasp Planning: This is the around this way for several minutes object-oriented hierarchical structures. recently completed work of Dr. Watabe. before a solution appears on the screen. Okino has constructed a conceptual In this system, there are several softmodel of computer-integrated manu- ware modelons representing the parts, Okino points out that speeding up facturing (CIM) that begins with all of the robot, and the environment. The the algorithm or competing with existsociety at the top and extends downward user requests a task, such as that the ing algorithms is not the objective, but by subdividing object classes until the

peg be put in the hole. The top level rather it is to understand modelons. lowest levels are reached somewhere modelon broadcasts this request to the In this regard, it is interesting to in a factory.

lower ones, all of which attempt to review the discussion we had about the Each level is composed of elements respond with a solution. Finding that shaft made of three shapes: a plain that have the same structure in princi- they cannot, they broadcast in turn to cylinder, a conical cylinder, and a ple. These are called “modelons.” Each any of their subelements whose threaded cylinder, all coaxial. Each shape modelon contains a common memory responses they need or could use, in a is supported by a generic feature that and a number of processes or methods divide and conquer approach. This is contains methods for drawing the shape, that operate using that memory. In repeated at lower levels recursively. calculating its mass, and a process plan addition, a number of lower level For example, the robot needs to know for how to make it. Presumably the modelons are attached to this memory. several sets of faces on the peg, such as process plan for the shaft is made by Modelons are independent actors, like parallel faces, parallel faces free-to- combining the process plans for the daemons or, in UNIX, processes. Each grasp before inserting, the same after three supporting features.

I asked a basic question that under- engine components and controls, radi- • High speed mass production lies the problems in all process plan- ators, dashboard displays, brake conning of this kind: what do you do in trol systems, and so on. It has manufac- • Mixed production (several varieties regions where the plans touch or inter- turing plants worldwide to satisfy many in one place) sect and presumably interact? That is, automotive manufacturers. Handling how does one compose process plans the wide diversity of product models • Just in Time production method from subplans? He agreed that this was and responding to the Just in Time a challenging question and replied that (JIT) ordering system philosophy have • Original equipment manufacturer perhaps one must declare the shaft itself heavily affected how Nippondenso (OEM) sales mainly to be the primitive element. A student designs and manufactures products. is beginning to work on the composi- Nippondenso is a high technology These conflicting characteristics tion problem.

company. Basic and applied research (especially variety, mass production, Okino's reply indicates that there cover materials, vacuum apparatus, and mixed production) have driven the are many problems yet to be solved by semiconductor fabrication methods, company into a variety of design and this approach. The value of having ceramics, robotics, vision systems, fac- automation methods. These are covered generic elements at the leaves of the tory automation software, simulation below. structures is clearly large and would systems for testing driver reactions, and give the approach considerable power. CAD/CAM. Major design thrusts over Automation and Product Requiring each specific shaft to have a the past 10 to 15 years include “manag. Development Techniques representation of its own is not effi- ing diversity,” designing new products cient. However, he is a software person faster, overlapping design tasks while An important feature of rather than an engineer and is pursuing managing risk, and dramatically reducing Nippondenso is an obvious long term the structural issues first.

the size and weight of products while enterprise-wide strategy for how to grow Prof. Kimura notes that object- increasing quality and performance. the company into a master of manufacoriented structures are good for some The company has 41,000 employees. turing products with these characteriskinds of data but not others, especially of these, about 5,000 are design engi- tics. Nippondenso has evolved a systhose that have strong interactions as neers and 1,500 are production engi- tematic approach to managing design well as, or instead of, hierarchical, neers. At the new R&D center where processes, designing carefully to supdecomposable structures. Mechanical the visit took place, there are currently port JIT operations, and developing design and manufacturing may not be 150 researchers. A major characteris- larger and larger systems of automaseparable enough to permit object- ticof Nippondenso, which I have noted tion. In the 1950s they had “spot” autooriented approaches to cover them in many previous reports dating back mation (what we would call islands); in completely.

to 1977, is that it makes much of its own the 1960s they had lines; in the 1970s

automation equipment and nearly all “areas,” meaning presumably several NIPPONDENSO

of its 3,500 robots. This commitment lines of the same type or several lines

to manufacturing equipment excellence connected; in the 1980s and early 1990s 29-30 July 1991

is one part of its commitment to man- "cube” or “totality.” Such increasing

ufacturing excellence in general. automation creates serious dangers for Background

The typical working year at a company whose customers switch

Nippondenso and in Japan generally is specifications, alter model production Our hosts for this visit were Mr. 2,200 hours, compared with 1,800 in volumes, demand instant response to Fukaya, General Manager of the Pro- the United States. The Government is orders, and increase variety of products. duction Engineering Department; Mr. trying to get this reduced to 2,000 by Nippondenso has only gradually Tsuchiya, Fukaya's R&D Manager; a 1994 and 1,800 by 1996. Each Japanese realized how deep the dangers can be young engineer Mr. Harada; and at

company is attempting to reach this and has instituted several procedures dinner Mr. Ito, the Executive Manag. target, facing various problems. for combatting them. These include ing Director of Production Engineering. Nippondenso characterizes its prod- simultaneous product-process develNippondenso Co. Ltd. is a former ucts as follows:

opment, a classification of levels of subsidiary of Toyota Motor Co. that

necessary flexibility in production, and makes a wide variety of automotive • Quality first

a classification of degree of innovation components, such as alternators, motors

in design projects. and actuators, air conditioning systems, • Wide range of product variety

« iepriekšējāTurpināt »