You are on page 1of 16

CHAPTER 8

Parametric Design

LEARNING OBJECTIVES

When you have complcted this chapter you will be able to

8.1 f NTROÜIJÜTIüN

Farametric design inchirÍes a nr.lmtler of decision-rnaking processes, just like


the othe: pliases of design. 'l'he processes use information (input) from'prior
phases lo arrive at logical decisions (output). Wh¿rt inakes pat'ainetric design
special and particularlS,'challenging is that'vye li,ill empio3,anal-vtical and
experimental rrethoCs to pl eciict and evaiuate the behavior cf each of the
design canriiciates to urakc these riecisions.
Let's exai:iine tlie iypes of input and output infolnratioit that we will
process as we make c¡ur decisions in the various design phases.

8.1.1 lnfornratlon [:low in the Design Phases


As we recall the phases of design as shown in Figure 8.1, information about the
customers'needs is processed in the formuiation phase usually resulting in a
Iist of customer requirementsiproduct functions, the importance of each, a list
of engineering characf eristics that quantitativelv describe how well the
functions are performed, a detailed engineering design specification, and
ideally, a House of Quality diagram.

177
()
178 Engineering Design Chapter 8 Parametric Design 179
(
( Similarly, we use th.at information as input to the configuration design
phase, wherein we make decisions about the product's architecture to decide
( wtrictr'parts should be standard or special purpose. Then we determine special-
( Customer Needs prrpor" parts geometric features, their arrangement and/or connectivity, and
ineir retailue dimensions. And we establish a list of attributes or variables that
we will need values for. And for standard parts, we select the specific type
( and

attributes that we will need values for.


( Custorner requirements
importance weights Intheparametricilesignphase,wedeterminethevaluesofthose
( En gineering ch aracteristi cs attributes, tyiically called rteiign variables. These usually telate to specific
House of Quality sizes, lengths-, radii, diameters, material types, and manufacturing process
Eng. Design Spe.cifications
requir"mJnts. Each set of design variable values is called a design candidate.
l.
Concept
( Design Wá predict the performance of altemative candidates using analytical method.s
all
and/or experimental methods. If the performance of the candidates satisfies
( Abstract embodiment we call them feasible designs. Then we evaluate the
the design constraints,
I
Pirl'sical principles
feasibie áesigns to detennine which candidate is the best'
Material
Geornetry
Finalyl during the detail design phase, we complete the remaining deci-
( sions, resuliing in éomprehensive product specifications, drawi1gs,
manufac-

( turing specifications, performancJ tests, and bills of niateriáis. Next, we


examine the parametric design of a simple flange bolt'
Speciai Purpose Parts:
(
Features
( Arl' an gern en ts/c on necti vi ty
Relative dirnensions 8.1.2 Parametric Design: Pipe Flange Bolt Example
( Attribute list (variables) to fasten a cooling water
Pararuetric At a petr.ochemical p1ant, eight bolts wili be used
Standard Parts:
I-)psign
Type pipe á tne inlet flange of a high-pressure chemical¡eactor vessel' The piping
Design r¡ariable values
A ttibute list (variables) iytt"* i, subjected to an operating pressure of 1,000 psi' Tire flanges. will be
( Ferformance Precii cti ons .rb¡..t.d to temperatufes up to 2501F. Each bolt wili be required to rvitlistand
Overall Salisfaction a 4,000 pound disign overláad, which is four times the typical operating- load
(
h4aterials of 1,000-pounrls. I'lie company has suggested a standald hex-head bo1t, shown
( Manufa ciuiin g processes
FJetail in nigure g.2. Some of tné deiign variables for bolts include: bolt diameter d,
(
Design overill length l, number of threads per inch, and type of materiai. Assume
that for this example, we have been asked to determine the bolt diametel only'
i Prod uct speci fica ti or:s The principil ¡unction ol a bolt is to fasten parts together so that they
Production ciravuings can be disássemül",1 1ut., for maintenance ór repair. As the bolt is
tightened, a
( Performance 'fests the
tensile force develops over its grip length. The greater the force. the better
Bilis of materiíils develops higher
( clamping action. Alio, since the head of the bolt consequently
h4fg. specificatioirs
friction forces, there is an improved resistance to loosening'
(

ffi L
I

( t

(
(
FIGURE 8.1 Design phase information input/ouqrut.
--t T

I
I

FIGURE 8.2
I{ead d Hex-head bolt.
( We use that inforrnation as input to the concept design phase, wherein we
( make decisions about the physical principles, abstract embodiment, primary I I I

rnanufacturing processes, and rnaterial classes. Shank Threads


(
(
180 Engineering Design Chapter I Parametric Design 181

However, a bolt can fail to function in a number of ways, including: The calculated diameter is then rounded up to the nominal size 0.25 in., so that
(1) the bolt head can twist off during tightening, (2) the bolt threads can strip a standard size could be purchased.
off, and (3) the bolt can experience excessive tightening or excessive operating As shown in Figure 8.3, as we increase the diameter, the calculated proof
loads that it permanently elongates or pulls apart. load increases exponentially. The required proof load is constrained to be
Let's consider only the last mode of failure for this example. 'Ihe ability equal to or greater than 4,000 (lbs.) and is shown as a horizontal line. The
of a bolt to withstand the tensile force without a permanent measurable minimum feasible diameter is shown as the dashed line at 0.245 (in') and,
elongation or "set" is called the proof load Fp,and is directly proportional to therefore, satisfies the constraint. Selecting a diameter larger than 0'2a5 (in.)
cross-section area A of the bolt and the rnaterial's proof strength Sp, by the wouid improve the safety and reliability of the design and could be considered
relation: in the final design.
This parametric design problem is rather simple, in that we have only one
Fo:AS, (8 i)
governing constraint (designed proof load must be greater than 4,000 [lbs.])
The Society of Automotive Engineers (SAE) has categor ized proof strengths and one algebraic relation to substitute for the area of a circular cross section.
of bolt materials as Grades 7*9, with strengths ranging from 33,000 psi to Thelefore, we can directly solve for the diameter that satisfies the constraint.
120,000 psi, respectiveiy. Assume that rve select a Grade 5 material, whose There is no need to iterate. Some engineers call this type of problem inverse
proof strength is 85,000 psi. analysis, because we can rearrange an analysis equation(s) to find the un-
Therefore, for this design problem, we ileed to design the bolt strong known value. Analyticai relations are often derived from experiment as

enough to have a proof load that is greater than the 4,000 (lbf.) design load. function = /(form). By inverting the analysis equation we find
This is a constraint for this problem. Consequently, \\e need to find an area lorm =,[ '(function).
and, tl:erefore, diameter such that:

¡-.. > 40C0 lbs (8 2)


v Fp versus d
1 2000
Ily sr"rbstitutir:g equation (B.i) intc the constr¿rint equation (8.2), r.+'e obtain:
r 0000
ASo>4C00ibs (8 3)
8000
We find that that the proof-load constraint can be satisfiecl if the area:
tn
*D 6000
, -- 4'000 (8.4)
¡t /-
se
.\
r.\
t". 4000

2A0A
,,
.na
r 4,000 (lirs)
(8 5)
fl
85,00ü (lbs/in2 )
,]

¿ > 0.017 't


Itl
tt "
(8 6) a (in)

and since cross-sectional area is relateri to diameter d as:


FIGURT 8.3 F-lange boit proof lcad versus diameter.

