UNDERSTANDING SYSTEMS ENGINEERING THROUGH CASE …sysengr.engr.arizona.edu/publishedPapers/CaseStudies.pdf · UNDERSTANDING SYSTEMS ENGINEERING THROUGH CASE STUDIES ... computed for

The Journal of NCOSE, Volume I, Number I, July-September 1994

UNDERSTANDING SYSTEMS ENGINEERINGTHROUGH CASE STUDIES

A. Terry Bahill, Ph.D.University of Arizona

Tucson, AZ 85721

William L. Chapman, Ph.D.Hughes Aircraft Company

P.O. Box 11337Tucson, AZ 85734

Introduction

Technical products define moderncivilization. The ability to create theseproducts quickly, accurately, and cost-effectively is what makes corporationssuccessful. There is a process for designingtechnical systems that is critical forcorporations to correctly understand and use.This paper analyzes case studies to helpunderstand the system design process.

Case Study Approach

While researching the system design process,it became necessary to develop a series ofcase studies of actual technical designs.Through the use of case studies we can createa bridge between systems theory [25], [55],[32], [37], [47], and actual design efforts.The design studies are grouped as illustratedin Figure 1.

This figure compares design difficulty versusthe resources used to create the designs. Toset a common limit, we have decided that thesystem design process ended when the firstproduction unit was built. The areas namedreflect the type of design needed. The largestarea, Consumer Products, is characterized bya design difficulty that is small to moderateand requires a small to moderate amount ofresources. The area that requires themaximum amount of resources, SevenWonders of the Ancient World, has a designdifficulty that is small to moderate. The upperleft quadrant is called Star Wars after the1977 movie. It indicates items that we canimagine, but probably cannot build becauseof the complexity of the product. The designsare so difficult that it is impossible to solve

these problems; thus, they remain intract-able.The final region is for high designdifficulty coupled with massive resources.This quadrant is called Moon Landing, toindicate the enormous nature of both thedesign effort and the resources needed.

Each design was categorized by a scorecomputed for each case study. The scoresreport composite scores of the constituentparts of each axis on the graph in Figure 1.1.Each constituent part is an ordinal rankingwithin the category. We recognize thatordinal rankings are not usually additive,however in this case we have adjusted thecategories so that the answers all pass areasonableness test. Extreme examples can beconjured up that will not fit these rankings,but this scale fits these specific case studies,although it is not an unassailable system.

The scores for the vertical axis of the graph,Design Difficulty, represent a combination ofthe following categories:

(1) Design type, which is a continuum fromredesign to original innovative designand finally to breakthrough design.

(2) Complexity of the knowledge needed tocreate the design.

(3) Number of steps needed to complete thedesign.

(4) Quality of the product.

(5) Quantity to be built from the design andthe expected sales price of the product.

145

Understanding Systems Engineering Through Case Studies

DESIGN

DIFFICULTY

Moon Landing

Seven Wonders

of the Ancient World

RESOURCES

Figure 1 Categories of Design

Each case study was scored using the scaleillustrated in Table 1. Many more categoriescan be created. For example, we found thatthe expected system life was another usefuland orthogonal metric, but we chose not toincorporate it here. We decided to derive aminimal set that would be useful for peopleembarking on a new design project.

By choosing ranges for these categories wehave in effect created the weights ofimportance for each.

Design type reflects whether feasiblesolutions exist, and how much originalthought goes into the project. A score of 1-6is awarded for continuous improvement. Ascore of 7-13 is given for original innovativedesign. A score of 14-15 is given for abreakthrough design effort.

Complexity of the knowledge needed tocreate the design is very hard to quantify. It is

determined based on an estimate of thenumber and availability of the people with thenecessary knowledge to do the design. A 1 or2 is given for common knowledge held bymany people, 3-5 for complex knowledgeheld by a sufficient pool of people, 6-8 forcomplex knowledge held by few people, and9 or 10 for undiscovered knowledge that canonly be found by specialists.

The number of steps needed to complete thedesign is defined as the number of discretesteps needed to design the system. It isrelated to the number of major components ormajor process steps that are needed toassemble the system. A 1 or 2 is assigned forany system with fewer than 50 steps orcomponents. A 3 or 4 is assigned for systemsup to 500 steps or components. A 5-8 isassigned for systems with more than 500 butless than 10,000 steps or components. And 9or 10 is assigned for systems with greaterthan 10,000 steps or components.

