Transcript
Page 1: Designing robots for industrial environments

Mechanism and Mechme Theory, 1977. VoL 12, D;~. 403-412. Pergamon Press Printed in Great Britain

Designing Robots for Industrial Environments

J. F. Engelberger~

Abstract First a compendium of likely industrial environmental hazards is presented and discussed. Next, the influence of these constraints on a specific robot design is considered. The field reliability required by industry is determined and justified at 98% uptime. Analysis shows how one design achieved a 400 hr MTBF as well as 98% uptime. Some of the harsher environments are then depicted. The final thrust is a discourse on the economic constraint that overrides all other design decisions. A graphical cost/benefit chart is used to evaluate economic justification for any labor pay scale, at any daily activity.

Introduction AN INDUST~AL robot is a highly sophisticated automation component. It combines computer-like electronics and hardware with advanced electro-mechanical servos to make a free standing piece of equipment with great application flexibility. The abiding design philosophy is that a great deal of flexibility can be built into a single piece of automation which can then be offered at an attractive price, if the breadth of usage justifies mass production. Breadth of usage automatically implies ability to handle a wide spectrum of industrial environmental conditions.

When designing for a specific task such as a military equipment mission, it is usually possible to attach a precisely specified environmental envelope to the equipment. Such is not the case for an industrial robot. There is no quantitative specification available for all of the environmental conditions that a robot might encounter. Still, drawing upon 4 million hr of industrial robot experience, Unimation Inc. has come up with a quality assurance test program which seems to presage a high degree of success in most industrial environments encountered. The test procedure and the product design are empirical and both are still evolving as the range of activities of the Unimate expands.

Figure 1 lists the primary environmental factors that must be considered in robot design. Because conditions are qualitatively described, there may be merit to expanding upon the tabulation with even more qualitative discussion to provide the would-be robot designer with a "gut feeling" of the job at hand.

Ordinarily, a human worker is not required to function continuously in an ambient temperature over 120°F and therefore this is a reasonable maximum standard for an industrial robot. Both the human operator and the industrial robot are afforded cooling air if the workplace temperature exceeds 120°F. In some instances particular attention must be paid to radiant heating where the worker is the target of open furnaces, lehrs and hot parts in process. Radiation shields are sometimes used and a robot may expect to be provided with a curtain quench for its extremities.

There are not too many instances when an individual robot must endure severe vibration conditions. It is usually lugged to massive floor members and vibration from associated equipment is minimal. On the other hand, shock can be severe. Some hammer forge operations develop shock so severe that it can be felt in offices 300 yd away.

One of the most frustrating environmental conditions to plague a robot designer is electrical noise. Unimation Inc. has been unable to create a noise standard which would enable us to

tPresident, Unimation Inc., Shelter. Rock Lane. Danbury, CT 06810, U.S.A.

403

Page 2: Designing robots for industrial environments

404

Industrial Env~ronmenta.l Hazards

l . A m b i e n t Tempey&~ure : Up to 1 2 0 ° F w i t h o u t c o o l i n g a i r .

2. R a d i a n t H e a t t n g : S o u r c e t e m p e r a t u r e up to 2000°g.

3. Shock: ¢'xcurslona up to ~ inch, repetitions to Z/second.

t'lectrical nolle: l.me drop-outs, motor s u t r e i n g tra.,laient#; R F he&ring.

5. L i q u i d S p r a y s : Wate~- an d o t h e r c o o l a n t s , o f t en co rl"o s i r e .

F~nl , e l k VI , pOL'I: P 1 " o c e l ! chef t l . tca lm, i t e a r n ¢ I e s n m g .

7. P i r t i c ~ t h t t e Ma. t ter : Sit.rid, rnet&ll ic d u a t . ho t s Iatg.

F i r e k ExpLoeion R i s k : O p en fi&rr:e, e x p l o s i v e g a l k v ~ p o r m i x t u r e s .

Figure 1.

extrapolate in-house testing to noise immunity in the field. Any new design is put into the field in operations which we have found to be particularly "dirty" as regards electrical noise.

An electrical line dropout that might cause a computation error in a computer, means only a burst of "garbage" data output. For an industrial robot it might mean physical action, damaging to the robot or to the equipment with which it is associated. Noise insensitivity is crucial and without a clear definition of the noise environment, design becomes an iterative process cycling back and forth between the field and the laboratory.