7r d2
A* ,
.1t
(E 7)
-t

The parametric design of most mechanical parts is much more difficult.


a

Let's exarnine the overall process to figure out how we can be more s)'stemat-
a/
7i a-
substituting, we find that 2 0.047 in2 (8.8)
ic, and thereby complete better designs with less effort.
4

a-o{l (4)
and further. that d2 > = 0.0598 ln- (B.e)
rT
8.2 STEPS IN SYSTEMATIC PARAMETRIC DESIGN
d >0.245 in. (8.10)
Systematic parametric design has five major steps, as shown in Figure 8.4:
(

( 8 Design 183
182 Engineering Design ChaPter Parametric

( Solution Evaluation Para¡aeters (SEPI), By reviewing the House of


ffi setectDesignvariabtes

(
Irl
I
r
ü
I Otoblem I Determineconstraints Quaüty chart and the engineering design specifications we reaffirm what
functions the customer wants the product to perform. We then select engineer-
ing characteristics to measure the predicted performance of the functions.
I
\
l* solution evaluation parameters are the engineering characteristics that are
selected to determine how well a candidate design "solves" the problem
I
( ll@a* @ setectvatuesforDesignvariables
""b (Dixon and Poü,1995).
Solution evaluation parameters depend upon the product and part being
( 'E.$t designed, but ofteu include: cost, weight, speed, efficiency, safety, and reliabili
O0
9-ol
I
ty. We also denote parameter symbols, units (of measurement), and any lower
l\
and or upper limits of the parameter. As mentioned in Chapter 3. we can also
§ J Ienalternati'es describe the customer's satisfaction with respect to each solution evaluation
aAv parameter, as a satisfaction curve. Finaliy, we agree on anaiytical or experi-
II
I I m PredictPerformance
mental methods to detetmine thei¡ values.
In the flange bolt exarnple, the custome¡ wants the bolt to clamp the pipe
t.routi"*, I Check Feasibilit,v: Functional? Manufacturable ?
flange to the pressure vessel flange. The clamping force can be measúred
I by the proof load ánd is, therefore, aq example of a solution evaluation pa-
Feasible irlternatives rameter. The higher it is, the better the clamping force, and the more satis-
{ fied the customer. We selectetl Jtp as its symbol and porurds-force as the
units. A lower linit. 4,000 (lbf.), is also established' The bolt proof load is
( Evaluate
i Optimal? Best feasible alternative? constrained to be larger than 4,000 (lbf.). An analytical formula will be
( 1"*:_J used to predict the proof load, although experimental tension tgsts could
have been used.
(
Design Variabtes (DVs). Parameters under the control of the designer,
which influence the candidate's performance, are called design variables.
relate part dimensions, tolerances, and/or material
ft.__
'_**_) Design variables usl]ally to
properties. In addition to the design variable names, we establish appropriate
wt I Optimized best a]ternative- iymbols, units, and upper and/or lower limits or bounds. For discrete variables
we determine permissible values (e'g., 2by 4,2 by 6 lumber, % inch diameter
FiGURE 8,4 Pararletric design clecision-inaking processes. bolts).
The diameter is selected as the design variable in the flange bolt example
Larger diameters result in larger areas and consequently, improved per-
Step 1: Formuiate the paran:etric clesign problen:.
formance. Permissible vah¡es are identiiied using handbook data'
§tep 2: Gener:ate alternatirre dcsiqns.
$tep 3: Anal,vzeii:i edict the pe; formance cf the, aiternalivr:s. Prablem Defini.tian Parameters (PDPs). Parameters that describe specif-
iirep 4: Ilvaluate the perfornr of each alternatii,e. ic concitions of use, such as operaling oonditions are called problem definition
( aTlce
parameters. in addition to ih"it nu,o"t, ,r. estabiish appropriate s-r'mbols,
( Step 5: Gptirnize/refine.
units. and values.
The operating pre§srue of 1,000 psi and the limiting temperature ol 250" F
( are identified as PDPs for the flange boit exampie.
Step L Formulate the problent. Practicing engineers often comment that the,v
( spend 40*50 percent of their time gathering appropriate data, clarifying im-
Preliminary Plan for Solving the Problem. For small design problems, we
portant details, considering different analytical and experimental analyses. and and start calculating things. For larger, more involved prob-
( often jump rigtriin
then pianning how the¡r will complete a project. All these activities relate to
lems, we need to make a prelirninary plan based on considerations including:
( problem formulation. As we start a parametric design problem, we need to
familiarize ourselves with the problem parameters and also plan ways com- 1. Do we have analytical models/formulas for our problem?
( plete the design. 2. Are the assumptions used in our models the same as our problem?
(
(
184 Engineering Design Chapter 8 Parametric Design 185

