The probe Some material used from:. EPMA - electron probe microanalysis Probe signals.
blank - NASA€¦ · PV = Pioneer Venus PVLP = Large Probe PVSP = Small Probe PVBO = Bus/Orbiter...
Transcript of blank - NASA€¦ · PV = Pioneer Venus PVLP = Large Probe PVSP = Small Probe PVBO = Bus/Orbiter...
COST ESTIMATION MODEL FOR ADVANCED
PLANETARY PROGRAMS - FOURTH EDITION
Report No. SAI 1-120-768-C9
by
Daniel J. Spadoni
Science Applications, Inc.1701 East Woodfield Road
Schaumburg, Illinois 60195
for
Earth and Planetary Exploration DivisionOffice of Space Sciences and Applications
NASA HeadquartersWashington, D.C.
Contract NASW-3035
April 1982
https://ntrs.nasa.gov/search.jsp?R=19840014522 2020-05-12T15:11:07+00:00Z
,
blank
FOREWORD
This study represents a portion of the work performed by Science
Applications, Inc. within Task 2: Cost Estimation Research of Contract
No. NASW-3035 for the Earth and Planetary Exploration Division (Code EL/4)
of OSSA/NASA Headquarters. The results are intended for use as a
decision-aiding tool to assist NASA in its development of long-rangemission plans for solar system exploration.
The author wishes to express his gratitude to those individualsboth within NASA and the industrial community who graciously provided theinformation vital to his study.
page i(eft
(eft blank
TABLE OF CONTENTS
Page
FOREWORD iii
ACRONYMS AND ABBREVIATIONS . . . vii
1. INTRODUCTION AND SUMMARY l
1.1 Background and Study Objectives 1
1.2 Cost Model Overview 2
1.3 Summary of Results 3
2. COST MODEL DATABASE 5
2.1 Development Project Cost Data 5
2.2 Development Project Technical Data 10
3. MODEL DEVELOPMENT 13
3.1 Labor/Cost Proxy Analysis 14
3.2 Functional Forms 17
3.3 Test Statistics 19
4. MODEL ANALYSIS 21
4.1 Labor/Cost Conversion Factors 274.2 Error Analysis 27
4.3 Simulation Analysis 31
4.4 Error Analysis Revisited 34
4.5 Benchmark Tests . 36
5. SAMPLE APPLICATIONS . 41
6. CONCLUSIONS AND FURTHER DEVELOPMENT 51
REFERENCES 53
APPENDIX A: Detailed Cost Database
APPENDIX B: Detailed Description of Cost Model Algorithms
APPENDIX C: Inheritance Model
APPENDIX D: Detailed Error Analysis
Page intentionally left blank
Page intentionally left blank
Acronyms and Abbreviations
Database Flight Programs
M64 = Mariner Mars 1964SUR = SurveyorLO = Lunar OrbiterM69 = Mariner Mars 1969M71 = Mariner Mars 1971PJS = Pioneer Jupiter/Saturn (10/11)M73 = Mariner Venus/Mercury 1973VLC = Viking Lander CapsuleVKO = Viking OrbiterVGR = VoyagerPV = Pioneer Venus
PVLP = Large ProbePVSP = Small ProbePVBO = Bus/OrbiterPVS = Science Instruments
Cost Model Categories
STD = Structure and DevicesTCP = Thermal Control, Cabling and PyrotechnicsPRP = PropulsionAAC = Attitude and Articulation ControlTCM = TelecommunicationsANT = AntennasCDH = Command and Data HandlingPWR = Radioisotope Thermoelectric Generator (RTG) PowerPWS = Solar/Battery PowerADM = Aerodeceleration ModuleRDR = Landing Radar/AltimeterIML = Line-Scan ImagingIMV = Vidicon ImagingPFI = Particle and Field InstrumentsRSI = Remote Sensing InstrumentsDSI = Direct Sensing/Sampling InstrumentsSYS = System Support and Ground EquipmentL30 = Launch + 30 Days Operations and Ground SoftwareIDD = Imaging Data DevelopmentSDD = Science Data DevelopmentPGM = Program Management/Mission Analysis and EngineeringFO = Flight OperationsDA = Data Analysis
Cost Model Parameters
N = Number of Flight Qualified UnitsDLH = Direct Labor Hours (1000 hours)NRL = Non-recurring Labor Hours (1000 hours)RLH = Recurring Labor Hours (1000 hours)URL = Unit Recurring Labor Hours (1000 hours)M = Subsystem Mass (kilograms)MD = Mission Duration (months)ED = Encounter Duration (months)PPL = Imaging Resolution (pixels per line)
vn
Cost Estimation Model for AdvancedPlanetary Programs - Fourth Edition
1. Introduction and Summary
1.1 Background and Study Objectives
In the decade of the 1980's, the United States' program for
unmanned exploration of the solar system faces increased competition
for the resources required for the achievement of its goals. Oneimportant implication of this situation is that the long-range mission
planning process will involve a greater degree of selectivity than was
seen in the past. This in turn implies that the total cost ofindividual missions must be forecast with a greater sense of confidence
than ever before.
Several techniques are used to develop cost estimates of futuremissions at the pre-Phase A level of mission difinition. Engineering,
or "grassroots", estimation generates cost estimates at the lowest
level of the project's work breakdown structure defined at the time of
the estimate. Analogy estimation derives costs by comparing mission
hardware and scenario definitions with those of similar past projects
and suitably adjusting the known, historical costs for such factors as
differences in requirements and capabilities and for inflation. Model
estimation uses cost estimating relationships (functions relating cost
to requirements/capabilities), derived from historical data, to predict
future costs. In essence, model estimation quantifies the analogy
costing process.
For nearly a decade, Science Applications, Inc. (SAI) has been
involved in cost estimation and analysis of the U.S. planetary exploration
program. The work has encompassed historical cost data collection and
analysis, development and refinement of a cost estimation model based
on the historical data (References 1 and 2), and extensive use of the
model for predicting costs of future missions.
This report discusses the development of the current versionof the SAI Planetary Program Cost Model. The Model was updated toincorporate cost data from the most recent U.S. planetary flightprojects and extensively revised in order to more accurately capturethe information in the historical cost database. The revision was madewith a two-fold objective: to increase the flexibility of the Modelin its ability to deal with the broad scope of scenarios under consid-eration for future missions, and to at least maintain and possiblyimprove upon the confidence in the Model's capabilities with an expected
accuracy of ±20%.
1.2 Cost Model Overview
The SAI Planetary Program Cost Model can be characterized by
the following features.
t The Model is based on all relevant U.S. planetaryprojects from Mariner Mars 1964 through PioneerVenus.
• Inputs to the Model are limited to informationgenerally available at the level of pre-Phase Amission definition. Generally, these consist ofestimates of spacecraft subsystem masses, designheritage, flight time and encounter duration.
• The primary output is manpower, expressed in directlabor hours. Total cost is obtained by use ofappropriate conversion factors which includeinflation indices.
t The Model views a mission program as consisting oftwo distinct phases: The Development Project, whichencompasses all activity through the mission'slaunch + 30 days milestone and the Flight Project,which includes all activity from L + 30 days throughthe nominal end of mission.
• At its most detailed level, the Model deals withcost categories which are derived as compromiseaggregations of the variety of work breakdownstructure definitions found in the cost database.
• The Development Project is further separatedinto hardware-related cost categories andfunctional support cost categories. The hardwarecategories are directly related to the missionspacecraft engineering and science subsystems.
• Hardware categories are further separated intonon-recurring costs (design and development)and recurring costs (fabrication and subsystem-level tests). Inheritance is assumed to affectonly the non-recurring cost,
• The Model is capable of dealing with a wide varietyof spacecraft designs, including inertia! or spinstabilized spacecraft, atmospheric entry probesand highly automated soft landers.
1.3 Summary of Results
The model development effort resulted in an updated and revised
Cost Model which adequately meets the objectives set forth in Section
1.1. Only the Development Project portion of the model was revised;
cost estimates for the Flight Project are generated using algorithmsfrom the previous version of the Model (Ref. 2).
A total of 21 revised cost categories were defined, 16 related
to flight hardware and five to functional support. Two separatealgorithms were derived for each hardware category: one which estimates
total direct labor and another which estimates recurring labor.
Non-recurring labor can be obtained by differencing the two estimates.
The hardware labor algorithms are, in general, power laws or exponential
functions of a single independent variable formed by the product of
the number of flight units and the subsystem (category) mass.
Statistical analysis of the historical cost data resulted in a
conclusion that factors derived as simple ratios can be used to convert
category labor hour estimates to total cost.
An extensive error analysis of the Model measured against theprograms in the database indicated that the information in the database
had been captured with an average error of less than 10%. However,
a simulation of the Model's performance, with number of flight units
as the parameter, showed that predictions made with the Model would
be highly sensitive to the number of flight units. A straight-forward
adjustment procedure was devised that effectively eliminates this
sensitivity but results in an increased average error of just less
than 20% as measured against the database.
2. Cost Model Database
Historical cost data for thirteen unmanned lunar and planetary
flight programs currently comprise the SAI cost model database. Table 1summarizes the present status of this database. For use in modeldevelopment, total program costs were segregated into two independentparts. The first, termed the development project, includes all programcosts incurred through the launch + 30 days milestone. All programcosts after this milestone are termed the flight project. Note thatsome programs in the database have multiple L + 30 milestones that arewidely separated in time (e.g., Pioneer Jupiter/Saturn with launches inMarch, 1972 and April, 1973). This does not present a problem insegregating the costs since it is a simple matter of continuing to trackhardware development of follow-on units after the first launch date.The indications in Table 1 regarding use of the data in model revisionwill be discussed in Section 3.
x
2.1 Development Project Cost Data
Cost data for the programs in Table 1 up to and including Voyagerwere used in developing the previous version of the SAI cost model(Ref. 2 ). At the time, however, the Viking Lander, .Viking Orbiter
and Voyager (then called Mariner Jupiter/Saturn) development projectshad not been completed and the cost data used in modeling were based onestimates to complete. Thus, prior to the present model revision effort,it was necessary to analyze and reduce the actual completion costs whichhad been collected for these three programs into forms useful formodeling.