There are lots of things that land on or diffuse through industrial robots in factories. These are often the same things that are designated as health hazards to human operators. The "black lung" human debility has its counterpart in the susceptibly designed industrial robot. Some examples will help make the point:

In investment casting, the atmosphere is heavily contaminated with alcohol.ammonia fumes which are highly injurious to any open switch contacts. The same operation also includes particulate contamination and gear trains and sliding bearings require absolute protection.

In one foundry application carbonized silica particles are continuously in the atmosphere and fall to the floor at the rate of 0.25 in. per day. The material is extremely abrasive and with any moisture at all corrosive to electronics.

Heat treatment processes often involve combination of high temperature and high humidity. This may be compounded by the salt solution which is used in heat treatment and which can build up on the industrial robot.

In some forging operations, die lubricant is applied copiously and the process impels the lubricant onto all surrounding equipment. This water-suspended graphite-based material builds up on the equipment and steam cleaning may be necessary daily. Steam cleaning itself is one of an industrial robot's environmental hazards.

Sparks fly continuously in resistance welding set ups and the fly metal particles will bond to open metal surfaces.

A robot does not stand aside when molten metal is shot into a casting machine and therefore it may be exposed to slag spurting out between die parting lines. The slag is hot, it must be endured and the robot must occasionally suffer hammer and chisel removal of the build up.

A man is inherently non-flammable and when admonished not to smoke he is non-igniting. When a robot stands in at a job where there is a continuing open flame, there must be protection against leakage of flammable servo oil that might cause a serious fire. In a job where an explosive atmosphere is created by the volatile carriers of paint for instance, the robot must not have any design element capable of creating a spark.

Page 3: Designing robots for industrial environments

405

Design Impact In designing industrial robots the primary influence is the nature of the jobs to be done.

Once manipulative power, sphere of influence, speed, strength and memory capacity have been established, environmental conditions can be brought to bear upon the design.

Some examples of design concessions to the environment follow: 1. Considering some of the hot and hostile places that a robot's hand must enter, it is

desirable to eliminate all electrics and serves from robot extremeties. 2. In many applications it is entirely proper to package the robot as a self-contained entity,

but there is an advantage to a design in which the electronics may be mounted separately. In extreme shock conditions it is convenient to be able to mount the control console on a shock absorbing pad and the remote location may be necessary to ensure life in a corrosive atmosphere. Going further, if the power supply of the robot can also be separated from the robot's arm then the arm may be introduced all alone into explosive atmospheres such as paint rooms.

3. With hot metal often flying about, it is important that all of the robot's skin be of non-flammable construction.

4. Wherever robot articulations are exposed in rotating or sliding joints, the joints should be booted to protect against abrasive dust collection.

5. The answer to fire hazard conditions comes with nonflammable fluid which hopefully is offered as an option because there is a significant cost disadvantage to its introduction.

6. If air is particularly dirty, air cooling may not be practical and the option to use water cooling should be made available. In any event, all cooling air should first go through filtration and enter enclosures to provide positive internal pressure.

7. Dust and dirt seem to be able to infiltrate the tightest crevices and therefore drive trains should use hardened gears and be pressurized to exclude contamination.

8. Robot logic design should be heavily protected from power lines spikes and noise pickup entering through any of the robot's communication links with surrounding equipment.

Reliability Required There are two concepts to be considered in a discussion of reliability: one is "Mean Time

Between Failure," MTBF, and the other is "downtime." Evidently, there is a relationship between the time interval between failures and the total amount of downtime; but, the correlation is not linear because there is the additional variable of time to repair. Thus, an otherwise satisfactory MTBF could result in unacceptable downtime if the time to repair is excessive.

An industrial robot is usually working in conjunction with another piece of productive machinery. When the industrial robot is down, and there is no provision for manual backup, the production machinery is also down. Industrial experience indicates that for most applications, uptime must exceed 97% to satisfy the users of'industrial robots. This "rule of thumb" is somewhat dependent upon the specific application. In an operation such as die casting, which inherently includes a lot of downtime, the process is not as sensitive to robot downtime. In a glass manufacturing plant, which is very akin to a continuous process, there may be need for uptime of 99.5% and if this is not attainable, provision must be made for backup labor which might be supplied by a backup relief human crew or by spare industrial robots.

If the design goal is a downtime of no more than 2%, and if we then conjecture that the robot manufacturer can always offer next-day service, and that all robots are working on a two-shift basis, the most likely downtime per incident becomes 8 hr. If the downtime per incident is 8 hr then a 400 hr Mean Time Between Failure must be the standard to hold overall downtime under 2%.