Table 1 Design Difficulty Score

TYPE

Range of 1-15

KNOWLEDGECOMPLEXITY

Range of 140

STEPS

Range of 1-10

QUALITY

Range of 1-10

QUANTITYAND PRICE

Range of 1-5

DESIGNDIFFICULTY

TOTALRange of 5-50

146

The Journal of NCQSE, Volume I, Number I, July-September 1994

Quality is measured by how closely the finalproduct must match the customer's specifiedtarget values. Customers provide a desiredoperating requirement. How well the systemadheres to this requirement over its expectedlife is the quality measure. A score of 1-3 isgiven for little expectation of compliance. Ascore of 4-8 is for increasingly tightexpectations, and a 5 is given for strict,unyielding expectations. It was very difficultto assign scores for this metric, that wouldsatisfy all of the potential customers.

The quantity to be produced from the designis the production build number. Bydefinition, the system design process endswith the creation of the first production unit.The number of systems built after the firstunit is of interest to analyze the level ofdifficulty to create the manufacturingprocesses. If a large quantity of items mustbe built, more effort is exerted in designingthe manufacturing system in addition to theproduct, thus yielding a more complex designproblem.

The price of the final product is related to thebuild quantity, and becomes a designconstraint. If there is a limit on the maximumselling price through market or other forces,then the design will be more difficult. A 1 isassigned for three or less units built, 2 for upto 100 units, 3 for up to 1,000 units, 4 for upto 5,000 units, and 5 is for greater than 5,000units built. Perhaps Quantity and Priceshould have been two separate metrics, butoften they are closely related, so wecombined them.

The scores of the horizontal axis of thegraph, Resources, represent a compositescore of the following categories:

(1) Costs to develop the product throughthe first production unit.

(2) Time from the beginning of the effortthrough the first production unit.

(3) Infrastructure required to complete thedesign.

The score for each case study was on thescale illustrated in Table 2.

The Resources Axis.

Cost is the amount needed to pay fordevelopment, including salaries, utilities,supplies and materials, through the firstproduction unit. This is not in absolutedollars, but in terms of the payer's ability topay. For example, it is easy for a rich man toafford a VAX computer, but not a personwith an average salary. Combine 1,000average salaries and these people can nowafford a VAX. For the U.S. government aCray computer has low cost, but for mostpeople a 486 PC is expensive. A 1 or 2 isassigned for affordable systems, 3-8 formoderately expensive systems, 9-13 are forvery expensive systems that are developedrarely, and 14 or 15 are for massivelyexpensive systems requiring major sacrifices.

The time score is for time spent from thebeginning of the effort to define thecustomer's needs through the first productionunit. A 1 is given for less than 3 months, 2for around 6 months, 3 for a year, 4-7 for upto 5 years, 8-9 for up to 8 years and a 10 formore than 8 years.

Table 2 Resources Scores

COST

Range of1-15

TIME

Range of 1-10

INFRASTRUCTURE

Range of 1-10

RESOURCESTOTAL

Range of 3-35

147


Infrastructure required to achieve the designis also hard to quantify. Infrastructure isdescribed as the physical resources neededfor construction (including machine tools,process shops, and assembly workstations),transportation, communication, utilities, thelaws and legal protections, the skilledmanagers, and the educational systemavailable. Infrastructure must be judged inregard to the designer's ability to get and usethe infrastructure over the needed designtime. A 1 or 2 is assigned if it is a common,low cost infrastructure (e.g. clean tap waterin the U.S.). The numbers 3-5 are given formoderate infrastructures requiring people onthe project to support it, 6-8 for large,complex infrastructures requiring largeportions of the cost of the entire project, and9 or 10 for a massive infrastructure requiringmajor portions of the available labor forceand the available equipment.

All of the measures depend heavily on thecontext of the design effort. It is impossiblefor a man in a primitive country to obtain theresources necessary to build a telephone in areasonable time. It is trivial to do this inAmerica. It may be impossible for a smallcompany to obtain $10 million to fund a newproduct. It is easy for General Motors. At theend of every case study the context for thescores will be given to help explain thereason for the scores.