3. Will we need to perform pilot scale or bench-top experiments to vaháate "physical reasoning." In other words, by becoming familiar with the causes
our analytical formulas? and effects, we can Iogically reason the "physics" of the problem.
4. How much time and money do we have to "solve" the problem? If no feasible design candidates exist, we select new values for the design
5. Are "ballpark" computations required, or do we need more thorough variableg and thereby generate new design candidates. These are subsequently
and precise calculations? analyzed for feasibiiity. This is the solid "redesign" iteration loop shown in
6. Do we have knowledge about acceptable industry standards? Figure 8.4. If we cannot find any feasible candidates, we might ask whether we
L WiU manufacturing specifications need to be generated from our analyses? have set our design specifications too restricdvely. Perhaps one or more con-
'8. Do n e understand the customers' function requirements versus satis- straints could be relaxed. lf so, we are "respecify"-ing as shown in Figure 8.4.
The formula for the area of a cylindrical cross section (8.7) and the formu-
faction weli enough?
la relating the proof load of a bolt to its area and strength (8.1) were used
9. Are we sufficiently qualified or competent to complete the required design?
in the bolt example. The performance consftaint considered i¡ the bolt ex-
For large design problems, design teams rvill consider the questions ample was that the proof load be larger than 4.000 (lbs.). Since fhe formu-
above and wili prepare a design project proposal for upper management to re- las were straight forward, we wére able to juggle the equarions inlo a se-
view and approve. Design project proposais include items such as a back- quence that did not require iteration. More complicated problems will not
ground sectir;n' goals or mission of tire project, a scope of work (of the work be "soivable" by equation jugqling.
tasks to be performed), schedule, and budget. Regardless of the size of your
Step 4. Evaluate the results of the analyses. The feasible designs are evaluated
pro.ject. however, it is airryays a good idea to prepare an outline of what you are
to determine the best dcsign. Usually one or more cfiteria are identified in the
going to do ancl wiren and haw you are goirrg to do it.
formulation plrase and used to deterniine the "b¿sl" feasible design alternative.
Ilecognizing the simple nature of the flange bolt calculations. u¡e u,ent
strai.sht to making a feu, calculations. The proof load was established during formulation as a rough measure of
customer satisfaction. We assume that the higher the proof load, the more
§rep 2, {Áeru**te af{er¡zr¡tíve desigr¿s'. We select different values for the clesign satisfied our customer ivould be. That would mean that a large bolt should
va;iables to generate different canCiclate clesigns. T'irese values cail come fi-om be chosen as the "best" design. This, however, ignores olher aspects of reai
úur own experience, from our coripa:ry's experience, oi fi orn industry design problens such as weight and cost limirations. Unfortunately, since
siandarc1s,Sonretir.¡es\,,,efieedtarrak.eec1ucaterJguessels. rve do not know how tl¡e customer feels about thesc issues. we can only
surmise that he might be satisfjed rvith the smaliest boir that meers the
§rep 3.A**{yze the *Íteruative rlesigrrs. §ie preiiict t}:e pe;:formance of eac}r
force constraint. As we sirall see in the next section, we should alwavs try
candidate clesign using analvtical and/or experimental n:ethods:
to ascertain customer satisfaction with respect to each soluüon evaluation
An*{jt:"¡cal fr'f eliwds. FonlruJas froni p}rvsics, niathe:natics, and rile engi-
parameter.
neering sciences are müst aften usetl. Sometirnes, aclvanced cornputel-aicleC-
dcsign packages, such as finite eJement analvsis, cciilputaticnal fluid dyuamics, Step 5. Optimizelrefine. Optimal design ruethods automatically regenerate
and inotion sin:ulation. are usecl new values of the design variables to improve expected performance and
hxperi¡nerct*! M#h*r§s üftentirnes. ti:e ccmplexity t:f ti:e desigr: is satisfaction (Arola. 1989; Papalambrc-.s and Wilcie, 1989: Rao, 1984; Reklaitis
bey6s¡1 the accut'acy or assumpticns of oul anailrtlsal raodels" in these cases" et al., 1.983; Sidciall, 1982: Vinderplaats, 1984). For single-altribute optirniza-
\ryt can buiid scale rnodels ancl/or full-scaie models crf the plcduct or critical tion. a single criterion is chosen, such as minimizing tlie weight of a part. The
parts of the prcduct. and test tire.ir performance. We cali use wind iunnels, for criterion is a function of the design variables, and is calied the objective
exarnple" to anah'ze the performance of complicated r,ving geometries" or function. Design variable values are usually constrained to some upper arid
lower limits. The predicted per{ormance of apafi is also reiated to the design
check control surfaces.
variables. For example, the strength of a bolt is a function of its diameter.
The performance of each design candidate is checked so tlrat e\¡er-v per-
Design variable functions and limits are called constraints. Equation (8.2), for
formance constraint is satisfied. These designs are called feasible designs. If
exampie, is an inequality constraint. Excel's Solver feature can be used to
the constraints are violated, we reiterate back to generating another alterna-
optimize many typical design problems found in mechanical engineering. This
tive and then anaiyzing it.
is shown as the dashed regenerate loop in Figure 8.4.
As we substitute new values for the design variables into the system of
analvsis equations we become familiar with how each design variable influ- A more detailed example using this systematic procedure is presented in
the next section.
ences part or product performance. Using this familiarity, we can sometimes
generate new values that satisfy the constraints. Dixon and Poli (1995) cali this
ChaPter I Paranretric Design 187
i 86 Engineering Design
pulley is
parameters (sEps) 'Thep:incipar funcüon of the
( 8.3 §YSTEMATIT PARAMETRIC DESIGN: EELT-AND-FULLEY EXAMPLE solution Evaluation to low speed' In the
íom a t'igi' ipeed
to traúsform the power of if'" *otot to a larger torque' ac'
( Let's apply the systematic procedure presented in the last section usiug more the smaller motor torque is converted
a
;;#;;;;";, belt-and-pulley system
detailed exarnple of a belt-and-puliey transmission. ;#;;"ffi;' Ánservation of energy law' A-tsg' the
itt ptñAp'iiáttig.t if bett §lipped o¡'if'the belf
would fail to perform .tf"
to .r.t"iui the customer would be more
i"ns'ion' Finally'
( broke owing
fi,3.1 §esign Froblern Forrnu§ation satisfied with a comPact desion' amount
(
Since we know that tiJitn'ioo
force in the belt is limited by the
A Yz-hp electric motor, running at 1,800 rpfil, will be used to drive a grinding driver pulley' up to the point of impending
( wheel operating at 600 rpm. A flat belt-and-pulley drive system configuration of friction between ,¡" Utlt on ti'" r¡'
;ü;;;;il;;;*ái," i"'q.e t¡x ihe be]t can deiiver to the pullev'
has been selected, as shown in Figure 8.5. The design team has also deter- 'i' supply' Aiso'r''
i
and compare it with the
i"'q*'' that the motor
mai*'i' can
mined that:
belt tension' F1 ' to make sure that
it does
r the drive motor will have a}-in. diameter pulley mounted, we should calculate tn' *t'i*"*
§ candidate designs should be able to transn:it the fuil horsepower, nOt exceed the 35 (ibs.) limit'
parameters in Tabie 8'1'
E We can su§lffiarire the solution er¡aluation
the customer desires a compact system design,
I tire drive pulley wiil slip first, before the driven pulley,
§ the purchasing department has ]ocated a venclor that can provide a flat uti on E, alu ati on P aram eters f or B elt -Pul leV- S,r'st eni
TABLE 8.1 S ol
beit that can withstand a maximum 35-pound tensile load,
Parameter S)'mbol
the coefficient of fricticn between the i:eit and puliey is Ü.3.
Jb
( belt torque
other ciesign engineers in your group will design tire mountiilgs, bearings, L

1 beit teirsion fi lbs


( and protective equipment, therefore, k

c in. small
a
.-1 center disiance
parametric design efforts should foci:s on tlie distilnce between centels
(
and the diameter of the clriven-putr1e3'.
diroctly affects
(
ffieslpr \¡arlahl*s (Bys) The vaiue of the csnLer rlistarice, Ü"

be deterniined by tlie designe'r' Also'


( tlie co;npacrness of ihe d*sig* and is to
rTlore of tire belt wraps sl'ound tlie pui}e1'
as tne center dista*ce is increase{i,
pulf i')' iRciea-i-lÍ,]"l:.ability of tho
( It4of.or Puliey Cir inciin g \\/heel Puile,v
(i.*.,is in ccntact rvith tlie surfece rif tire
(drivcr) (rlriven) the tolque requiren:ellts *f the
( \,d 1,úi,rt, r.2,t|2,ú2.n, belt to grr, the pulley anr] therel:-r,satisfy
ere suillffial ized in T-abie 8'2'
1nüto]-. The iiesigri va¡ iables
I

(
TABLrE.?L)esiF.n\¡ariiibiesiorBeit..Pulle-.vS"r,stem
l jtit-
Ile.sign Variable S-vr:r'o-oi Ifl:----!!**i !:lt--!ppqt
( 1 center ciistarrce c - ill' sn:all

driven puiie-v di ameter


d" ) J!'