Potential Reliability of a Robot Given a rough standard on downtime, it is interesting to consider what one might expect on

a theoretical basis for a particular industrial robot design. In this discussion, the reference design will be that of the Unimate which has a multi-year design history. The design was evaluated against a reliability standard set through exhaustive study of all system components and their theoretical contribution to a total system failure rate.

Page 4: Designing robots for industrial environments

406

For Unimation Inc., the reliability feasibility evaluation was made by an independent organization, Bird Engineering Research Associates, Inc.. which specializes in reliability assessment. The Bird Report to Unimation Inc. concluded that Unimation's system design justified achieving an MTBF of 500 hr without incurring prohibitive costs in component design, system design, manufacturing, and quality control procedures.

The complete analysis was, of course, voluminous, but an appreciation of the method can be gleaned from an overview.

To predict the failure rates for all of the components in the Unimate, Bird relied upon notebooks which were prepared under U.S. Government contract to aid prediction of reliability of space vehicle systems. For electronic components, the Rome Air Development Center Notebook, TR-67-108, was used and for mechanical hydraulic components Bird used the U.S. Navy's Failure Rate Data (FARADA) Notebook. Both of these references were cross- correlated with other similar data banks.

Figure 2 is a tabulation of reliability feasibility for electronic/electrical elements. It will be noted in this table that the expected MTBF for parts only, is 1800hr. Because of the contribution of other (non-parts) the estimated MTBF drops to 1217 hr. The non-part failure rate is a system failure rate due to tolerance buildup, critical interface tolerances, customer abuse, unforeseen environmental problems, etc. This type of failure occurs in analog systems at a rate proportional to the complexity of the system and for the Unimate this was established by our consultants by comparing complexity to similar systems for which massive data exists. Evidently, the upper limit for an MTBF would be the 1800 hr due to parts only. This upper limit could be approached only at high dollar cost and therefore if the more modest 1217 hr goal were not sufficient, it would be necessary to examine other alternatives including going back to the original design concept.

Figure 3 is a table which integrates the electronic/electrical and mechanical/hydraulic failure rates to predict an overall attainable MTBF for the Unimate system.

Component or Element

I I II

Common Group Power Supply Shift R ~ t s t e r s MemOry Relay Tree Co.m01 Panel Memory Sequence Controi Home Ol~).onll (3 x 60) Sc ~.ne¢ Compara¢ot Sequence Control Opecaie E~ce~aal Wait Exxemai Counter-Demod (Common)

Sub ' , ta i

Pert Fellu~'e Rates (X 10 "6)

I I

54 103 L12 125

11 16 11 3

14 7

18 15

4

493

Coumt~r-Demod 4 Secvo Poem" Amp 16 Serve Switch & Dir . Stoce 13 Encoder Eleca'onlcs 29

Subtotat 62

Total FAec~omc/£1ectrlc ai Failure Rata (x 10-6):

Pacts Only SSS Other {non- i~"t s) 267

Ov~al l System 82Z

£1eC~omc/]~lec. MTBF (hrs): Plots Only 1800 Othq¢ (rim1 -s~nct s) 3745

Overall Syetem 12 t 7

Figure 2. Reliability feasibility of electronic/electrical elements used in unimate 2000 series design,

Page 5: Designing robots for industrial environments

407

Failure Class~ticat~on

P~t-*, Fat| tzxes Only: £1ectron~c/EleC't.r~ cel MecnamcaJ/Hydr eubc

Non-Pests Fe~h~res: )'lecMonlc/Electric~! MechamceZ./Hyclreu~c

S y s t e m }'allures: P~'~: s Only Non-Tolm'ence

Combined

Failure Rete (x 10 "6)

$88 673

267 4; 'S

MTBF (Hours)

i

1800 1485

3745 2100

1228 818 742 1350

1970

~',stlmeted l).ellmblLtty Fo ls ib i l i ty , Ummate 2000:. MTB7 * S00 Hours

Rgure 3. Unimate system reliability feasibility estimate.

Once the reliability feasibility has been established, the hard work really begins. Aiming for an overall MTBF of 400hr (about which more momentarily), Unimation Inc. set up a management system designed to bring individual components up to standard and assure statistically that the system, as shipped, will meet the overall goal. Figure 4 shows the reliability control points in the Unimate life cycle.