During this process of data reduction, two issues concerning the
data and its use in modeling became apparent. First, some allocationsof raw cost data into the model's cost categories did not appear to beconsistent. Second, the assumption used in the previous model forseparating non-recurring and recurring costs no longer appeared to be
valid. Both of these issues made it necessary to reevaluate specificelements of the entire database.
LUOO
CQ
<u
OOO
CD
00
I—
OO
O I—i—i c_>OO LU>—i *~3> OLU C£o; Q.
Q LUo 2:
LU LU00 Q
00
•-0 QOCXL OQ- 00
I—I 00—I OU_ Q.
o; cc: ce: o o o< < ec I— I— I—Q. Q- Q.
o o o o
i—i i—i X X
CJLU"-2O
Q
.
~o_
-
^C£
OCCQ.
o:LU
i i,2.
DCO
LU
o;ID00
o;LU1—t— 1CQrvO
ct:
— ^_i
CTlUD
—C£.LU
^o:^2T
i— ij *
—ceLUz:2c2!
<Cooo:LUI—i— io."~2
rvLULU
OI— IQ.
COr--.—C£LUz:o:rs!
o:LUQ
t_J
C_3z:2HH
=»
cr:LU1—1— 1CQo;oC3
^2i — i>
00~
^LU
>
fV CCLU LUC3 LU
>- Oo >-•>• Q_
o:LUi—t-H
CQC£.O
OLU—1_J
^CD
LUCQOCC0.
oLU
1
^C3
oo l—00 LULU >-
LU CHI— ta =tLU O I—_i a: <ca. Q- QO z o
O i—i X
During the process of examining cost allocations, a decision wasmade to broaden and redefine the model cost categories. As with previousmodel versions, these categories are separated into two related areas:flight hardware categories and functional support categories. Thesecategories are defined to be compatible with the wide variety of workbreakdown structure definitions used by the system contractors who
develop the mission hardware. Each specific category definition wasarrived at through an iterative process involving both the cost dataallocation and statistical modeling efforts.
The flight hardware-related categories are defined as follows:t Structure & Devices - Spacecraft main structure, support
trusses, adapter, scan platform, booms, solar panelstructure, miscellaneous mechanisms and other hardware,ballast, bioshield, pressure vessel, landing gear,HGA structure.
• Thermal Control, Cabling & Pyrotechnics - Passive andactive temperature control, cabling and wire harness,pyrotechnic devices.
• Propulsion - Propulsion system inerts.
• Attitude & Articulation Control - Celestial and inertia!sensors, attitude control electronics, articulationdevices and actuators.
• Telecommunication - Transponder, receiver, transmitter,telemetry, modulation/demodulation.
• Antenna - S/X antenna, omni's, low and medium gainantennas, waveguides, feeds, rotary joint.
• Command & Data Handling - Command computer & sequencer,flight data, data storage.
• Power - Solar cells & slide covers, battery, conditioning
and distribution (does not include RTG units).
t Aerodeceleration - Heat shield, aeroshell, parachute andmortar.
• Radar - Altitude marking/terminal descent radarantenna(s) and electronics, radar altimeter.
• Imaging - Camera and electronics (vidicon or line scan).
• Particle & Field - Magnetometers, high-energy radiation,plasma, micrometeroid sensors.
• Remote Sensing - Radiometers and spectrometers.
• Direct Sensing & Sampling - Atmospheric and surfaceinstruments.
Similarly, the functional support categories are defined asfollows:
• Program Management/MAE - Project management and control,
administration and support staff, division reps, preflighttrajectory and navigation analysis, mission engineering,ephemeris development, planetary quarantine support.
• System Support & Ground Equipment - Spacecraft designteams, system configuration, system assembly and testing,quality assurance, reliability, safety, electronic partsacquisition and screening, mission and test computers,ground data system, ground data handling, ground handlingequipment.
• Launch + 30 Days Operation & Ground Software - ETRoperations, command team test and training, simulation,sequence development, flight command and control software.
• Image Data Development - Development of capabilities forimage processing lab, image data software, imaging scienceteam and support (pre-flight).
0 Science Data Development - Development of capabilities forscience teams and team support, science data processingand analysis (pre-flight).
8
The fully reduced and allocated cost data are presented in AppendixA for the seven major programs used in the present model developmenteffort. Although technically speaking, Viking was a single program, theLander capsule and Orbiter are treated separately since each system wasdeveloped under a separate contract. Conversely, all Pioneer Venusspacecraft were procured within the same system contract and thereforethe functional support costs are aggregate for the entire program. Noattempt was made to prorate these costs to the various spacecraft types.
Previous versions of the cost model were predicated on definingthe separation of non-recurring and recurring costs as the point in timein the project schedule when the fully-assembled proof-test-model wasdelivered to the spacecraft test facility for initial system testing.This definition, though arbitrary, was felt to provide an adequateaverage basis for model development.
Recently completed development projects, however, appear toinvalidate the use of this definition. Specifically, several of themajor flight components of the Viking Lander were almost totallyredesigned after initial system tests were started. Conversely, almostall of the Voyager flight hardware was fully fabricated well beforeassembly of the PTM spacecraft. Finally, the Pioneer Venus project didnot fabricate PTM spacecraft. This latest case is also indicative ofcurrent and future project planning, i.e. to not fabricate, assemble andtest a proof-test-model spacecraft.
Since a new definition of the non-recurring/recurring costseparation at the system level could not be found which would adequatelyapply to the projects in the database, it became necessary to analyzethe data at the subsystem/major component level. As a result of thisassessment, it was decided to separate recurring from non-recurringcosts at the start of fabrication of flight qualified hardware. Thisnew definition was applied as closely as possible to the major componentlevel. Occasionally, there was not sufficient information to determine
this breakpoint in cost. For such cases, either a single milestone inthe schedule was applied to all subsystems or considerable direction was
taken from the Work Breakdown Structure (WBS). For example, using WBSsubaccount definitions, non-recurring could be equated with "engineering",
and recurring could be equated with "manufacturing".
Table 2 presents percentage ratios of recurring labor to non-
curring labor, normalized to one flight unit, for each of the hardwarecost categories for the Mariner '69 (M69) through Pioneer Venus (PV)development projects. Cursory examination of this data, as exhibited bythe large standard deviations, leads immediately to the conclusion thatuse of simple ratios for determining recurring cost from non-recurringcost, as had been used in previous model versions, would no longer bevalid. A more complex functional form would be required.
2.2 Development Project Technical Data
Table 3 presents the project-related technical data used in formu-lating the cost model. Except for the number of flight qualified units(N) for each project and imaging resolution (PPL), all other data aresubsystem masses. No other information, such as power requirements, wasfound to be necessary for developing the cost model algorithms.
Note that for those projects that use radioisotope thermoelectricgenerators (RTGs) as the main power source, the mass of the RTG units isnot included. Also, for the Pioneer Venus probes, the masses of thesmall omni antennas are included in the telecommunication subsystemrather than considered separately in the Antenna category (ANT).
Special considerations were required for certain aspects of thePioneer Venus program. For example, many subsystem masses of what isidentified as the Bus/Orbiter are composites of averages of commonhardware components plus components of each vehicle which are unique.This approach was required because of lack of resolution in the detailedcost data between bus hardware and orbiter hardware. The PV probe andorbiter science are treated together as a separate subproject because ofinsufficient resolution in several of the instrument contracts to allowadequate proration of costs to the appropriate PV mission.
10
CM
eur—tO
cc.oPQ_j
S
§i. »LU
1
O" y^ ~/r^
==
01
£jLU01
OOopn<=c
h-LUC_J)
1 1 1Q_
gg
Z eteC >
00 Q
LU0•=£o:1 1 1
CO>
oCO
Q-
Q.COQ_
Q_i
D.
ONX
CJ
5>
coD.
,_,
^
cnIO•sr
LU >-or CE<C 03 IDQ LUo: i—
CM CM
J- I-HCM CM
O CM
CO OOCO CM
, ,
00 LO
CM fCO CO
IO CO• •
t-H i-H
i— H cn( .
i — . enr*~ CM
CM enO COCM CM
co coCM O
r-~ en•=t in«=f CM
cn t-Ho LOt-H CM
LO tor co
Q CL.
CO 1—
r . coco *a-CO r-H
cn T-HIO COOO CM
1 1
LO CM
LO <d-LO CVJ
1 1
1 1
r r-O <£>CM CO
LO LO
r-H CO
r ^ cnto o«HT co
LO cnCM LOi— i
CO t-H
en co
Q. Oa: <=cQ- «C
COCM
CO
,
LO
CM*
CO•
CM
t— 1•
OO
CO
COCM
00
10
oo
CO
oI— 1
enCO
2:oH-
» « » „ « -co r-.. o oo co coCM T-I CM i— 1 i— 1
o Is*. CM en en CMCO VO CM .—I CO COCO CM CO CM t— 1 CM
1 1 1 1 1CM
I ' l lVO CO I--CM t-H r-H
10 ^ f o1 • 1 • • I
cn I-H r- -t— 1 i— 1 t-H
LO cn «3-i i i
CM <3" OOCM CM CO
t-H 1- 1 ,• • 1 • 1 1
LO t— i r-.CO <* LO
CO t— I *J- O O
(S3 i-H 00 to <— 1
LO -=d- en**Q 00 rot-H CO *d-
LO LD C\J
•=3- en *-"t-H t— 1
O LO
r-i ' COt-H
Z Q 3 3 0 Q
» „ « « »IO t-H (£> CO U3
i— I CM CM
CM LO cn co toco *d~ I — » h*- oCM CM ^J- CO CM
r-H cn cn CMi . . .
t — co r-** LOCM OO OO CM
1 1 1 1 1
1 1 1 1 1
-
1 1 1 t 1
to to1 • 1 " I
to cnCM CM
CO ' t-H1 1 1
LO tO
CO CO «*
to o toCM OO CM
to vo1 • 1 • 1
- ooCM CM
CO O1 • 1 • 1
cr» tnt-H
g~ >" [j tS) I/")•H HH O_ CV r j
11
CO
HIr—J3(O
1 —
<CQ
1LU £>-si — QJ "— 3 CO -r~
of £
1 — <g
Ll-J>5.