(
(
oo'e find a
( prohlem sefinition parameters stuclyi*g the design.problem.data,
,,givens,, that define design páut** roñoiiiolls such as the friction
number of Therefotre'
( and motor pulley diameter'
coefficient, beit strength. motor power,
(
c4 we identifl, these as the problem
definition parameters in Table 8'3'
tl
( FIGURE 8,5 Belt-and-puliey-drive system for motor and grinding wheel.

(
(
T
I
I

1 88 Engineering Design Chapter 8 Pararnetric Design 189

TABLE 8.3 Problem Definition Parameters for Belt-Pulle,v System h=n zsin-1
W) (8.13)

Parameter Symbol Units Lower Limit Upper Limit


Similarly, we know that a belt, with coefficient of friction f that is in contact
1 friction coefficient f none 0.3 0.3

2 belt strength 4ru* lbs 30


with the pulley for an angle of wrap fi , has a maximum belt tension 4 , on
the taut side of the pulley, and is reiated to the belt tension Fz on the slack
3 rnotor power w hp Yz

side of the pulle"v according to:


4 motor pulley d¡¿meter d1 in. 2

lt s uro (8.14)
F2

Plan for Solving the Design Problem Using analytical relations from physics Upon sketciring the free-body diagraill, shorvn in Frgure E.6, we see that for
and mathematics we can use a hand calculator or develop a spreadsheet to static equilibriurn, we can sum the moments about the bearing B to obtain the
calculate the engineering characteristics, including:
torque I,,, delivered by the belt to the driver pulley of radius r'7, iisl
1. giinrling wheel pulley speed, iir,
Fr)ry
2. angle of wrap as a function of the center distance, c,
Tb = (R - (8.1s)

3. belt torque, Io, As a point of interest, we can show how the angie of wrap affects the torque
4, maximum belt teilsion , Fr, capacity of the belt at im.pendine slip, b), substituting equatiort (8.14) into
(8.15). By examining the resulting equation (8.i-6), we see that as the angle of
-5. slack-side belt tension, 4 , and
wrap increases and the torque increases.
6. initial tension (before torque is applied), Fi
( F^\ _.(. 7)
-4¡ = ¡ll 1 --*
\ e"')l,i \ e'' )l',
T'hen we u,ill check that tile constraints are not violateci. Specificaliv, r,ve rvill Tt =(¡i - P,)r, = i 4 (s.16)
make sure that the belt will deiiver the frill motor torque to the grinding-u,hee,1
pullev and that the belt tension does not exceeci the belt strength limit. The maximum torque delrvered bv the r:otaÍ, Tm can be determined by
re.Íirranging eqriation (8.10) as:

w s2s2 0.5 (5252)


(ft lbf) =7J.5'2(in. Ibf.) (8.17)
8.3.2 Generating and Analyeing Tu, =1.46
t"t, l8Cü
irollorving our plan, we need to develop ait anal,vtical rneans to predict l.he
behavior of the system. We can model the behavior of the system using
relations fronr ph),sjc.s and rnalhematics and clevelop a svstem of equations to
analyze the performance r¡f each roídidute clesisn as rve substitute rJiffereni
.,,alues for the eic;sign variables.

For exai::ple" we knoiv that a motor will deli\¡er power Vt (hp), to a pul-
lev rotatrlrgdL n rpnt rryiren producing a torque 7-,, lb. ft. accoi'dii:g to eclitation
I
(8.1 1): I

, 'l- vl
¡\t
14/=# (hp) (8 1i) - .-.-.1-
I
5252 t
\
If the belt is not permitted to slip on the pullevs. the pulle,r, speeds are related
to the ratio of the pulle,v diameters.
nrf nr- dtldz=\lrt (8.12)

We can determine the angle of wrap l, using basic geometric relations.


I

FIGURE 8.6 Free-body diagram of motor pulley.


(
(
190 Engineering Design Chapter 8 Parametric Design 191
(
( The belt must be abie to provide a torque equai to or greater than 17 .52 (in. \\/e cileck the beit torque using equatir-rn (8.15).
lbf.); oiherwise it wili slip.
( Tn = (f1 - Fr),i = Q7 .5 - 20.0)1 =IJ .5 lb'in (8 2s)
Using equation (8.12) we can predict the driven-pulle,v speed, fi,2, 3-s a
( function of the design variable, diameteÍ dr. We note that to satisfy the torque constraint, a 37 .S-Ib. tension will be neces-
sary for Fr. But that level of tension exceeds the belt strength constraint of 35
(
txz=rr|=1800 (8,18) lbs. Therefore, a center distance of 4 in. is an infeasible value. It causes a
( fioo*> violation of a constraint and will need to be increased (to increase the angle of
Note that equation (8.18) is an equality constraint, having one unknown design uzr&p, and consequently reduce the tension in the belt).
(
variable , dz.A feasible candidate design must satisfy this constraint. Since the To reduce the effort in computing the expected system performance a§ a
\ grinding-wheel speed has been specifred as 600 rprn, we can obtain a feasible function of center distance, c, we can develop a spreadsheet as shown in
value of the driven-pulley diameter by rearrariging equation (8.18) Table 8.4.

dz=drL=r(ry)
L 'nz =u(in.)
\ (8 le)
[600/ TABLE 8.4 Flat Beit & Pulle-v Design

Therefore. the only feasible value of the design variable d2 rs?in.


Problern Befinition Parameters
Pararneier Symbol Units Yalue
Generating an tnifial Ya]we of the Design Variab]e c Sjnce the customer rvill friction coefficient f none 0.3
be r¿ore satisfied wiih a compact design, we q,ouid like the eJistance betrveen belt strength F*u, lbs -15

( the pulley centers, ú to be small. TIie ciosest ihat the tivo puiieys can be is motor power W hp 0.5
motor speed ny rplx 1800
when their radii are almost touching, theoretically spe.aking, or
( motor pulley diarleter Dl inches 2
Cmin ='l'1.*l'z -!l'lz*6¡'z=4j*.
ZZ
(8.20) flesign Yariables
\¡ariable S),rnbotr Units Lorver \/alue Upper
Now we can use er¡uation (8.13) to find the angle of u,rap crn the driving clriven puliev diameter dz inches 6

pulley. center distance c inches 4.A 10 12

Perfbr¡nance Calcslafions
#t = n *zsin-1f:*l =2.üe racl = fta des (8.21) Eng. Charatteristic Syrmbol IJnits X/al*e 6p-
[ 2(4), motor torque T* lb-in 11.57
grinding wheel speed ¡12 rprn 600 = 600 Satisfied
To f-ind the tensile force Ft that satisfies the motor torque constraint \\¡e Ltse angle of wrap ú, degrees 120.0
equatiolr (8.16) to obtain bcit tension-taut Fl
belt tension-siack L-z lhs 24,0

=
r'*(^ T-- - -= , 1J ''!2 ,- initiai belt iersion ¡; Ibs 28.8
i;r -3J .5 1bs (8 22)
I 1--
tl-l,I ¡ t---- I )-r'r '
i;elt torque Tn lb-in t?jl ¿ 17.51 Satisfied