Case Studies

In the following paragraphs we give a briefdescription of our 17 case studies.References and a fuller discussion are givenin [24].

(1) Resistor Network - A serial or parallelcombination of up to four resistors,designed to achieve a specific resistancevalue. Designing a resistor network isan NP-complete math problem, but issimple if only four resistors are used.[35].

(2) SIERRA Train Controllers - Students atthe University of Arizona havedeveloped numerous versions of acontroller to run two toy trains. Thecontroller must prevent collisions and

can be accomplished with three to sixstate machine designs. Despite beinglimited to nine components, studentshave found dozens of successfulsolutions. [25],

(3) Bat Chooser - Bat Chooser is a smallconsumer product developed todetermine the ideal bat weight for anindividual baseball player. Bymeasuring the swing speed of a givenbat, the ideal weight of bat for theplayer can be selected. [22].

(4) Pinewood Derby - A Pinewood Derbyis a Cub Scout race of wooden cars.Creating a round robin schedule formatfor 15 cars and 3 lanes proved to be anNP-complete math problem. Severalbright engineers spent an inordinateamount of time chasing a problem withno solution. [25], [19].

(5) Second Opinion - An expert system thatruns on a personal computer andprovides an evaluation of childhoodstuttering. It provides a second opinionto clinicians who are evaluatingchildren. [20], [21].

(6) American Airlines Scheduling -American Airlines schedules its aircraftand crews using computer algorithms.These sophisticated algorithms search afast solution, but not necessarily theoptimum. [15], [31], [51], [16].

(7) Superconductors - Superconductors arematerials that exhibit no electricalresistance when very cold. Recentdevelopment of advanced copper oxideswas a breakthrough design effort. Infour months Paul Chu of the Universityof Houston raised the world recordfrom 30°K to 90°K. [36], [12], [13],[8], [4], [41].

(8) Incandescent Light Bulb - In 1879Thomas Edison developed the firstcommercially viable light bulb. He wasridiculed throughout the effort by themainstream press, engineering andscientific communities. His break-through occurred in two months andincreased the life of a bulb from 13hours to 560 hours. [26].

148


(9) Boeing 111 - The 777 is a newcommercial aircraft being designed bythe largest airplane manufacturer in theworld. Boeing is spending $4 billionand using a new design process,centered on teaming and computermodels. [2], [1], [49], [18], [23].

(10) The Apollo Moon Landing - The Apollomoon landing was one of the mostdifficult and costly projects mankindhas ever undertaken. The rockets werea major portion of the design effort.The development of the Saturn V rocketused only two prototype launchesbefore manned flight, compared to 91prototypes for the Mercury program'sAtlas rocket. [39], [29], [44], [52].

(11) A House - The family home in Americais easy to design, but it is a majorinvestment. Modifying existing designsis the normal design approach. [38],[45].

(12) Central Arizona Project - The CAP is a336 mile aqueduct built from theColorado river to the central Arizonacities of Phoenix and Tucson. It cost$4.7 billion and took more than 20years to build. [17], [46], [43], [14].

(13) The Great Pyramid at Giza - One of theSeven Wonders of the Ancient World,the Great Pyramid is still admired for itsengineering. It was built in 2575 B.C.and took 20 years and 50,000 laborers.[HL [40].

(14) A New Car - The redesign of new carsis one of the most costly ventures inmodern industry. Japanese andAmerican corporations do it differently.The Japanese spend on average 46months, while the Americans take over60 months. The difference can mainlybe attributed to the Japanese use ofrapidly developing prototypes. [53],[33], [34], [27].

(15) The GM Impact: An Electric Vehicle -The Impact is an electric vehicledesigned with the sponsorship ofGeneral Motors. It has impressivespecifications for a battery-operated car.[30], [10], [9], [28], [5], [54], [6], [7].

(16) Batteries for Electric Vehicles -Batteries are preventing the widespreaduse of electric vehicles. A breakthroughin energy storage is needed. This maynot occur; instead, incremental designchanges may eventually give a decentbattery. [28], [42], [48], [54], [3].

(17) C3PO - This handy robot in the StarWars film is an intractable designproblem for 1994 America. [50].