OL_0
LjJ(=1
00
Q-
OCO
Q.
Q-ooQ_
Q.—1
D_
OiCD
O
o_J
oo•~DCL.
T— 1
s:
^Os:
i-H 1 1 1
j- co oCM i— 1 LO CT.
I-H
co CM r--. i
CM CO
I-H CM r^ iCO .-Hr— I
CO CO LO
CM r**. t" CTt
CM •— 1
CM O CTl 1^co co r~-CM i-H
CO O CO CTtI-H o r-vCM i-H
LO O O
CM CM O •— 1LD CM « — 1
o o oCM IO CM (--.
LO LO CO<— 1
•* CM ' sf
co r-» vo CTiCM d- i-HI-H
Q Q. Q-•z. i— o a:
00 1— Q.
1 1
CO CMr-H l— 1
CO
1 CM
2
CO CM
i-H COLO LO
CTl O
•* oLO LO
co ~O LOCM CM
r coLO O
I-H
0 O
CO* *£>CO CO
LO CTl
co co
c_> ~
$ "
1 1
CO O
CO i-HCM
1 CO
CM
1 LO
i-H 0
LO OLO
CO 0
<£> CM
CO 00
co r-.
LO *d-LO CTl
i-H O
•=* CMCO
r--.1 LO
co
•Z. Qet 0
1^ O
CTi LO
CO
i r-- i i i iCO
CTl O
1 LO •* 1 1 1CM
0 0
1 LO CO 1 1 1i-i O
i-H
en t— i -~^CO 1 1 1 1 CO r— 1CM CO CO
• CM1 CO 1 1 1 LO CO
co -=i- I-HI-H t— 1
CM LO >* CO
LO 1 ^3- LO CM 1co I-H co i-i
CM
r-.. cor--. i i i -=J- iI— 1
O O-— -. CM1 LO 1 1 1 VO CO
r-x CM co
CO LT>^->. LO
1 CM 1 l 1 CM •=!•LO CM CTl
fy* {f) " fy* 1 ^ J••c - f~ f~ r* F* O-,
•— i r-- r--.CM LO CMCM CM CM
1 1 1
1 1 1
1 1 1
CO O
O CTl 1«3- CO
CM
CM
1 1 COLO
co r*CO CM 1CM
o1 CM 1
1 LO 1CO
S 2 a
12
3. Model Development
Results of advanced (pre-Phase A) mission planning and analysisstudies typically involve detailed recommendations regarding sciencerationale, trajectory analysis and mission sequencing. Mission-relatedhardware definitions, however, are usually not as well defined. Grosspayload requirements, in such terms as net mass delivered at the targetor net injected mass at Earth, can be fairly well predicted. However,detailed mass definitions at the subsystem and component level are noteasily obtained. Hence, detailed mass estimates are typicallygenerated prior to Phase B studies by a combination of selectingappropriate subsystems/components from previous successful designsand/or current designs, and use of mass scaling relationships.
This approach can generate reasonable early mass estimates andalso yields information concerning design heritage. However, such anapproach cannot take into account what impact technological advancesand changes in general design philosophy might have on component masses.Furthermore, little, if any, information is generated regarding suchdetails as part counts, number of spare components, reliability, powerprofiles, command structure, communications link parameters, etc. Thatis, spacecraft designs resultant from advanced studies contain none ofthe detailed engineering parameters, required in performing a so-called
bottom-up cost estimate.
The approach for achieving early "top-down" cost estimates, there-fore, is to develop a cost model commensurate with the level of missiondefinition and related flight hardware details generally obtained fromadvanced mission planning studies. On the other hand, as seen inSection 2, the useful database of historical programs will provide onlya relatively small sample size for statistical analysis. Therefore,the individual algorithms which constitute the model should be parsimon-
ious, requiring the smallest possible number of estimated parametersfor adequate representation. Parsimony will thus lead to uncomplicated
functional forms while preserving as many statistical degrees offreedom as practical during the model fitting process. The techniques
13
employed for model fitting are ordinary least squares regression andregression through the origin.
3.1 Labor/Cost Proxy Analysis
All previous versions of the SAI Cost Model have used direct labor
hours as the primary dependent variable for both estimation andprediction, and the present version is no exception. The majorarguments underlying the use of labor hours include decoupling of fore-casts from inflation and ease in costing low volume production.Decoupling from inflation places all forecasts on a normalized,comparable basis, comparable both to past programs in the database andto other forecasts. Mass production techniques have not been utilizedfor deep-space missions and thus the marginal cost of mission hardwareis not substantially decreased through additional production. Themission development effort is labor intensive, and therefore, the costof mission hardware is a direct linear function of the manpowerinvolved in design, manufacturing and testing. This implies that.itmay not be necessary to analyze how development cost breaks down amonglabor, overhead, materials, other direct charges, etc. The relation-ship between mission parameters and resources can be better understoodand evaluated for accuracy when that resource, i.e. manpower, ismodeled explicitly. Finally, forecasting manpower requirements inaddition to project cost provides management with additional informa-tion to aid in the program planning process.
It is not sufficient, however to categorically state thatmanpower, expressed in direct labor hours, should be a reasonable andpractical proxy for cost. Figure 1 compares database averages ofcategory labor hours with category total cost, as percentages of totaldevelopment project hours and cost, for each cost category. On inspec-tion and for most categories, the correlation between labor hours andcost appears to be adequate.
This favorable comparison can be firmly established with astatistical analysis. In a statistical sense, the percentage ratios
14
en•— «ccen0_
UJoGC
= CJs cc^ LU
0.
<noCJ\ccCDCOOL
tnoc
(CO00cr
D(_)
a:aLJ
UJ _1CC CE
cr cct— t-D 0
ocrcr
ccQ-
at—en
incv
ot
G
O«
ino
•-
oo
•
LO
O»
O
lN3Wd013A3Q ItflOl
15
iN33U3d
shown in Figure 1 are simply sample means. Therefore, a t-testr ' for
equivalent means can be applied to the data. The assumptions and stepsfor doing so are as follows. Since the sample size is small, normality
must be assumed in the data because tests for normality would not be
powerful enough. (This assumption actually holds throughout the
modeling process for the same reason.) A fairly powerful test isrequired in order to decrease the probability of incorrectly accepting
the hypothesis of equivalent means. Since there is no control oversample size, the best available option is to test at a fairly highsignificance level. For this purpose, 20% was selected.
The t-test is based on equivalent sample variances. Therefore,(2}an F-testv ' for equivalent variances was first performed. Results
from the F-test indicate that for all 21 categories shown in Figure 1,
the variances of the labor hour ratios are equivalent to those of the
cost ratios at the 20% significance level, and therefore the t-test
can be applied.
(1) The t-test evaluates the statistict* = d /TT/s(d)
against the Student's t distribution with n-1 degrees of freedom, to test the hypothesis that twonormally distributed populations with the same unknown variance have the same mean. The test is
performed using n paired observations from the two samples. In the above equation
d = the mean of the differences between the two samplesn = the sample size
s(d) = the standard deviation of the differences between the pairedvalues
For a selected significance level, a, if
|t*| < t(l-a/2; n-1)then we may conclude that the two sample means are equivalent.
(2) F-test, as used in this context, evaluates the statistic
F* = sf/Sa
against the F distribution with n-1 and n-1 degrees of freedom, to test the hypothesis that two
normally distributed populations have the same variance. The test is performed using n pairedobservations from the two samples. In the above equation
Sj = is the standard deviation of the first sample
S2 = is the standard deviation of the second sample
For a selected significance level, a, ifF(a/2;n-l;n-l) < F* < F(l-o/2;n-l;n-l)
then we may conclude that the two samples have the same variance.
16
Results of the t-test show that in 16 of the 21 categories themean ratios of labor hours and total cost are equivalent at 20%
significance. This result is fairly acceptable since at 20%,approximately 4 out of 21 categories are expected to be not equivalent
purely by chance. Relaxing the level of significance to 10% results
in 19 of 21 categories having equivalent means where 2 of 21 are notexpected by chance.
The implications of these results are that direct labor hours
should provide a good proxy to total cost as the primary dependent
variable and that factors derived as simple ratios should be adeqiiate
for conveTt'inq from labor hours to total cost. This latter conclusioneliminates any need to analyze the cost data in terms of breakdowns
among labor, overhead, material, etc.
3.2 Functional Forms
Having argued that 1) the type and amount of independent variables
for cost prediction of advanced planetary missions is limited, 2) that
the cost category algorithms should be parsimonious, and 3) that cost
category direct labor hours should be the primary dependent variable,
the following general relationships are postulated. For the flighthardware categories (subsystems), total direct labor hours (DLH) is a
function of the number of flight qualified units (N) and the subsystem
mass (M):DLH = F (N,M) ,
with separate functions independently derived for each category. Forthe functional support categories, total labor hours is postulated to
be a function of the total hardware labor hours (all categories):
DLH = G (zDLH hardware).Table 4 presents examples of some of the functional forms that
were analyzed for possible relevance to the cost model. The single
parameter type functions attempt to capture the separation between
non-recurring and recurring costs by modeling each quantity separately.
The dual parameter functions essentially ignore this separation and
17
Table 4
EXAMPLES OF POSSIBLE FUNCTIONAL FORMS
Single Parameter Functions
1) DLH = NRL + N*URL
where NRL = f(M)
URL = k*NRL , k constant
2) same as 1) except URL = g(M)
3) DLH = (NRL + URL) + (N-l) * URL
where (NRL + URL) = f(M)
functions f(M) and g(M) are of the general
forms a + bM, aM or exp (a + bM)
Dual Parameter Functions
4) DLH = a + bN + cM
5) DLH = aNb Mc
6) DLH = a + b(NM)
7} DLH = a(NM)b
8) DLH = exp [a + b(NM)]
9) DLH = aMb + cNMd
10) DLH = (a + bN) Mc
18
attempt to directly model category total cost. Function types 1 through
8 in Table 4 can be fitted using the standard techniques of ordinary
least squares regression. In some cases regression through the origin
was used. This technique constrains, on an a priori basis, the estimate
of "a" to be zero in function 6 and to be one in function 7 for example.