[' ufa ) [' eo'3i2'ae) I

We find the tension on the siack side of the beit as

D - !L= :J;]^- = zü.a rbs


t1-;ff -W@ (8.23) Rerlesigning: Finding Feasible \Ialues Since tire belt tension is constrained to
a maximum of 35 ibs., the initial value chosen for c is infeasible. We therefore
The initial tension in the belt before the torque is applied is obtained as increase the center distance and recalculate the belt tensions. lJsing the
spreadsheet, we obtain the follou,ing values for belt tension, F1tnTable 8.5.
t:_LaLL-
P, + F,-j#=28.8
31 .5 + 20.0
lbs (8"24)
'27

I
Chapter I Parametric Design 193'
192 Engineering Design

8.3.3 Evaluating
f¡om among those fea-
TABLE 8.5 Belt Tension for Alternative Center Distances
During evaluation, we choose the best design candidate
and analy¿e¿ earlier' Often our company will
Center distance c (in.) Belt Tension F1(lbs) sfute iesigns that we generated
Other times we may want to maximize an as-
*ánt m" í.utt design.
4 31.5
"*p.orÑ,
ot ryrt". prior*un* tJobtuin th" fustest, or the üghtest'
or the most fuel-
4.96 (Goalseek) 35.0 f".t
'efficlentiesign. Like many design teams, we could, therefore' select one of these
6 33.5
i.best,i to determine our final recommendation.
8 32.0 tUJ*easure of
10 31.2 "iit.¡uMostdesignproblems,however,havemultipleattributesorcriteriathat
",
for example'.we
usually involve'trade-off decisions. tn the bált-pulley system'
center áistance and belt tension (i'e''
,.r¿ io make compromises between
compactness versus saf etY).
' these neces-
strength constraint is satisfied. Holever, to provide soffie extra capacity or Therefore, we need a rational rnethod that can help us make
As in concept design and configu-
compensate for sorne belt wear, or a slight decrease in friction coefficient, we sary compromise§ amóng muttiple criteria'
method can be used and is recommended
usuatrly select a larger value. perhaps B, or even i0 in. Selecting c = 8 in., for ration OÁign, the weighted-rating
erarnple, \ ¡e r.r,ould obtain a 1.09 factor of safety F,S,frorn: whenever possible.
F
tullov,able Step t Establish a set of evaluation criteria'
F,S_ =35 l32=1.09 (8.26) criterion'
tL
Step 2: Rate the feasible designs for each
design
Step 3: Weight the ratings according to importance'
Factor cf safety is a term used to express ratio of loacls or stresses as exernpli-
Step4:Sumtheweightedratingstocalculateanoverallweiglrtedrating.
fied in equation (8.26). Factors of safety less than one mean that the design
rr¡iil not support the load without "failing." Factors of safety greater tharr one StepLEstablishasetofevaluationcriteria,irnportanceweigbts,andsatisfac.
rve tleed to
inijicate that the design is "safe." Factors of safety &re ?,lso discussecl iater in tion levels. if we rvish to inaintain high standarcls of quality
the cliapter. incorporatethe..r,oiceo{thecu§tome¡.',Therefore,evaiuationcriteriaare
or fror¡r
Finding feasible vahles for the design variabies is what generatin.t ancl oitrri A.r.toped from the iist of solution evaluation parameter§.'
of quality' andior
arialvzi¡iq is ali about. X¡rf'errslbie designs aTe those t,,hose design variable errgi,reering áharacteristics. Recall, that we use the house
(1) functioÚai
vairres clo not satisf'*¡ tire constraints. We found that i,vlren d.r.. 2 tn. an,i c = other mar-keting research tools, io irlentify the customer's:
(that meas.
requirements for. tlre product, (2) key engirreering,characteristics
'
I in. al1 the constraint.s ere satisfied (speed. torque, aird tension). But are tliesc of each
ure horv well the functions aie pertáLmeO)' and (3) the importance
feasible values the best values?
functional requi¡ement.
Tr.?de-$ffs As sho\r,il in'fable 8.5, as tve increase the center distance, we Fo¡ our belt-pulley system, let's assume that the product development
was fully utilized'
clrtain lorver values of belt tel:sion, and consequenti), higlier factors of safetr'. team determined that as long as the moior horsepower
the two most important engineering characteristics were beit tensioir and
And that's goor1. lJrit. \\,e are also iilcreasing t}:e size of the s,vstem. {J;rfortu- tbese earlier during the pa-
center distance (conrpactl')ess). We identilied
natelr¡ this ciesign prroblem. like most design ploblcms, exhibits a trarle-*flf', evaluation parameters'
rametric design formulation step as solution
u,herein orie attribilte ir::1-rro\/es as the other rlegrades. 'i racJe-offs aie causecl that the
Let's furrher assume that the feám was also able to deterrnine
bi,the iriterdepenclency of variables. li,pically referred to as couirlimg. Both center distance as
customer considers belt tension as "very important" and
cümpactness and belt tension are coupied to the center ciistance. \\ie tracie-off with a weight oI
"important." Assume that we interpret "very important"
higirer belt tensions as rve satisf,v tl:e customer's desire for compactness. We nith a weight of 0'4' Note that we can certainly use
0.6 and "important"
could ask, "What is the best desisn?" Is more safet-v (lower belt tension) and a othernumerical values for imporrance levels' The very act
of choosing
Iess compact svstem better than iess safety and more cornpactness?
numericalvaluesoftenleadstolotsofdiscussionamongde§ignteam
This question is difficult to answer. However, as it has been said, "beauty members. And whiie it may be time-consuming and frustrating'
team dis-
is in the eye of the beholder." Whv not consider the trade-offs from the cus- cussion is very constructive in that it leads to a better understanding of the
tomer's perspective. That is, how does the customer feel about the importance customer's desires.
of each of these attributes? Also. is there a way to assess overall satisfaction of weighted
the customer? This is the process of evaluating, which we consider in the next Step 2. Rate the feasible ilesigns for elch criferion' We can use the
very good)' as we
section.
rating method and the rating icale e1 ¡r-4 (unsatisfactory to
(
I Design 195
( 194 Engineering Design
Chapter Parametric