Table 3 and Table 4 show the summaryscores for all of the case studies. Figure 2plots all of the case studies together on onechart. Many, many more case studies can beextracted from the literature. We were tryingto derive a minimal set that would be usefulfor systems engineers embarking on a newdesign project. We started by reading two orthree books and numerous articles on eachcase study. Next, we condensed the systemengineer's roles and the critical designrequirements into a case study ranging from 3to 10 pages in length. The scores for eachstudy were derived from multiple discussionsbetween the authors, students in graduatecourses at the University of Arizona, andemployees of Sandia National Laboratory.

Usefulness Of A System DesignMetric

This system of rating system design efforts isvaluable for several reasons. First, the bestway to instruct engineers on the methods ofsystem design is to use case studies. Designis as much an art as a science, and bystudying how well other design efforts weredone, lessons can be learned and applied tofuture projects. The evaluation schemepresented here is one method of showingwhat designs were successful given theresources available.

149


Table 3 Design Difficulty Scores For The Case Studies

Case Study

ResistorsSIERRABat ChooserPinewoodSecond OpinionAmerican AirlinesSuper ConductorLight Bulb111ApolloHouseCAPPyramidCarElectric VehicleImproved BatteryBreakthroughBatteryC3PO

Type

1263551514912245710815

15

KnowledgeComplexity

125264106671244769

10

Steps

111234239104546533

10

Quality

112232327103313323

5

Quantity andPrice

11113224321115245

4

DesignDifficulty

Total57151020173229344111151525272335

44

Table 4 Resources Scores For All Of The Case Studies

Case Study

ResistorsSIERRABat ChooserPinewoodSecond OpinionAmerican AirlinesSuper ConductorLight Bulb777ApolloHouseCAPPyramidCarElectric VehicleImproved BatteryBreakthrough BatteryC3PO

Cost

11.5222323131510121597352

Time

11326533793910741062

Infrastructure

1121.5354281076964453

ResourcesTotal33.575.51113982834202734221517167

150

DesignDifficulty


Resources versus Design Difficulty for the Case Studies

STAR WARS45 T •C3PO

40 • •

35 • • «Battery

•Superconductor30 " ' •Light Bulb

•ElectricCar25 • •Car

20 - •Second Opinion•AmericanAirlines

15 • •BatChooser

10 " • tPlnewood•Sierra

5 • • •Resistor

0 5 10

•House

CONSUMER PRODUCTS

15 20 25

Resources

MOON LANDING

•Apollo

•777

•CAP •Pyramid

SEVEN WONDERS OF

THE ANCIENT WORLD

30 35 40

Figure 2 Actual Scores For The 17 Case Studies Plotted In Relationship To Each Other

The second, and main reason for such asystem of measurement is to rate proposeddesign efforts before they even start.Corporations that must make a profit tosurvive must only attempt those projects inthe Consumer Products region. Any designeffort outside this region will requireresources outside the scope of most ventures.Those ventures in the Star Wars arena arebest left to academics and research labs,where lack of immediate success is notpunished. If your customer asks you to builda system that fits into the Star Wars quadrant,you should tell them at the beginning that itcannot be done with the allotted resources.Those that are considered one of the SevenWonders of the Ancient World are strictly forgovernments. Not only do these ventures notpay for themselves, but they also requiredespotic power to implement, since there isso little reward for the citizens involved. Andfinally, those in the Moon Landing arearequire government and industry involvementto succeed, since they require too large ashare of the national resources.

We expect systems engineers to look atrequirements for a new design, assemble aninterdisciplinary team of customers,designers, manufacturing engineers, sales,product support, etc. and develop aconsensus for the amount of resourcesrequired and the design difficulty. We believethat an analysis, such as that presented in thispaper, would lead the system design team todecide the scope of the project and theappropriateness of the resources to the task.Additional effort is now being directed atusing the case studies to decide whensystems design requirements must beformally documented and when a full timesystems engineer is needed for a design task.

REFERENCES

[1] "777 Revolutionizes Boeing AircraftDevelopment Process," Aviation Week& Space Technology. June 3, 1991,pp. 34-61.

[2] "Boeing Selects Design for 777Candidate," Aviation Week & Space

151


Technology. December 18, 1989, pp.106-107.