Since field experience is crucial to determining true reliability, the entire process of building toward this reliability includes placing machines in the field and feeding back the results of this experience into the reliability control system. In the case of the 2000 Series Unimate, the opening experience produced an lvlTBF of 145 hr and over the ensuing years of production, this MTBF was slowly brought up to 415 hr.

Typical Applications In Fig. 5 a Unimate services a die casting machine, suffering the perils of its stance at the die

parting line. Note the protective baffle on its boom cover and the skirt below. Figure 6 shows a Unimate transferring billets in and out of a rotary furnace for heating just

prior to a forging operation. In high energy forging, the billet is formed in one pass. Figure 7 shows the robot reentering

the press bed to extract the formed part. The chips created in machining are a hazard to robots. A boot around extension rods plus

tight seals on wrist gearing are essential to long life in the application shown in Fig. 8. Figure 9 shows the Unimate on a spot welding line. The robot must endure sparks, oil leaks

and cooling water spray.

Economics When one talks about harsh environments it is easy to conjure up sociological benefit

arguments. If a job includes a health hazard, perhaps the robot's cost is not critical. In the real world this is false reasoning. The standard for the cost of the robot has to be something less than whatever it is that will induce a human being to accept the working conditions. Barring actual legal action, manufacturers simply won't introduce automation that is not cost effective in comparison with human labor.

Figure 10 is a plot of return on investment for any number of hours worked per day, replacing labor in a range of different hourly rates. The plot presumes choice of a typical Unimate in comparison with human labor. Evidently, ff the robot cost were doubled, any return on investment drawn from this chart would be halved.

Page 6: Designing robots for industrial environments

408

I ¢ONFiC~UTION &' '=

I~CT]ON ;

® , , ,

*P~PEC T'I ON

, ,,~ ~

J

I_

,~ TEST I

'~ '~=~ I ] l - i - - r

I~TI~LJ~T I aN ,

Figure 4. Reliabil ity control points in the unimate life cycle.

Page 7: Designing robots for industrial environments

409

Figure 5.

Figure 6.

Page 8: Designing robots for industrial environments

"n

I:1

! ~

J

Page 9: Designing robots for industrial environments

411

Rgure g.

ioo

80

,40 .,

Lobar rate- $ I0.00 hr (including fringes) I

$ e .OO/nr

$6.00/hr

$ 4 .00 /hr

-20 .LAsr, umptions:

Figure 10.

8 I0 12 14 16 18 20 22 24

Average labor hr replaced per day

Robot investment $ ~5,000 AnnuOI depreciation 4,37,5 (8 Yr. St. Line) g# Working doys/yr 250 Robot' upkeep $.0.85/hr Robat mptoces one

operator

Conclu=ion

In conclusion it is possible to offer capsule advice to robot designers. If a robot can withstand the industrial environment, maintain an upfime of 98%, endure 10,000 hr of work between major overhauls, and provide a return on investment in excess of 25%, it is probably a viable product.

DIE KONSTHUKTIOE ~01~ ROBO~N FDH DIE ~W~LT~EDII~GUNG~ 11,: D~ INDUST~IE

J. F. Engelberger

~urzfauun~ - Zin IndusCrie-Roboter ist ein hochen~vickel~es Au~omatisieru~,os-

Page 10: Designing robots for industrial environments

412

element. Dutch die Verbindun~ yon computer~hnlichen ele~t~on£sc~en Bauteilen

und Hardware mit ~ortschrittllchen Servomechamlsmem stellt de~ Induat~ie-Robo-

re= elm unabh~n61ges Ge~t mit weit~eichenden Anwendun~sm5611chkelten da~. Al-

len Kcns~ruktlonen llegt dam l~£nzip zu~runde, elm einzelnes Automatisierungs-

6er~t mlt h~chstm~gIiche~ Vielselti~keit zu scha£fen. Es kann dann preisg~nsC£g

an6ebo~en we~den, wean eAn ausgedehntea Anwendun6sspekt~um die Massenp~odu~tlon

rechtfer~ig~. Vielseitig~eit £n der Anwendung beinhaltet zwangsl~ufig auch Fun~-

tlonafEhlgkeit unter den verschiedens~en industrleLlen Umweltsbedingun~en.

~id I z~hlt die wichtigsten Umweltsfaktoren auf, d£e be£ der Roboter~onst~u~-

tlon ber~cksic~ti6~ werden m~ssen. Sie we~den selbstve~st~ndlich e~s~ dann in

Bet~aoht ~ezogen, wenn die wlchtlgsten Arbeltsplatzamforderungen, wie z. B. Manl.

pulle=f~hAgEelt, Relch~eite, Geachwindlgkeit, ~-aft und Spelcher~apazit~t, fest-

gelegt sind.