(
from geometry can be used as in
our
used in Tabl e 4.12. Irüote that the people doing the rating may or may not using aspreadsrreet. Basic formulas Trendline
( programr. ,*n É* used. And Excel's
reflect the voice of the customer. exarnple. speciar computer linear'
is quite capable in curie-fitting the following relationships:
( R"ather than using a rating, we can use a custoüler satisfaction curve or feature
exponentia-1.' and power'
function. We arbitrariiy let minimum satisfaction be equal to the numerical pofy"o*ial, logarithmic, fit a sirnple,
since the curves are straight rines
in ttris example, we -can
value of 0 and maxinnum satisfaction be equal to the number 1". These are using the two-point formuia
( simiiar to the ratings of 0 and 4, discussed before. The main clifference is that straigilfin"int*rpolation *quutlon
_ y,
\
in satisfaction curves or functions, we link the value of each criterion, or §EP,
\ry *
( yt,_ ly, )(, _ r,) (8.27)
to a value of satisfaction. This linking of satisfaction versus SEP values is Vz- \)
facilitated using graphs and curve-fits. Let's see how we can use these to evaluation parameter for
evaiuate our belt-pul}ey designs. we substitute satisfaction s for y, and solution of 's is 0'
\ maxi*u* ,alue of s is J., and the minimum value
Let's assume that the product development team also prepared estirnates x, noting that the as
curve we find the satisfaction
( of customer satisfaction with respect to belt tension ancl center distance as For an sEp with a decreasing satisfaction
grapheri in Figures 8.7 ancl Figure 8.8. .1 (r, --t) (8.28)
( r]- -
{rr* *r}
as
curve, we find the satisfaction
For an sEp with an increasing satisfaction
q (\
"-tr+ = /
l'-''l. \
(8 2e)
\xz*xt)
x,'
(g.zg) are cnly vaiid betr,ryeen xzand
(i
q*
idore that equations (g.2g) and
V)

in equatians (8.30) and (s.31)


We set uP the belt-pulle3, suli,factions
(¡s - r') (8.30)
0.0
for belt tension F, as Sr¡1 = ------------:-
(:s -:o;
Be.]t Tension (lbs)

f tGURE 8.7 Customer satisfaction versus belt tensian.


and for cornpactness c as §, :gq4
(zo - s1
(8.31)

( }n this step we muitipll


§tep 3. §veighÉ the r*áing accerdimg to. inrp*rtance"
( importance weight'
the satisfaction r ating by the i-espective
(
weighteel r*tiN,rgs tq) calear[aés i]m
over*]] rt'eüght*d raÉEmg' we
st*p d" smmr tɡe
the impoi-
sat.isfaction ü as i]:e sltm of
( can estirnaie ihe custoñreí's oürail
1et 5, be the level of the
tance-weignred satisfaction ie,els.
s¡reci{icariy,
( u
evaluation paranreter and r+, i:e
equal
q;
w
customer's satisfaction for i-th salution
( Then the overall cusi.onrer
to the imüortance weight for that parameter"
. 1 r -_

( satisfaction can be estimTiT,


(8 32)
( 5 Cenrer distance c (in) ZA n,¡,s¡ - tü1si + w.:sr *...

as
FIGURE 8.8 Customer satisfaction versus center distance. we can calculate the overall satisfaction
! For a senter distance equal to 6 in.
(8.33)
u'¡ §4 + w.S,
¿ After the satisfastion curves are determined, we c¿ln rate each feasibie O=7, w¡S; =

design by reading values off the curves and entering them into the weighted-
(
(
rating tabie.
To facilitate the weighted-rating calculations we can find analytical for-
e=a6ffi+04ffi (8.34)

mulas that fit the curves (i..., curve-fits) and then automate the computations
(
(
Chapter B Parametric Design 197
196 Engineering Design

Q=aJ=IP+049L9
(3s - 30) (20 - 5)
(8.3s)
Satisfaction versus Center Distance

0 = 0.6(c.3)+ 0.4(0 e3) = o.1B + aH = ü.55 (8.36)

Other values of overall satisfaction are calculated and summ arized in Table
8.6. Ir{ote how values of the design variable , c, aÍecoupled to the values of belt
tension and compactness. Also. note that cquation (8.34) can only be used
when both belt tension and center distance are within their respective interpo-
o
lating points. Outside of these, a value of 0 or t wiii be rnultiplied by the §
a
0.6
:J
é.
respective importance weight. (n
'Ihe satisfaction vaiues are also plotted in lrigure 8.9. Tire satisfaction
curve for belt tension illustrates hou, as c increases, the customer is more
satisfied. This occurs because the belt tension decreases. resulting iri higher
factors of safety. And similarly, as c increases, the customer is less satisfied
because the design is less cornpact. 0.0 +
'I']ie overall satisfaction curve, orl the other hancl, illustrates that the cus-
051015202s
tomer's satisfaction is somewhat maxiuiized rvhen the center distance is about Center Distance c (in')
10 in. Also. the figure indicate.s that the overall saiisfaction is fairly insei:sitive
from about c = 9 in. to about c = 14 in. f'hat appcars to jndicate that the -** Y:**:19:::il-,
i'' ---u--- Bett Tension - . s. - Compactness
customer may be fairly indifferent to values of c bet\\/een 9 and 14 in.
The overali satisfaction folmula can be used as an aggregate objeciive
FIGuRE 8.9 Satisfaction ctrr\¡e's for belt-pulley s'vstem'
functioll to. optimize the belt-pu1lsr,' dcsiun prcblenl. An aggregate objcc-
tive fi-rnctioit is used to optimize probiems rvhen rnore tllan one critelior:
exists, in other u,o::ds for mu}ái-af tribute optimization. Aggregate objective satisfaction formula as an
Excel's soi,,er featrire and the overail
L_rsing
value of c = 11'38 in' resulte'ci itl
futrctic.txts combine a number of separate objective filnctions into oi:e scaiar ¿in
aggregate objective function, un optimal satisfaction
function. 10 in. results in an overall
overail satisfaction o,ia.723, wheleas =
c

The set of the desisn variabie values that satisfies all the Constraints
of. C.177,
custoiner's satisfaction is there-
(speecl, torque, and te*sion) and maximizes the
TABL.E 8,6 Weighted Satisfactions Based Oir Differe rit Ccntcr Disrances c iirr. rt^= 2ln. and t: = I i'38 iri' '
Ceutcr 'fension Compac.tness Ovsvuli
riistance Satisfar:tion
F1 (lbs) S¡, u'4 §E c (in.) S, t4'rS.
c (in.) STÜRAGE TANK EXAMFL§
$YSTÉhIIATIT PARAMETRIÜ üESIGN;

8.4
4.96 3s.00 0.00 0.00 4.96 1.00 0.40 ü.40
continues the use of spreadsheets
blt
6 33.5t] 0"30 0.18 6 0,93 0.37 0.55 Let's consider a classic'al problem thaf for finding the
I sufficient conditions
32.U0 0.60 0.36 E 0.80 0.32 0.68 also invoives using the necessarv and
i0 it.z) 0.16 0.46 10 4.61 4.77 0.12
optimum of a function of one variable'
).4 30..x0 0.92 0.55 14 0.4c 0.16 0.71
18 30.00 1.00 0.60 1B 0.13 0.05 0.65
ZO 29.9A 1.00 0.60 2A 0.00 0.00 0.60
8.4,'t Design Problem Formulation
our engineering team to design an
Our companY's Canadian office has asked
tank should-hold a minimum of 10'000
above-grorrá fuel storage tank' The a tank'holding
liters. However, theywould be very happy if they could afford
I T
(
Chapter I Parametric Design 199
( 198 Engineering Design

of $10,000, but
liter or more. Their budget is limiteg to a maximum
t
15,000
would be very satisfied if the tank would cost less than $5,000'

The design team has also determined that:


o
10 mm thick'

local codes require sturdy steel walls at least


q-.,{
I (¡)

be cylindrical, ñ
I to facilitate routine maintenance the tank should a
then welded together,
I two half-cyiinder shells would be rolled
top and bottom'
I two circular plates would be welded to the 5,000 Cost, C ($) 10,000

building codes restrict height to a maximum of


¡ L5 meters,

t compared to diameter for calculating purposes?


wall thickness is negligible FIGURE 8.10 Satisfaction as a function of storage tank cost. Note that satisfaction is largest for a

r steel weighs 7,830 kgim3, tank costing less than $5,000.