Function types 9 and 10 can only be fitted by using non-linear
regression and were only briefly examined.
3.3 Test Statistics
In order to identify which of the possible function forms best fit
the database, some statistical measures of goodness-of-fit are required.For this purpose, two standard measures were used during the regression
analysis: the correlation coefficient of the fitted data and the
t-statistic of the parameter estimates. (The t-statistic in thiscontext is defined as the parameter estimate divided by the standard
error of the estimate. It should not be confused with the t-test
described in Section 3.1, although it is evaluated in a similar manner.)
These measures are fairly adequate for determining a best fit ofthe data at the individual category level, but are not sufficient for
analyzing model performance at the project level. Therefore, twoadditional measures were defined for use at both the category and
project levels: the mean percentage error (MPE) and mean absolute
percentage error (MAPE). Given a theoretical function of the form
y-j = a + b xj , i = 1 , ... , n
the fitted function is then
y-j = a + b x-j ,
and the residual errors due to the fitting process are
The two measures are then simply defined as
MPE = -J- i -^-
and
MAPE = -4- Z -TT"- •
19
Mean percentage error can be viewed as an indication of bias in the
fitted function. Mean absolute percentage error can be viewed as anindication of the gross error inherent to the fitted function and thus
the confidence with which it can be used in making forecasts.
Specific cutoff values of these statistical measures were not used
during the modeling process. For the most part, the "best fit"functions were chosen on the combined basis of having the highest
correlation and t-statistics and lowest MPE and MAPE. However, certain
qualitative, or subjective measures were also applied, such as a desire
to avoid negative constants in linear (straight-line) fits.
20
4. Model Analysis
The first model formulation analyzed was the single parameter type,
which separately models non-recurring and recurring labor hours, for the
obvious reason that this form had been successfully modeled in previousversions. Briefly, however, it was now found that algorithms based on
non-recurring labor as a function of subsystem mass had poor correla-
tion and unacceptably large percentage errors. Furthermore, as might
be expected from the data in Table 2, recurring labor could not be
expressed as a constant fraction of non-recurring labor. Algorithms of
recurring labor as a function of subsystem mass did have moderate
correlation and marginally acceptable percentage errors.
Reexamination of the cost data allocations did not reveal any
obvious discrepancies to which the poor correlation could be attributed.
Therefore, it seemed apparent that a new model formulation would have to
be investigated. Consequently, a wide variety of theoretical functions
based on the dual parameter formulation were analyzed. Multiple lineartype functions, i.e. having two independent variables, exhibited
moderately good correlation, but, in general, the strongest correlation
was observed in algorithms that modeled category total labor as afunction of a single variable formed by the product of subsystem mass
and the number of flight units. Simply restated, total cost is a strong
function of total mass. This is merely a statement of observed
correlation in the data and is not necessarily evidence of a fundamentalrelationship.
Figure 2 illustrates the preceding discussion with the structure and
devices hardware category as an example. Each graph shows the sample
data and the fitted function, in each case a power law. Considerable
scatter and thus poor correlation can be seen in Figure 2a, non-recurring
labor versus unit mass. By comparison, a significant trend is observed
in Figure 2b, total labor versus total mass. Similarly strong correla-
tions between total labor and total mass were/observed for most"all other
hardware cost categories.
21
Ni NK X"
V
N g|
\L^
K
NN\
XX 0»
)KV,N K
^, sxK £
\j^^ 1
^= « Nsi N1 a N b .w *- C X? ££ *~ t £
E £ . o *" •$£vL A N^ ^j \
XX
X
z:x"Z.~~to(ACCz:_ju:*—o
•O01-M<r-M-
^_O
-QrO ro
ro^> ~Oi-o -aO) ro
4->ra E0 0
r— -Mro (J
O 31— 4-
>>
Oa>O)
ro
4->co0oCO0)o
,f—>OJ
o3
COOJS-3
•t~>O
sanoH uoabi loaeia 10101toa>o
ai
c/oS-o
co
O
Os_•)->C/l
n3
O
•t->O
* \ f
L \ s^K\
\\\
si .0 ' x
\\\
r- *
L \ g* J ;
01 \ y »
^ ^\
\
£ i X
g ° „ \5 § r \K t M/ N
E £ ^ v Sih. 0 £
, , \ M;
5^ O \X\
>*.
"OO)
+>*-» "r~E "^
s-to o
i 3St- """ T3— CJ)Z C "O= -r- C
S- ro
O 0O> -r-S- -!->1 OC EO 3Z 4-
_3LL.
-t->ECDi-OJ
t
0
M-O
COE0CO
CroQ.
OCJ
OJ
(DS-
0)
synoH yoaoi IDBUIQ
22
All of the available information in the database was not used indeveloping the cost model algorithms. The earliest projects,specifically Mariner '64, Surveyor and Lunar Orbiter, were largelyexcluded from the modeling effort because of the early technologystatus inherent in these projects. Exceptions to exclusion were madewhen a specific cost, category data point correlated strongly with thedata from later projects. Conversely, if a data point from the laterprojects was observed to be an obvious outlier from a significanttrend, that data was excluded from the regression. An example of thisis seen in Figure 2b in which Mariner '64 is included among the fitteddata but Pioneer Jupiter/Saturn is excluded. This selective use ofthe data is not arbitrary. Data included from earlier programs mayindicate that the effort required to design, fabricate and test theparticular subsystem is somewhat independent of the technology in use.Data excluded from recent programs indicates either extreme cost over-runs for high outliers or extremely high heritage for low outliers.
The decision to eliminate outliers led to the complete exclusionof Mariner '73 from the modeling process. This is not surprisingsince this particular project used a considerable amount of residualhardware from previous Mariner projects. Specifically, all hardwarecategory costs for Mariner '73 were observed to lie well below trendsindicated by the other projects in the database.
Even though non-recurring cost could not be successfully modeled,this quantity in terms of labor hours was still needed in order toapply inheritance algorithms (see Appendix C). These algorithms weredeveloped on the premise that design and hardware heritage affectsonly the non-recurring portion of development costs. Developing newinheritance algorithms which would operate on total category cost wasbeyond the scope of the current model development effort. Thus, aprocedure was needed to extract the non-recurring portion from thetotal labor estimate. As was previously mentioned, the recurring
cost data exhibited moderately good correlation with subsystem mass.
Obviously, then, an estimate of non-recurring labor could be recovered
23
by differencing the total labor and recurring labor estimates. Duringthe process of analyzing recurring costs, a slight improvement incorrelation was observed between total recurring cost and total massover that of single-unit recurring cost and unit mass. Algorithms oftotal recurring labor as a function of total subsystem mass were there-fore developed. An estimate of single-unit recurring cost can beobtained by simply dividing by the number of flight units.
In the functional support categories, costs for system supportand ground equipment and for launch + 30 days operations and groundsoftware were found to correlate well with the sum of flight hardwarecategory costs. Costs for pre-launch development of capabilities forimaging data processing and for non-imaging science data analysis wereobserved to correlate with imaging system resolution and non-imagingscience payload mass, respectively. Finally, the cost of programmanagement correlated well with the sum of all other category costs,both hardware and support functions.
Table 5 summarizes the complete set of algorithms developedduring the modeling process. The flight project algorithms from theprevious model version are also presented. Graphs of the data andfitted functions together with associated statistics are presented inAppendix B. Figure 3 is a flow diagram illustrating the key elementsof the cost model.
MissionInput
Parameters
vpCost
CategoryLabor
Algorithms
CostCategoryLabor
Estimates
N
/ Labor Hours \I to 1-\ Labor Cost /
ReducedLabor
Estimates
-4f Programv> 1 «!,„_
VCost
Figure 3. Cost Model Schematic
24
Table 5
Summary of Cost Model Algorithms
Development Project - Flight Hardware
Structure & DevicesDLH = 1.626 (N * M)0.90«t6
RLH = 1.399 (N * M)°'7l+l+5
Thermal Control, Cabling & Pyrotechnics
DLH = exp (4.2702 + 0.00608 N * M)RLH = 3.731 (N * M)0.6082
PropulsionDLH = 56.1878 (N * MjO-^eRLH = 1.0 (N * M)0.90ii
Attitude & Articulation ControlDLH = 21.328 (N * M)°-7230
RLH = 1.932 (N * M)
Telecommunications
DLH = 4.471 (N * M)1-1305
RLH = 1.626 (N * M)1-18^
AntennasDLH = 6.093 (N * M)1-131*8
RLH = 3.339 (N * M)
Command & Data Handling .