[3] "Chrysler Corporation Plans To Showa Model of Flywheel Auto," Wall StreetJournal Jan. 3, 1994, pp. 10.

[4] "Cooperating on superconductivity,"Spectrum. August 1992, pp. 50-

55.

[5] "Electric Car Pool," ScientificAmerican. May 1992, pp. 126-127.

[6] "Electric Vehicles," IEEE Spectrum.November, 1992, pp. 18-24.

[7] "EV Watch," IEEE Spectrum.December 1993, pp, 65.

[8] "Fact vs Fantasy: CeramicSuperconductors Scrutinized," IEEESpectrum. Vol. 25 No. 5, May 1988,pp. 30-41.

[9] "GM Drives The Electric Car Closer toReality," Business Week. May 14,1993, pp. 60-61.

[10] "GM IMPACT," Discover. March1992, pp. 91-98.

[11] "Pyramids," Encyclopedia Britannica.New York, 1972.

[12] "Superconductors," ScientificAmerican. December 1993, pp. 119-126.

[13] "TRW Superconductor ReducesSatellite Power," Aviation Week &Space Technology. December 6, 1993,pp. 56.

[14] "Who will pay for CAPl,"TucsonCitizen. 11 Dec 1993.

[15] J. Abara, "Applying IntegerProgramming to the Fleet AssignmentProblem," Interfaces. 19:4, 1989, pp.20-28.

[16] R. Anbil, E. Gelman , B. Patty and R.Tanga, "Recent Advances Crew-Pairing

Optimization at American Airlines,"Interfaces. 21:1, 1991, pp. 62-74.

[17] Arizona Department of WaterResources , "Impact of GroundwaterManagement Act and CAP WaterSupply on Agricultural WaterConservation Programs," Proceedingsof the 20th Anniversary Conference onWater Management in the 90's. SeattleWA., 1993.

[18] S. Ashley, "Boeing 777 Gets a BoostFrom Titanium," Mechanical Engineer-ing. Vol 115, No. 7, 1993, pp. 60-65.

[19] A.T. Bahill and WJ. Karnavas,"Reducing State Space Search Time:Scheduling in the Classic AIChallenge," AI Expert. Sept 1993, pp.29-35.

[20] A.T. Bahill, K. Bharathan, and R.F.Curlee, "How the Testing Techniquesfor an Expert Advisory SystemChanged Over Eight Years,"Proceedings IEEE InternationalConference on Systems, Man andCybernetics. Le Touquet, France,October 17, 1993.

[21] A.T. Bahill, K. Bharathan, and R.F.Curlee, "Making an Expert SystemChanges the Expert," Recent Trends inResearch, Education, and Applications,Volume 4, Eds. M. Jahshidi, R.Lumia, J, Mullins and M. Shahinpoor,ASME Press, New York. Proceedings4th International Symposium onRobotics and Manufacturing, Santa Fe,November 11-13, 1992, pp. 711-716.

[22] A.T. Bahill and WJ. Karnavas, "TheIdeal Baseball Bat," New Scientist.April 6, 1991, pp. 26-31.

[23] D. Brooks, "Engineering SimulatorsEnhance 777 Development,"Aerospace America. Vol. 31, No. 8,1993, pp. 30-32,41.

[24] W.L, Chapman, The System DesignProcess is Intractable, but Robust.

152


Ph.D. Dissertation, Tucson, Arizona:University of Arizona, 1994.

[25] W.L. Chapman, A.T. Bahill, andA.W. Wymore, Engineering Modelingand Design. Boca Raton: CRC PressInc., 1992.

[26] R.Conot, A Streak of Luck, ABiography of Thomas Alva Edison.New York: Seaview Books, 1979.

[27] D. DeVera, T. Glennon, A.A. Kenny,M. Kham, and M. Mayer, "AnAutomotive Case Study," in QualityProgress. June 1988, pp. 35-38.

[28] P. Dvorak, "The Shocking Truth AboutElectric Vehicles," Machine Design.September 21, 1989, pp. 86-94.

[29] Editors of Time Magazine , The SpaceIndustry, America's Newest Giant.Englewood Cliffs, N.J.: Prentice-Hall,1962.

[30] Electric and Hybrid Vehicle Advance-ments. Society of AutomotiveEngineers, 400 Commonwealth Dr.,Warrendale, PA., March 1993.