T steel sheet costs about $3/kg,


: footprint should be no larger than J"0 meters square,
insignificant, and
¡ welding, rolling and other incidental costs are
o/o,35% respectively.
of
cost and volume have a importance weights
t 65
(,
o

/a ,
parameters (sEPs) The principal function of the tank is +.¡
a a'
solution Evaluation
to safely store fuel. The walls must be sturdy to resist bucking, corrosion and a a
wilt depend on getting a a
local weather conditions. Our affiliate's satisfaction a
It will be more satisfied if cost C
tank with as much volume V aspossible.
also
10,000 Volume, Iz (liters) 15,000
is controlled.
would be 100% satis-
For evaluation purposes we wil assume that they
They would be not be satisfied FIGURE 8.11 Satisfaction as a function of storage tank volume. Note that satisfaction is largest
fied if the volume were 15,000 riters or more.
for tanks bigger than 1"5,000 liters.
liters, nor with a cost greater than $10,000' we
with anything less than 10,000
that they wourd be satisfied with a cost of $5,000 or less' The
can assume
in Table 8.7. The satisfaction
solution evaluation parameters are summarized
curves are presenteá in Figure 8.10 and Figure 8.11,. Note that the satisfaction
o
(ro,ooo - c) . (v - to,ooo)
,,less-is-better" shape and that the satisfaction for volume is Y _- 0.65.
v'v¿ _ \"Jr
+t 0.35 _ (8.38)
(ro,ttoo - 5,000) (ts,ttoo - 10,000)
curve for cost is a
and "more-is-better"
a ,,more-is-better,' shape. Having bóth "less-is-better"
means that we will have a trade-off to make' For
satisfaction curves usuaily Equation (8.38) should be used with caution, The linear interpolation formula
(resulting in less cost satisfaction), to
example we may trade a higher cost tank can only be used between the end points. This can be handled with an appro-
satisfaction).
gain more storage vorume (iesulting in more storage priate nested "if" statement for each term, for example:
customer's overall satisfaction 0 as the sum of the
we can estimate the *
cost and volume as =IF(cost <=5000,. 65,IF(cost> 1 0000,0,.65 0000-cost)/( 0000-5000)) ).
importance-weighted satisfaction levels for
(1 1

(8.37)
0 =»r-w¡S¡ = wcSc + wr^Sy Design Variahles (tlVs) The two rolled half-shells are shown in Figure 8.12,
The top and bottom circular disks are not shown. The radius r, thickness, f and
height h are variables that determine the cost of the tank. Local codes restrict
g.Z Sotuiion Evaluation Parameters fbr Storage
Tqlk
iÁglf the wall thickness to 10 mm. Therefore the controllable variables are radius
Parameter Symbol Units Lower Limit upper Limit and height. They are therefore defined as the design variables as shown in
Table 8.12.
cost C$ $5,oot) $10,000

volume V liter 10,000


200 Engineering Design Chapter 8 Parametric Design 201

8.4.2 Generating. Analyzing and Evaluating


We can model the tank with a system of equations and thereby analyze the
performance of each set of design variable values, which we will call design
candidates.

Modeling For example, we know that the material volume V * of the tank
walls, top and bottom is the product of area A and thickness f. Further, the
weight t¡¿ i* the product of the volume and the weight density y as

W=TV¡t=ytA (8.3e)

The area of the tank walls, top and bottom can be found from

nnl
FIGURE 8.12 Rolled walls of storage tank before seam welding.
Which simplifies to
A-

W
A,uatts+

A=h(Znr)+firz
\op*

A=Zn (hr +r')


The weight and cost are consequently modeled

- TVu =ZtYt(hr +rz)


Ahouo^

+tTrz

as
(8.40)

(8.41)

(8.42)

(8.43)

TABLE B.B Design Variables for Storage Tank C = cntW (8.44)

Design Variable Symbol tlnits Lower Limit Upper Limit


C= cuTnyt(hr +rz) (8.4s)
1 radius rm 5

2 height hm 15
The storage volume, V canbe calculated from the formula for the volume of a

right circular cylinder as


V * hrcr2 (8.46)

Problem Definition Parameters (PDPs) The givens for this design


From equation (8.45) and (8.46) we see that cost and volume are directly
problem include the unit cost, density of steel and wall thickness (local
related to the radius and the height. We note that this design problem
standard). We identify these as the problem definition parameters in
involves two equations and two unknowns. For a tank of constant volumo' we
Table 8.9.
can eliminare nlight, hby substituting equation (8.a6) into (8.45) obtaining

FIan for Solving We will use a number of analytical relations from math C = c ¡,tLnYt ((V t r,r')r + r') (8.47)
and physics to estimate values of cost and volume as a function of radius
and height. Then we will check that the constraints are not violated.
Specifically, we will make sure that the thickness is exactly 10 mm, the Which simplifies to c =cuznyt(Y *r'l (8.48)

minimum volume of 10,000 liters, the footprint is less than 10 (m2) and
\nr )
the $10,000 maximum cost are not violated Since all the parameters in equation (B.aB) are constants except for radius, it is
a function of one variable. Using the necessary condition from calculus for
TABLE 8.9 Problem Definition Parameters for Storage Tank determining a stationary point we can take the first derivative of the cost
function with respect to the radius and set it to zero to obtain
Parameter Symbol Units Value

t cM $/kg dC
material cost 3.00
+( ,.znytft*,'ll =o (B 4e)
2 weight density I(g/m3 7,830 dr d,[-M-'t \o, ))
a
J wall thickness t mm 10
202 Engineering Design
Chapter 8 Parametric Design 203

=o (8.s0) TABLE 8.10 Storage Tank Design


#=c*znyt\#+zr)
Problem Definition Parameters
Parameter Symbol Units Yalue

Finding the oPtimal radius as r _l,,


-\u))"' (8.51) wt.
cost/unit cM $/kg a

wt density y Kg/m3 7830.0


thickness rmm 10

we next use the sufficient condition for an optimum to test the second
Design Variables
derivative as
Variable Symbol Units Lower Upper
radius r m 05
(8.s2) height h 515
#=l#*z),0 Solution Evaluation Parameters
m

zero for Parameter Symbol


Since the second derivative of the cost function is greater than Units Lower Upper

positive, non-zero values of r, we readily conclude that the stationary point r- cost C$ 5,000 10,000