Nachs~ehend elni~e ~elaplele fttw die Anpassung de~ Konstru~Clon an die Unfelt:

BeA de~ [ona~u~.O~ Z~ be~q~icht~;ende Fa~to~en

~. I n A n b e ~ a c h ~ dez T a t s a c h e , da~ d i e Rand des Robote~s o f t i n h e ~ e ~ und u n -

g~Anatige~ Um~ebun6 a~beiten :u~, ist es wGnachenswez~, die E~tremit~ten des

Robote~a ohne elektrlsche Bauelemente und Se~vomechamlsmen zu ~onst~uie~en.

2. FGr viele Anwendun~be~e£che ist es du_c~aus an~eb~acnt, den Hobote~ als

eine in sich Gescnlossene Einheit zu planen; a~lerdings hat eine Eonstru~-

tion, in dez die elek~'-oniscncn Haue!emente gesondert installiert we~den

k~nnen, auch ihze Vort~i~e. Bei extremen $to~belastun~en ~ es vor~eil-

i~aft, das Steuezpult au~ eine~ abgefederten ~a~en aufsCellen zu k~nnen,

und in einer ko=~osionsfa~dernden Atmosph~e kannte zu~ Sichezstellung de=

Lebensdaue~ Fernsteue~ung no~wendig sein. ~e es att~erdem maglica, die

Energievezsorgung des Roboters yon don hoboterazm getzennt zu halten, So

~6n:~te der A~m allein in explosionsf~hi6e Atmospharen, z. B. in elmer

Lackie=ezei, eingeft~zt we~den.

3. ~a oft heiBe Metallteile herumfliegen, ist es wichtig, dab die Au~enhaut

des Roboters nich~ entfl~mbar ist.

~. Wo Dreh- und ScAubgelen~e des Robote~s Staub auscesetzt sind, de~ Abrel-

bunsen ve~ursac~en kannte, mGssen die Gelenke nit einer SchutzumhGllung

vezsehen werden.

5. Da~ Problem de~ Feuer~e~aaz ;.i~d mi~ nicht emtfl-mmbaren FIGss£6keiten ~e-

18st, die wahlweise an6eboten we~den, da sie mat wesen~llch h6he:en Kosten

ve~bunden sind.

6. Bei besonders hoher Luftverscn~utzung kann sich d i e LtuftkGhlun~ als unprak-

tisch erweisen; dahe~ sollte WasserkGhlumg wa~lweise angebo~en werden. Auf

jeden Fall sollte die ~ungsluft zue~st sefiltert und damn in ein Geh~u-

se geleitet werden, um positiven Innendruck zu gew~ihrleistem.

7. Btaub und Schmu~z scheAnen dutch die e~sten Spalten einzudrln~en. Getrlebe

aoll~en dahe~ mat geh~te~en Za~de~n ausge~stet sein und unte~ b~oerd~uck

etehen, um Verschmu~zun6 zu verhinde~n.

8. Die Robote~loglk sollte gut gegen Netzlmpulse und Starungen geschGtzt seln,

die dutch die Ve~bimdungen zwlschen de~ Roboter und angescnlossenen Get,ten

elmt~eten kannt em.

~ i x ~ a c h a ~ t llchkeit

T~otz dex bedxGckenden A~pekte v i e l e r A x b e i t s p l ~ t z e in Fab~'lken we~den Un~erneh - mex Ro~otmr nux damn e i n s e t z e n , wenn d i e Kos ten Lm V e x g l e i c h zu m e n s c h l i c h e ~ Ax-

be i~ nied~Ige~ mind. ~id ~0 ze£~'t die Ea~tal~emdlte bel untexechledlichen Ro-

bo~e~p~elaen und Lohna~tzen sowe£ beA untezschAedliche~ Schichtdaue~.

Nenn eAn Robote~ den Ua~ebungsei~li~aen in einem Zndustriebet=ieb wlde~stehen

kann, e~ wg~end 98 % de~ Arbei~azelt bet~iebsf~hig Ist, gra~ere Znstandhaltungs-

arbeltem uur nach 10.0OO Stunden Betrlebszelt bem6tlgt und elne Kapltal~endlte

Yon 25 % ble~et, dann let da~aus zu schLte~en, da~ er sich wahrscheinlich dutch- met zen wlrd.