DLH = exp (4.2605 + 0.02414 N * M)
RLH = exp (2.8679 + 0.02726 N * M)
RTG PowerDLH = 65.300 (N * M)0.355<*
RLH = 7.88 (N * M)0-7lso
Solar/Battery PowerDLH = exp (3.9633 + 0.00911 N * M)
RLH = exp (2.5183 + 0.01204 N * M)
Aerodeceleration ModuleDLH = 3.481 (N * M)0-8ti6
RLH = 4.662 (N * M)0-5
M in kilogramsDLH, RLH in 1000 hours 25MD, ED in months
Table 5 (continued)
Summary of Cost Model Algorithms
Landing Radar/AltimeterDLH = 11.409 (N * M ) ° - 9 5 7 9
RLH = 1.2227 (N * M)1-23"
Line-Scan Imaging
DLH = 10.069 (N * M ) i . 2 5 ? o
RLH = 1.989 (N * M) 1 ' 1 * 0 8 9
Vidicon Imaging
DLH = 4.463 (N * M)L0369
RLH = 1.0 (N * M ) i - i 5 2 0
Particle & Field InstrumentsDLH = 25.948 (N * M)0.7215
RLH = 0.790 (N * M)1'39™
Remote Sensing InstrumentsDLH = 25.948 (N * M)0.5990
RLH = 0.790 (N * M)°-8393
Direct Sensing/Sampling Instruments
DLH = 6.173 (N * M)1-2737
RLH = 1.0 (N * M)i.^oo
Development Project - Support Functions
System Support & Ground Equipment
DLH-=. 0.36172 (z DLH hardware)0-9815
DLH = 0.5095 (z DLH hardware) [viking Class Missions]
Launch + 30 Days Operations & Ground Software
DLH = 0.09808 (z DLH hardware)
Imaging Data DevelopmentDLH = 0.00124 (PPL)1-629
Science Data DevelopmentDLH = 27.836 (non-imaging science mass)0-3389
Program Management/MA&EDLH = 0.10097 (zDLHall categories^. 9670
Flight Pro.ject
Flight Operations / vn, u / zDLH Hardware \ °-&,,DLH = I 3100 ) (10-7 MD + 27.0 ED)
Data AnalysisDLH = 0.425 (DLH Flight Operations)
26
4.1 Labor/Cost Conversion Factors
As was demonstrated in Section 3.1, simple conversion factors,
derived as average ratios, are all that are needed to convert category
direct labor hours into category total cost. Two independent factors
were derived for this purpose: the first converts labor hours tolabor cost while the second converts labor cost to total cost. These
conversion factors are presented in Table 6 for all cost categories.
Derivation of the labor cost to total cost factors was straight-
forward, involving simple ratios of the allocated cost data. However,derivation of the.labor hours to labor cost factors first required
elimination of the effects of inflation inherent in the raw cost data.i-
For this purpose, the NASA R & D Escalation Indices for Space Systems
Development (February 1979) were used to adjust the annual funding for
each category in each project to a fiscal year 1977 basis. Once this
was accomplished, simple ratios were again used to derive the
conversion factors. Projects completed prior to 1970 were notincluded in this process because such projects, having median funding
nearly a decade or more from the basis, would introduce too much
variance in the ratios.
4.2 Error Analysis
The first step in analyzing model performance was to test the
model against the development projects in the database. In order to
obtain global, consistent measures of model performance, the error
analysis was confined to the seven major projects contributing to the
model development. Errors from cost categories excluded from the
regression analysis (such as PJS structure and devices) are includedin the global error analysis.
Table 7 summarizes the percentage residual errors at the hardware
level (i.e. exclusive of the support categories) of the seven projects.
Also shown for comparison are the hardware level residual errors
obtained from a model based on separate algorithms for non-recurring
27
Table 6
Labor/Cost Conversion Factors
Cost Category Labor Hours to Labor Cost Labor Cost to Total Cost
(FY77 dollars/manhour)Development Project
Structure & Devices 10.45 3.303
Thermal Control, Cabling & Pyrotechnics 10.26 3.317
Propulsion 10.54 3.616
Attitude & Articulation Control 10.63 3.347
Telecommunications 9.99 3.352
Antennas 9.96 3.466
Command & Data Handling 9.68 3.163
RTG Power 9.51 3.177
Solar Battery Power 10.41 3.148
Aerodecleration Module 10.73 3.296
Landing Radar/Altimeter 10.08 3.158
Line-Scan Imaging 10.57 3.604
Vidicon Imaging 9.52 3.586
Particle & Field Instruments 10.62 3.395
Remote Sensing Instruments 10.65 3.286
Direct Sensing/Sampling Instruments 9.55 3.454
System Support & Ground Equipment 10.55 3.076
Launch + 30 Days Ops & Ground S/W 10.71 3.214
Image Data Development 11.46 3.130
Science Data Development 12.76 3.987
Program Management/MA & E 11.57 2.685
Flight Project
Flight Operations 10.44 3.247
Data Analysis 10.44 3.42528
Table 7
Comparison of Hardware - Level Residual Percentage Errors
Hardware Project
M69
M71
PJS
VLC
VKO
V6R
PV
MPE
MAPE
Total DLHModel
1.8-3.1
17.1
-1.2
-5.4
3.0
-8.6
0.6
5.7
Separate NRL/URLModel
26.4
-4.1
26.2
19.8
9.3
8.4
-19.4
9.5
16.2
and recurring labor. The global errors, MPE and MAPE, support expecta-
tions based on the individual algorithm correlations: a model
formulated on estimation of total category DLH would exhibit less bias
and higher confidence than a model based on separate NRL and URL
algorithms.
The next step in the error analysis was to determine the global
errors at the development project level in terms of both percentage
labor hours and total cost. Estimates of support function labor, for
the most part, are based on the summation of hardware category labor.
Thus errors in hardware labor estimates would be expected to propagate'
through the support functions. Also, the factors to convert from
labor hours to total cost are simple averages and the variances
implicit in such averages would also be expected to propagate through
the conversion to cost. These considerations imply that the global
errors at the project cost level might be worse than those observed
at the hardware labor hours level.
Table 8 presents a'summary of the global errors at the project
level. The first grouping, termed "Fitted", are those errors which
are obtained when the actual hardware level labor is used to estimate
29
ft3
3CO
s-o
0) S-
CO
tO
O)oi.Olo.
•CD
eaE
toLU
-oQJ
[ *•r-U_
r— O(O QJ
0 01— S-
D.
1 _> J~
i- OO T-
CL U^3 C
CO 13Li
QJ
03
t3
rO3:
-Mi— (J<O O)
0 01— S_
Q.
i ^ c~*
S- Oo •<-CX-MQ. 03 C
CO =3U.
QJ
res
-o
or
•-i O• •
c\j r-»i
r-~. co
oo «d-i i— i
vo r-.
O Lf>
LO «d-• •
T— 1 (^3
1
co tn• •
<& CxJ1 i— I
ID r-.O Lf>
UJ ^
D; ooo c;CO =)=£ 0
«-< CM• •
r-i OO1
CO LO
t*O (1 I— 1
LT> cr>r— 1 IO
CO CO• •
co r**>i
LO CO
Lf> CM1 i— 1
Lf) CT>
1—4 VO
S g
—1 .
g §f ^
30
the function labor. The group labeled "Estimated" is the errors
observed when the estimated hardware labor is used. The conclusion
from these results is that use of estimated hardware labor in the
function algorithms and averaged conversion factors has only marginal
effect on the hardware level global errors. The model as formulated
captures the database with a marginally conservative bias of approxi-mately 1% overestimation and a gross error less than 10%.
Appendix D contains the details of the error analysis from which
the results in Table 8 were obtained.
4.3 Simulation Analysis
The error analysis discussed in Section 4.2 provides an assessment
of how well the model has captured the data from which it was developed.
However, it provides little information regarding how well the model
will perform in practical applications.
Because of the large number of input variables, a parametricanalysis of the model would be cumbersome to perform and difficult toevaluate. Monte Carlo simulation can provide a feasible method for
assessing the overall performance of the model. By artificiallyviewing subsystem masses as stochastic variables, the simulation
reduces to selecting probability distributions which adequately express
these variables, and randomly sampling from these distributions. A
simplifying assumption is made that individual subsystem masses areindependent of other subsystem masses. The sampled masses from each
trial of the simulation can be input to the model in a deterministic
fashion and the model output can then be averaged over the number of
trials. Given a fixed number of trial runs, the number of flight
units becomes the only control parameter.
Selecting probability distributions is the only matter of specula-
tion. Both uniform and triangular probability density functions were
used to assess what affect, if any, such different assumptions mighthave on the outcome. Figure 4 below represents a typical triangularprobability density function.
31
Pr(M)
'mm max M
Figure 4. Subsystem Mass Probability Density Function
The lower and upper bounds, Mm-jn and Mmax, for each hardware cost
category were chosen as the smallest and largest subsystem masses
observed in the historical database from Table 3. Three different
modes were analyzed for the triangular pdf's: 25%, 50% and 75% of the
difference between Mml-n and Mmax. The number of flight units was
systematically varied from one to five and 1000 trials were run for
each case. Only the hardware total labor was analyzed since the
support categories are direct functions of this quantity and their
inclusion would not provide additional information.
For purpose of comparison, a simulation of the previous version
of the SAI Cost Model was also made. Table 9 summarizes the results
of the two model simulations with number of flight units as the
parameter. That the total simulated mass is the same for each
respective pdf attests to the fact that to maintain consistency the
same pseudo-random number stream was used in each case.
The first point to notice in the data is that the results from
using different probability distributions are consistent. Therefore,
any conclusions which may be drawn from the analysis are not
dependent upon the assumptions used to randomly select subsystem
masses.
The major result indicated by the data in Table 9 is the non-
linear increase in total labor as a function of the number of flight
32
CO•r—CO
ro
Co
ro
3E
*r—oo
cn o>-o
<v o
ro
o
L>-
o
3COo>
rO
LOII
COi-3
-E
Oo <s-O IIi— 1 Z.
S-o
_!
i — COrO II
O
O)
rO5
S- CMrd IIIE 2:
CUCT>
ef
CO
"OCU CDCO+-> COrO fO COS- i — roOJ 3 S>_E
CO ro
Or—
LL.QQ.
CO
•r—r— -»->O> ro
o 's2: E
0u.