This behavior is illustrated in Figure 8'13' volume V ttt3 10,000 none


is a minimum.
Performance calculations Satisfaction

.dnalyzing and Evaluating Recognizing that a minimum weight tank


can be V(l) r*(m) h*(m) w (kg) c ($) wcSc wvSv 0
designed using the optimál radius and height, we can develop a spreadsheet to 10000 1.17 2.34 2011.92 6035.75 4.52 0.00 0.52
comlute the üeight, cost and weighted satisfactions as a function of volume.
11000 1.21 2.41 2143.90 6431.74 0.46 0.07 0.53
12000 1,.24 2.48 227194 6815.82 0.41 0,14 0.55
As show, ln Table 8.10, we see that as the volume of the tank increases, 13000 1.27 2.55 2396,47 7789.41 0.37 0.21, 0.58
the weight and cost increase. The weighted satisfaction for cost decreases
as
14000 1.31 2.61 25t7.84 7553.52 0.32 0.28 0.60
the tank becomes more expensive. If we were to merely minimize cost, we
15000 1.34 2.67 2636.35 7909.06 a.27 0.35 0.62
would select a 10,000 liter dásign with radius 1..I7 (m) and height 2.34
(*). This 16000 1,37 2.73 2752.26 8256.78 a.n 0.35 0.58
decision would result in an overall satisfaction of 0.52. The weighted satisfac- 17000 1.39 2.t9 2865.78 8s97.33 0.18 0.35 0.53
tion forvolume increases with larger tank sizes reaching 0.35 for tank sizes of 18000 1.42 2.84 2977,09 893I.26 0.14 0.35 0.49
o'best?"
15,000 (liter) or more. Note how these trade-off. Which tank
size is the 19000 1.45 2.89 3086.35 9259.t¿5 0.10 0.35 0.45
20000 1.47 2.94 3193.72 9581.15 0.05 0.35 0.40

Cost versus Radius


If we were to make our final decision solely based on a single attribute,
we would select a 10,000 (liter) tank size. To maximize total satisfaction we
25000
should use overall satisfaction Q, which is a sum of the two weighted satisfac-
20000
\ tion values, Overall satisfaction reaches a maximum of 0.62 at a tank size of
15,000 (liter). This reflects the increase in volume for a marginal increase in

@ 15000
\ cost. A tank larger than 15,000 (liter), however, does not improve satisfaction
because the customer is indifferent (no change in satisfaction) for tanks larger
(r)

U 10000
\ than 15,000 (liter). Therefore we select a 15,000 (liter) tank with radius 1.34
\ á¿-¿ (*), height 2.61 (m) and a cost of $7,090. These trends are illustrated in
5000 Figure 8.14.
Lastly, we must also check that constraints are not violated. The thick-
0 ness is equal to 10 (**). The cost is less than $10,000. The tank volume is
0.0 0.5 1.0 1.5 z'0 2"5
greater than 10,000 (liter). And the footprint is nr2=n(134)2=5.6 (m2).
Radius (m) Our selected design maximizes satisfaction and does not violate constraints.

( a radius of 1.17 (m)' The


FIGURE 8.13 The cost for a 10,000 liter tank reaches a minimum for
( optimal radius r8 can be calculated using equation (8.51).

(
244 Engineering Design Chapter 8 Parametric Design 205

and Creveling, 1995). During parametric design, these variations in material


properties, ¿lii.ntionr, and opárating conditions can. be simulated' Alternative
Satisfaction versus volume
d.signs.un b. generated and analyzed for their sensitivity to these variations'
1.0 Two methods used quite fréquently are probabilistic optimal design and
as Monte
0.9
the Taguchi method. Probibilistic optimal design uses methods such
Carlo iimulation combined with nonlinear optimization algorithms to
0.8 design
É 0.7
parts with high levels of reiiability (Eggert and Mayne, 1993; Haugen' 1980;
€ 0.6
ilao, 19g4; SI¿¿utt, 1983). The Táguchi merhod exploits the use of statistics
# o.s
from design of experiments to reduce the sensitivity of products or
processes
€ 0.4
v) 0.3 to "noise" (i.e., variations) (Ross, 1996).
0.2
0.1
0.0
8.6 COMPUTER.A¡DED ENGINEERING
14000 16000
and hard-
Computer-airled engineering (CAE) refers to computer software
Volume (m3) ware systems used in the unutyilr of engineering designs to validate functional
performance. Computer-aideá design (C¡.O) is often used interchangeably
TOtal Qg5[ --.ri-- VOiUme withCAE.
-¡ -
CAEsystemscanbeseparatedintofourmaincategories:dynamicsanal.
ysis,finite element analysis (FEA), general purpose,,and other'
FIGURE 8.14 Weighted satisfaction for cost decrease as tank size increases. Similarly the that is,
weighted satisfaction for volume increases with tank size. The overall satisfaction reaches a Dynamics analyses involves'the kinematics and kinetics of bodies,
the motion and the forces and tolques that cause motion. We find
that solid-
maximum for a tank size of 15,000 liters.
and that we
modeling packages, for example, will mate parts into an assembly,
can maliipulate-uu.iout parts to animate thei¡ relative motion' However' in a
package such as Working Model, ADAMS, or DADS' we
8,5 DESIGN FOR ROBUSTNESS dynamics analysis
aiply specifi"á loais to the parts and the package will calculate their resultant
A robust product is one whose performance is insensitive to variations.
,ótio, usirg fundamental iquations of physics and a variety of nume,rical
Variations come from a variety of sources including: manufacturing, wear, and
methods. Th-ese packages simulate behavior based on physics, whereas
cAD
solid-modeting pactages animate motion, independent of physics'
the operating environment. When a product is manufactured, minor variations Since
in material composition can occur. Sometimes an alternative, less costly, kinematics a;d-kinetics-based packages are based on the physics of the
material with similar mechanical properties is purchased. In other cases, we
problem, they can arralyze a number of behaviors including:
might heat-treat or surface-coat the material in manufacturing processes that
have resulting variations. And, as important, most of the manufacturing
. positions,velocities,accelerations,
processes, such as machining, casting, and sheet metal stamping, will produce
. contacts and coliisions,
variations in final dimensions.
. joint forces, shaking forces, and
As the product is used, its moving parts undergo abrasive and adhesive
. relative motions.
wear. Also, some materials have properties that degrade over time. Steels be- a part
Finite element analysis (FEA) is a method that essentially divides
come brittle with exposure to hydrogen gases and some polymers become into smaller discrete elements to analyze the functional performance of a part'
opaque, etc. Then, we might also consider situations when the customer abuses used in many engineering fields, the process begins with the creation of a geo-
the product, by subjecting it to higher-than-expected loads. metric modei using CAD software. §oüd-modeling packages can expof
files
The operating environment can also cause variations in performance. Ex-
to FEA packages and some can communicate directly with FEA packages'
cessive humidity, dust, dirt, and extremely hot or cold environments can alter
The part is usuilly subdivided into a mesh of smaller finite elements connected
product performance. Steel, for example, can be brittle at sub-zero. Corrosive
at nodes. Loads are then virtually applied to the part' Using fundamental
gases or vapors, such as in chemical plants or food-processing plants, can also
equations from physics and engineering, an FEA package can analyze
a
cause performance degradation.
number of behaviors including:
Design for robustness is a term used to describe a number of methods
aimed at reducing the sensitivity of product performance to variations (Fowlkes

You might also like