[31] I. Gershkoff, "Optimizing Flight CrewSchedules," Interfaces. 19:4, 1989, pp.29-42.

[32] J. Grady, System Integration. BocaRaton: CRC Press Inc., 1994.

[33] D. Halberstam, The Reckoning. NewYork: William Morrow & Co., 1986.

[34] D, Halberstam, The Next Century.New York: William Morrow & Co.,1991.

[35] Handbook of Electrical Engineering.Boca Raton: CRC Press, 1992.

[36] R.M. Hazen, The Breakthrough, TheRace for the Superconductor. NewYork: Summit Books, 1988.

[37] B. Hollins and S. Pugh, SuccessfulProduct Design. Boston: Butterworths,1990.

[38] P. Hotton, So You Want To Build AHouse. Toronto: Little, Brown &Company, 1976.

[39] R.S. Lewis, Appointment on theMoon. New York: Viking Press, 1969.

[40] D. Macaulay, Pyramid. Boston:Houghton Mifflin Co., 1975.

[41] A.P. Malozernoff, "SuperconductingWires Get Hotter," IEEE Spectrum.December 1993, pp. 26-30.

[42] H. Oman, "28th Intersociety EnergyConversion Engineering Conference,"in IEEE AES Systems Magazine,November 1993, pp. 2-8.

[43] D.B. Paul and M. McDonald, "Use ofgeographic information systems fordesign and construction of newWaddell Dam," InternationalSymposium on Mapping andGeographic Information Systems. SanFrancisco, CA, June 21-22, 1990, pp.183-192.

[44] C.R. Pellegrino and J. Stoff, Chariotsfor Apollo: The Making of the LunarModule. New York: Atheneum, 1985.

[45] Projects and Plans. Elmhurst, Illinois:Craft Pattern Studios, 1976.

[46] D.C. Rogers, T.F. Kacerek and R.S.Gooch, "Field data for verifying canalunsteady flow models," Journal ofIrrigation and Drainage Engineering.Vol. 199, No. 4, Jul-Aug 1993, pp.679-692.

[47] A.P. Sage, Systems Engineering. NewYork: Wiley-Interscience, 1993.

[48] O. Suris, "Ford Plans Rely On aBattery Seen as Risky,"Wall StreetJournal. December 17, 1993, pp. B2.

153


[49] C.B. Tenning, "Evolution of theBoeing 777 electrical power system," inProceedings of the 27th Inter societyEnergy Conversion EngineeringConference, San Diego, CA, August 3-7, 1992.

[50] Twentieth Century Fox, Star Wars,1977.

[51] A. Valasquez Marquez, "AmericanAirlines Arrival Slot Allocation System(ASAS)," in Interfaces, Vol. 21, No,1, 1991, pp. 42-61.

[52] T, Wolfe, The Right Stuff. New York:Farrar, Straus & Giroux, Inc., 1979.

[53] J.P. Womack, D.T. Jones, and D.Roos, The Machine That Changed theWorld, New York: Rawson Associates,1990.

[54] F.A. Wyezalek, "Heating and CoolingBattery Electric Vehicles - The Final

Barrier," IEEE AES SystemsMagazine, November 1993, pp. 9-14.

[55] A.W. Wymore, Model-based SystemsEngineering. Boca Raton: CRC Press,1993.

AUTHOR BIOGRAPHIES

A. Terry Bahill is a Professor of Systems andIndustrial Engineering at the University ofArizona. He has written six books andnumerous journal articles. He is a Fellow ofthe IEEE. Terry received his Ph.D. inElectrical Engineering from the University ofCalifornia at Berkeley.

William L, Chapman has been an engineer atHughes Aircraft Company for 15 years. Hehas worked in Manufacturing, Design andTest Equipment. He holds a patent on a radarplatform and is the author of a collegetextbook on systems engineering. Hereceived his Ph.D. in Systems Engineering atthe University of Arizona.

154

Documents

UNDERSTANDING SYSTEMS ENGINEERING THROUGH CASE …sysengr.engr.arizona.edu/publishedPapers/CaseStudies.pdf · UNDERSTANDING SYSTEMS ENGINEERING THROUGH CASE STUDIES ... computed for