CO
COCOCM
LOooCM1— 1
CMcnCOCO
COCO
COCOc— 1CM
B J-
r-
0
.,_c
:=>
cn t— i CMi— i co r^
*d- co coi— 1 t— 4 CM
i s §§ £ a
CO O COt — «3- 1— 1»— i cn cnLO LO CO
•—l CM LOO CT) i-i<d~ r - CMCO CO *d"
CO O « — 'CO LO COcn i — i coi— I CM CM
CM LO CO ,f"» r . i*LO LO CO
LO O LOCM LO rS- S- S-rO rO ro
3 3 3CO CO COC C Cro rO ro
£ £ sit— 1— H-
So>•z.
cn •— i coLO 'd- COo r^ •— iLO "=3- LO
CM O LOCO CO CMCO CO t~^
•=}• O CMLO CO f-HCM cn co" CO ^J"
co cn co^d~ cn cnco LO coCO CO CO
co cn LOCO i— I CO«d- CM <3-CO CO CO
•vl- CM LOr*""" cnzj r "*""LO LO LO
LO OCM LO
E S- S-S- ro roO r— r—
•r- CO COC C C
Z3 <O ro"r— *^S- S-
K- 1—
CO
0•r—>
CDS-
Q_
COi — iLOLO
cnCOoLO
LOCMCO
"*
oCOI— 1
COr .CO
CO•=3-CO
LOr^-i.to
1 —— JCOcro
•t—
1—
OroCU
S-o
01
ro
si
OOO
it-CD
<DCOroS_o;
33
units. Figure 5 is a plot of the data for the 50% triangular pdf. The
increase in labor for the previous model is linear, which is consistent
with intuitive expectations, ignoring such factors as economy of scale.
Since the hardware cost category algorithms of the new model are a mix
of power laws and exponentials, the increase in labor as a function of
flight units was not expected to be exactly linear. However, the
highly positive slope of the function in Figure 5 was not expected.
To investigate this further, simulations were also made of models
based on formulations 2 and 3 in Table 4. Results of these simulationsare shown as the hashed region in Figure 5. These results exhibit
consistency with the results from the previous model simulation:
increase in total labor is a linear function of the number of flightunits.
4.4 Error Analysis Revisited
From the previous discussion, it appears that the model formula-
tion, as presented in Table 5, is highly sensitive to the number of
flight units. A procedure is needed to reduce this sensitivity while
maintaining the high correlation of the category algorithms with thedatabase. Two considerations indicate that such a procedure may be
feasible. Of the ten hardware systems in Table 3, five consisted oftwo flight qualified units. Furthermore, at N=2, results of the new
model simulation are very near to the other model simulation results.
This suggests a procedure which may be termed anchoring and adjustment,
that is, to anchor an initial cost estimate at two flight units and
adjust this estimate up or down by the appropriate amount of recurring
cost. Expressed mathematically, the procedure is
DLHa = DLH(2,M) + (N-2) * RLH(2,M)/2
where DLH(2,M) and RLH(2,M) are the algorithms of Table 5 evaluated
for N=2. The correct sign for the adjustment term is automatically
accounted for in the (N-2) factor. For N=l, the recurring cost of one
unit is subtracted from the estimate and for N13 the appropriate cost
is added. For N=2, the initial estimate is unaltered.
34
Figure 5
Comparative Summary Results of Simulation Analysis
10
O
Ol
1 5-os_fO
? 4
new model
region ofseparateNRL/URL models"
previous model
triangular mass pdf50% mode
averages of 1000 trials
I2 3 4
Flight Qualified Units
35
Simulations of the new model based on this procedure were performed
and are plotted as the upper bound of the hashed region in Figure 5.
The error analysis of the model database performed in Section 4.2
was reassessed using the adjustment procedure described above.
Table 10 summarizes the results of this analysis. The estimated errors
from Table 8 are repeated for comparison with the errors arising from
the adjustment procedure. The adjusted errors show that, due to the
adjustment procedure, the model still has relatively low bias, but the
gross error has increased. Development projects in the database with
other than two flight units contribute to this increase in mean absolute
percentage error. Despite the factor of two increases, the gross errors
of both labor hours and total cost are still within the acceptance limitof 20%.
4.5 Benchmark Tests
The last step in analyzing model performance was to run benchmarktests of the model against project costs from other sources. Two
sources for such tests are available: actual costs of past projects not
included in the model database, and independent project estimates of
current and future programs.
For the first case, one project is readily available from the SAI
cost database, specifically, Mariner Venus/Mercury 1973. This particu-lar project readily lends itself to the benchmark tests since no data
from it was used during model development, yet it was accomplished
during the same time period as those projects in the model database.
Two projects were selected for the second case: the Galileo Probe to
Jupiter and the Venus Orbital Imaging Radar (VOIR) spacecraft. The
estimated total development cost for the Galileo Probe was obtained
from the POP 80-2 fiscal year cost estimates in Reference 3, adjusted
to FY'82 constant dollars. The benchmark cost of the VOIR development
project was obtained from a JPL cost review (Reference 4 ). The cost
of the synthetic aperture radar system is not included in either the
benchmark cost or the model cost estimates.
36
cu
"3
COS_Os_s_
O)oO)ex
o5-aai
CUa:
•oO)
CO*~^
• 1—5
T3<C
CU•M
E•r—+JCOLU
r— Uro <L)+J '""jO OI- S-
Q-
i- 00 -r-Q.4JQ. U3 C
OO 3LJ_
0)S-03
s_ro
r— O(O CU•4-* '"O 0
1— S-CL.
i- 0O -1-r^ ^_)Q. CJ^J E
C^) 3U.
CU
03
O3n:
CO CM• •
CM IDrH
i-H CO• •
CM CT>1 r-t
r- LOr 10
r-H
•-• 0• •
CM l~~1
r-~ oo• •
CO f1 I— I
^o r^O LO
LU m
s_ too s-rO O
CO CO• .
O Is**1 I— 1
• •
LO CM1 CM
r-H 0
CM VOi — I
i— 1 CM• •
i— I OO1
oo uo• •
LO LO1 . <— <
LO CTl
T-H to
s I<o "J^
^ 5
37
Table 11 summarizes the results of the benchmark tests. All costs
are given in fiscal year dollars appropriate to the particular benchmark
cost. Subsystem-level heritage was taken into account in generating all
model estimates. Estimates from the cost model are shown first without
using the flight unit adjustment procedure. The model estimate for
Mariner '73 is well within acceptable error limits. However, the
initial estimates for Galileo Probe and VOIR are significantly below
the benchmark costs. Applying the adjustment procedure has no affect on
the Mariner '73 estimate since two units are costed, but the estimates
for Galileo Probe and VOIR are now both within acceptable limits.
Based on the previous analyses, it appears that the revised model
performs well for applications involving either two flight units or one
flight unit with the adjustment procedure. Firm conclusions cannot be
made for applications involving three or more flight units since
appropriate benchmark projects are not available to test such cases.
Considering the results of the benchmark tests, although limitedin number and scope, together with the results of the database error
analysis, the uncertainty of cost predictions obtained from this modelcan be stated as follows. Flight hardware mass estimates associated
with future missions have uncertainties which can vary greatly depending
on the design concept and degree of heritage. Thus, for missionconcepts that generally fall within the scope of missions in the model's
database, the estimated uncertainty in cost predictions is approximately
-20%. For mission concepts which are outside the scope of the database
missions yet can be costed with this model (e.g. penetrators or rovers),
the uncertainty of cost predictions cannot be estimated and must be
treated on an individual basis.
38
QJ
JD03
V)QJ
S-03
U
QJCO
oto
toOJo:
oo
-oQJ
to OE s-
••- s_i * | ! |
01LU
"O to QJOJ O -M4-> O <0to E3 r— ••-•r-j QJ -MT3 "O 01<t _O LU
-oQJ
n3 OE s-
•r- S_
toLU
•4-J10 QJ0 •*->
<_> "3
QJ -!->T3 01O LUs:
.*
E Mc 5O)oo
Jc "*
'*~ c:
c. ^
E "*"* QJf^ /it ^O i-
^?0QJ °-o
&9ini
r .>-u.
s:Pv.COI— 1
^5uni
£>-U-
^COr— 1
r--
u.
oCO<— i
c.
QJCO 01r * n3
J3na
S- -l->QJ (OC Q
S- i— ito eCS <A
S^r— l
1
OJCO>-u_
^>cr>o.—i
^5i — iCO
00CO>-u-LOr »
^
OvJOO
U.
cio1 — 1
I— <
QJ
OS-
CL.00
0 iQJ 0
•— CO•r—i— CL.ra O
C3 Q_
S^« J-
1
0co>-U-
2.
OJOJ-fae-
s-sCOOJ
0co>—u_
s:in
^
oCO
u_
CO«— IOJ
t— 1
-p010 2O QJ
l_l 1 >O Q- QJ
39
"PAGE MISSING FROM AVAILABLE VERSION"
5. Sample Applications
Two sample applications of the cost model are presented to
illustrate its applicability to a variety of project implementation
concepts and mission scenarios.
The first example is a comet rendezvous mission using a low-thrust
trajectory with the stage concept for solar electric propulsion. The
mission is the Hal ley Flyby/Tempel 2 Rendezvous as defined inReference 5 .
The following two pages present completed input data worksheetsfor this mission. The first worksheet defines the project scenario
and provides necessary guidelines for generating the cost estimate.
The guidelines indicate that only the Mission Module spacecraft and
its associated mission operations are to be costed. The second worksheetpresents subsystem mass estimates for the Mission Module flight hardware
together with estimates of the inheritance classifications for each
subsystem.
Figure 6 presents the computer-generated output of the cost model,
showing the raw, detailed cost estimates for the Halley/Tempel 2 Mission
Module hardware development and for the mission flight operations and
data analysis. Within the Development Project subheading, the first
group of columns display individual category labor hour and cost
estimates without accounting for inheritance. The last three columns
show labor and cost estimates after factoring in the effects of
inheritance. The estimated cost of hardware design and fabrication
is approximately $134 Million, after an estimated 30% savings due to
inheritance. The cost reduncing effects of heritage in the flighthardware propagate into the functional support categories and the
flight project categories so that the total development is estimated
to cost $214 Million and the total program is estimated at $292 Million,
without contingency. Table 12 summarizes the cost estimate in a format
which is typical of actual reporting practices. "Science Development"
includes the instrument hardware and imaging and science data development
41
SAI PLANETARY PROGRAM COST MODEL
Input Data Worksheet Page 1Project Scenario Definition
MISSION*: Hatt(Ly flyby/Twpet 2 Rendezvous
HARDWARE CONFIGURATION:
SPACECRAFT ELEMENT*
Module.
NO. OF UNITS* DESIGN HERITAGE
WASA Standard,Gatlte.0, Vo yag eA
LAUNCH VEHICLE: ShuttlfL/lUS
FLIGHT MODE*: Solan Elzc&Uc PJiopul&ion
MISSION PROFILE:
LAUNCH NO.
7
BASE FISCAL YEAR11
LAUNCH DATE*
Ju£(/, 1985
FV79S2
FLIGHT TIME*
4Z MQwtfa
ENCOUNTER TIME*
25
COST SPREAD OPTION PROJECT START DATE:
SPECIAL COST GUIDELINES:
Co4^xS 0($ SEP Stage., SEP OpeAdtionA, and Hattzy P/tobe not -oic£uded.
Module powa/L v-ia Stage..
201
*Necessary Information
42
SAI PLANETARY PROGRAM COST MODEL
SPACECRAFT ELEMENT: HF/Tf
ENGINEERING:
Structure & Devices:
Thermal, Cabling & Pyro:
Propulsion Inerts:
Att & Articulation Control
Telecommunications:
Antennas:
Command & Data Handling:
Power*: Solar _v/_ RTG
Aerodeceleration:
Landing Radar/Altimeter:
SCIENCE
Imaging Mass:
Imaging Resolution:
Particle & Field:
Remote Sensing:
Direct Sensing:
Input Data Worksheet Page 2
Flight Hardware Definitions
INHERITANCE CLASS PERCENT BY MASS
199.3 kg
65.0 kg
BLOCK EXACT MINOR MAJORBUY REPEAT MOD MOD
0.0 0.0 33.0 33.0
0.0 0.0 33.0 33.0
NEWDESIGN
34.0
•34.0
-0- -kg
: 59.5
40.7
6.2
45.4
18.6
-0-
15.0
32.0
800
82.2
11.0
-0-
kg
kg
kg
kg
kg
kg
kg
kg
PPL
kg
kg
kg
33.6 49.4 6.0 0.0
20.9 50.6 28.5 0.0
0.0 0.0 100.0 0.0
0.0 100.0 0.0 0.0
0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0
0.0 33.2 0.0 0.0
Vidicon CCD S Fax
0.0 9.5 23.1 27.2
0.0 0.0 40.9 0.0
11.0
0.0
0.0
0.0
100.0
100.0
66.8
40.2
59.1
*For RTG Power Systems, do not include mass of RTG units
43
CMCOI
DiQ-
ICO
uQUl«a.
toOO
a:OCD
s:(n
CMCO01
[_ro*iiIV
f—
rouvi
u.
inCO01
D
L>
Drt-1
in»-
u.
0)
3•aos:co
«!
CM
r— .
t)G.E H-0 o1- LU^s, r~>>> 0C> ci
t— e^r—
0 H
lil• « jrE r...13 OL, _lcn iLtO >I. i.-j
c\_ c*
»-iH-
O WU
•— cn^C I—
oi y"Z.
~ V)i-LI I—
O -^
"* SI>« 0— s:
to
< —-JS
LU
^ £
1-
m ca
I-
o wu
:r ti
z A:
a: *
CxL .C
I 10,-J L.Ca .C
V
to toC/J C'l< -VIE
•***IT, >,
+J O-.- rr.C 6^ -4"*
c?•!-> U.cC-i
•F-
*—tJ-
*— I
^j-fsj^-fr^csjCJuiLntximojOl CO M CM <J* CO CO "-i CO CO
Ln^Dro^rcocriroro(M-«sj1-4 f— t t-« .— | .-H (V) _H
^CM^-LO^LO-CO^
»-H r *- co co co tn *^ r vo i.o c^JOvjrs»-'~ | i i i^ i 'COLO^-*LOCOCOCSJ — * — • W ^-* — <
^CML^CMVOCMVOCO^r^
^HCococncncn-^LOiDSjcnCO — i CM LO LO CO U3 I.O — « CO F--CM -H CM CM « CM CM LO —•
^^^^^^^ fe^CMS,
ro co CM
coroiocot3&(S}te!*SjCOfi;CO CO CM 6} CM^f
«3 6. t LO 65 S! S5 eg fM in 6s
fej fej d ^ fe; t5> Sj CO CT^ j"* UJ G< CO
^^vom^^^^e t ta
CO CM
cocor^ — covD'a-corMi^ —r ^LDCMI^- * lOCOCOLO-H_— < COCM CM ri — i ro —
oo r in rvj [ ^ -et 10 vo L.O -
£ £ 5 5 S £ £ ~ ~ 2 SCM t CO "? CM CM -H
•* 6? e; ^ r- ITS i-. Si CM co •*
ss25asw5SS!:
r» r in co -«r rv CM vo co o co
LO >-< S) r in 01 r L,-J r-.: LD -si25CMLOCTlCOCs;LOLn^-CO^lro — • in ^ in CM CM L.O -H
COS>Lnr^CM"*LDt5iSCMS>
Cl !O CTl 13; -,O l/J CC- LO CM CM '-Htn ID in tr •* >-i — co C3 — i— •
O 13 a!.tx1 7*" L* '>- HH t-" K
O- -J Ul t/3CO O Ql SI i—tjjo^ (/> 2J LL! •— ' *— *C/5O t ~ '-'-"C ^=- ! — 1~.'.*— « CI' O O ^IO I O »— «>2^Ci'-* O_c£CU ILU>— «| — J— <if *y "Z*. (jj <j;f j j ^^ -^T !— ^- r\' <— ! «— r. "?*"
rT5 O (_> ^ Ct< <•" C^ L'_ *-•«3 < O *-* CS LU Q < 00
o 2: t— <t s; °^ ^uj I— ^> t^HciH-t ujC ^ - i 2 ^ r C C / > <f LiJC/1
"» . .! -~r *~ ••x .0 f«'j c n f; „ »H < " O ": v. ^ 6 o iu(j s: «3 o 2: -^ «< i-* u •— r—r «: iu ui :>: <r o ;-• :— o&'. u ! — i !— x^ i ^••\ iO •?•* '2£.I — m ' — IM r^ '—' i~- «.-." *— ? ",-" t.u!tc i— < ',— < O cr; _j >• a! »ai
r^ ^?co r^
co mCO CM*-4
r.
U3^J-
CVJ
«— *
CiJ
fCOCOCM
CO"»SQcn•—i
-"
roCMCM
CO
sj
^_
_]
CM^3-
CO
Gl
^1
f^in
c0
^jUu
"O0d
LU•,y ^Jer nS O00c-r< -P
»Vi
— J -'.'irT t.J— (i:O r.v
44
CO CM S> CM Cvlco CM « S> cr.
cn co co C; coCO -^ »- —i
Qrt^^ r tCO f £NJ 'P CO•^ f O U) SXi h»-rv fsj ^-* CM
tr> in ^H CM jx.r*^ cri cr* co ^y
l£i CO CO Cj O^in -• ^ -•
cr. LO in t - £3
««• LD V2 PS CslLO ro LO rj coS? CO -H CO
-•
1- r- ij;
c"'1 1- LT! -3u. 3 ^ "^C
Q uj a. :d0 2-. 21 O ~sE K 2. _! !-fi 'J? O Li i ^T.CJ3 — 1 > 111
o<i UJ LU i!«C! > O LU
f" L'l O1- >• a <( <r*< 1- :a. Ci < << -rt
CO CO <C.;- ,-;, ; . > r
~ 1C O <Ul O L.J " C.<t- :•"' u> L-J eiC'7". ~> •"?" •— • O'-- ":•" """ ' » rv^W J '- °"; n-
LO r?a; 10
n co-* CMCM
6J
inr-1
COCO
01r^CTlcr.CM
^
••B-OlC-'JIf)
Co-*-
•poi
1 — ~o-z wUl Ci"f. .!>
O in-J Clii O
I >J •*)r:, •-
Ift i j
<*" *-1- pO r,.
r-< COL.T ro
OtZ.Ul
cd
o
C/1t/l
uUl
O
e_
O
-i CO
•? c.is: i—u: toO- 5-O _l
t/i -s;t-T i-
LO UDcr. c\itM
L3 —i CO^-i fj Ujtj •* CM-H in
LOCO
LO co cnCM in «»-4 r^
t5
inCM
-oO
coCOco
CM
o>Q.ECl)
S-o
3Q.
rso
01T3O
COO
O)
CD•r-Lx.
cO
ODX"eDi
: +>. r
categories. "Spacecraft" includes the engineering subsystem categories
and the system support category. With contingency, this program is
estimated to cost $355 Million.
Table 12. Cost Summary for Halley Flyby/Tempel 2 Rendezvous
FY1982 $MProgram Management/MA&E 13.9Science Development 66.9Data Analysis 25.3Spacecraft 119.9Launch + 30 Days Operations 13.2Mission Operations 56.5
Subtotal 295.8APA/Reserve «a 20%) 59.2
Total 355.0
The second sample application deals with a multi-mission projectto deliver atmospheric entry probes to the outer planets Saturn,
Uranus and Neptune. The project implementation scenario is based on
an assumption that all six spacecraft (three probes and three probe
carriers) are developed under a single hardware system contract. The
probe design is assumed to rely heavily on the current Galileo Probewith suitable modifications for use at the other giant outer planets.
The carrier design is assumed to benefit from the contractor's
experience in designing low-cost spinning spacecraft. Other guidelinesinclude a presumption that the project is charged for RTG units, 15%
contingency is to be applied, and only the development cost, i.e.
costs to launch + 30 days, is to be estimated.
Figure 7 presents the cost model output for the Outer Planet
Probes Development Project. Hardware cost categories are shown separately
for the two different spacecraft. Functional categories, however,
are shown as totals for the overall development effort and not prorated
to each spacecraft. The total development is estimated at approximately
$277 Million without contingency. Table 12 summarizes the development
45
SAI PLANETARY PROGRAM COST MODEL
Input Data Worksheet Page 1
Project Scenario Definition
MISSION*: OateA Planet ?nobn
HARDWARE CONFIGURATION:
SPACECRAFT ELEMENT*
P/iofae
Piofae COMA.&L
LAUNCH VEHICLE: Shat££e/ZUS
NO. OF UNITS^
3
3
FLIGHT MODE*:
MISSION PROFILE:
LAUNCH NO.
DESIGN HERITAGE
GaLiie.o P/iofae
Contnactoi'A
BASE FISCAL YEAR*: FY 79*2
COST SPREAD OPTION PROJECT START DATE:
SPECIAL COST GUIDELINES:
System cottt/axc-t:
Co i to Launch •*• 30 dat/4 only
3 RTG'4 (I
15%
46
Suiiwgfay
LAUNCH DATE* FLIGHT TIME* ENCOUNTER TIME*
Ap c£, 1992 (Scutum)
Jan, 1994 (U/ia.nu-4 & Weptane tandem £aanch) .
*Necessary Information
SAI PLANETARY PROGRAM COST MODEL
Input Data Worksheet Page 2
Flight Hardware Definitions
SPACECRAFT ELEMENT:
ENGINEERING:
Structure & Devices:
Thermal, Cabling & Pyro:
Propulsion Inerts:
Att & Articulation Control:
Telecommuni cations:
Antennas:
Command & Data Handling:
Power*: Solar / RTG
Aerodeceleration:
Landing Radar/Altimeter:
SCIENCE
Imaging Mass:
Imaging Resolution:
Particle & Field:
Remote Sensing:
Direct Sensing:
INHERITANCE CLASS PERCENT BY MASS
58.4
•o: 23.9
-0-
trol: -0-
12.9
-0-
g: 15.6
13.4
9?. 9
r: ~°~
-Q-
-0-
-0-
6.9
21.4
kg
kg
kg
kg
kg
kg
kg
kg
kg
kg
kgPPL
kg
kg
kg
BLOCK EXACT MINOR MAJOR NEWBUY REPEAT MOD MOD DESIGN
0.0 100.0 0.0 0.0 0.0
0.0 100.0 0.0 0.0 0.0
0.0 65.0 35.0 0.0 0.0
0.0 65.0 35.0 0.0 0.0
0.0 100.0 0.0 0.0 0.0
0.0 65.0 35.0 0.0 0.0
Vidicon CCD Fax
0.0 100.0 0.0 0.0 0.0
0.0 62.8 37.2 0.0 0.0
*For RTG Power Systems, do not include mass of RTG units
47
SAI PLANETARY PROGRAM COST MODEL
Input Data Worksheet Page 2
Flight Hardware Definitions
SPACECRAFT ELEMENT: fnab<L
ENGINEERING:
Structure & Devices:
Thermal, Cabling & Pyro:
Propulsion Inerts:
Att & Articulation Control:
Telecommunications:
Antennas:
Command & Data Handling:
Power*: Solar RTG _/
Aerodecelerati on:
Landing Radar/Altimeter:
SCIENCE
Imaging Mass:
Imaging Resolution:
Particle & Field:
Remote Sensing:
Direct Sensing:
INHERITANCE CLASS PERCENT BY MASS
183.6
•o: 52.6
16.2
trol: '21.1
31.0
9.1
g: 50.3
/ 35.5
-0-
r: -0-
-0-
-0-
-0-
-0-
-0-
BLOCK EXACT MINOR MAJORBUY REPEAT MOD MOD
kg S.2 1.0 0.0 90.3
kg 0.0 0.0 0.0 100.0
kg S7.0 1.8 0.0 11.2
kg 34.7 65.9 0.0 0.0
kg 22.2 77.8 0.0 0.0
kg 77.0 77.6 77.4 0.0
kg 24.6 75.4 0.0 0.0
kg 6-4.8 0.0 0.0 35.2
kg
kg
kg
PPL Vidicon CCD Fax
kg
kg
kg
NEWDESIGN
0.0
0.0
0.0
0.0
0.0
0.0
0.0
• 0.0
*For RTG Power Systems, do not include mass of RTG units
48
CMCO
in sO wO
6) coi- co
•-"in •s!-cococn5j-HCoc5( •«•3: !__l JC -=f -H i» CM S> IO CO LO .-l
>-<
tf\ *—VJ *•—
o wo
-1- L.
CO!"•
&CM
\£>
•f •» r—! 10 cri
Ci> CO CO
r- co ro
r-- i & CMCM LO id UJ
Eg co -=r -a-CM CVJ *-•
co cn tp co
VO Xco en
&! r— i CO*"*
*
r is; >o CMt} V CM •=*
in co - coin — ' -H
*>• 5-5CO ~<
1C COI***- coCM
_c in to co en co i;» ri CMn cniarMcocoino-^ »t
Oin
«— 2;
) in -* in tg en 10co r- to is; co co f-< M
IE S-_i _c 13 co co co cn fe) en CM nQ .* o '.o in E.O '.o in co cn -a-
CO •-* •-< CO "H CM ^CM
en 63 -H
•* in —•ro CM ID vote> ro ro
UO)
CMCOI
co
CMCnen
L.O.
jzucDra
oUJU C^Z LU
Ul COn: co
•*- &3&2inm25iLn€3i>t~ co p) co eo
O
tu cozca
! C5 ~* CM fc^ ID CO
CO CO CM « CM IO
I— ^ C O V D e i - J & j C M C Oto SOto (sicnin—IIM—'i^coO w — r-j co —4 -«t —•
ai.0oS-Q.
OJC
10 tsi co »*-H i^ cr. roco 01 ID enoo CM c-i
inen
i.o
Qi
6-^t-ot—ia
h-coOO
a: :.i CM m m ro « en ^ •«• co
—« M ^ in CM
cMcois—iin-*MO" icrv. in — vo c\! en en ~ co
—< co co
rn in_l !-ol jc '-o -?J 10 r in n "3, — • co
ojr^ — foLOte jcoco r^CM CM CM c\j —• in ~* r--
-. co T CM in
ex-l->
o
O)-ao
l/lO
Oca
co co —i — < CM
o .c -"r tn urt — m -^ tr. co «ij; ^-H—it^r^rt^-j- en
r-4-Hfvi— COPJCO J5CM
co co 65 en *».-•-• CM
ro — CM co in -H ci co r\iCO
en CM ID co
in «.o ro co -«rin CM co r- tsto in in co_ CO
0)
oc?•oO
CL.
C'-
6a.
4JIVc
r—O-
1.
G-4^DO
• •PC31_Cn0L
O-
1-
lilr-^
Ooin.
H•z.UJs:tj0— JL>J
^5-
I'.1
C
0G'
JOot-n.
••i— i
•PrO£l!'
• —UI
(/) in
ff^(n >,* i-
+J O|~.:
C &
Z3 <fJrt
•P O.cDl---' —L'_
CO
•crcncni.o^i-cncTi'a-
ra -r> !>.! 10 ro — 10 «Lf> CM — • — -H CTi (Si
1-O U! LU CO
O- -J Clco o d o \— uiiu o^s f/i :-': LU s: t/) _ jt.' «-^ - '" ~^-- ^ P—t— < ; jt rr* !T! o "^ t— < "> 2 1-1 c_ O <UJ 1-1 1— < l-H tj (/l
Cl — 1 -^C ! — > ! — ^ *•*cn O <il t^. •-'X '••* -'
oj) < I-H Q LU ai C.1 t,T(j - : K ui -z. -z.
L'J 13 =3 1— _l UJ lilCV< *• T " <t| LJJ CO I/?;TI __i "iC d -" c.^f— 't O i- V. Lu UI I—U !E O < ni cu i- CJ."• pi U! :~ <\" C> O t'-it;.i t-i --i ';--t i Cki l-Z -'*'•1 — r^ I'.- O O i-.! '--.' l~*C/T i— i— o « < ct-: ci
•<*
CM
c0
Ll
~o
r--*UIr. ' j.>< ir,-5 O0 (JPi<£ +*ri- in
W! ij
< i.j - tilO f=-1 ^^
inu0)
oC.'
_T(Ci_r, .
•*CO
•['
G*F.-:0>
i—L:.!
CO <0C/J C'
f*"X
01 >><^ L-•P O•-- TT1
n «Z3 +J
<v•p tJ.CO1
>^-< —l'_
o
s.oiocM-H^'Heoin
co in — • ro co u~i co
0 0'ci :*:>. l-H —D. _l tO
(/) O -
( j t r-^ i^ J—« MJ 06 ^5 n--> ^ CE^ '— 1
LU i— i— i— «tf On _; rr < !- ",
CK O O < ~-<>a < o 1-1 Q —
O ^ 22UI O 1— 73 c3 OiGi •• *-• Gi ;~ CO L-.lrD _: (/) <f s: < o ".;1- -: _i Q-s: T Qo ^; a <j o 2: x (x.rj c.; tv. u; ui •-ni iu O !- -J 1- I !J1- :i:.x:- us .. f~to i— LI. < ;— < o ct;
-*
39
9.
uct Io
n
"OOJ
r*^
UJ* ' +^< in5 0o uf*< -p:r. c
4)_l U
H fUO Q.1- —
c. * zi: ujO to Li.) vj
Q c.
f^ tn i a rz<jl > iu
«« LU s:*i( e^ LIJ
C.I C ''\--X--f.
ct. o <^- <£O ^! Q SICO CO
•(• u • m".' Z 0 :*.
ijj u :•: o.'.
01 Z5 n 6>• < o r-'(,T _J !.": 1.1.
uc t Io
n
h"* "32: auj u:ci jj6 i.l-1 OUJ (J
>UI +>o. =
rt!
-1 U< 1-1- CPOo-i
49
cost estimate for this project. "Science Development" refers only
to the probe science since the carrier spacecraft has no science hard-
ware. "Probe System" and "Carrier System" each include a portion of
the System Support category prorated on the basis of total hardware
cost. With contingency, the total development cost is estimated at$353 Million.
Table 13. Cost Summary for Outer Planet Probe Development Project
FY1982 $MProgram Management/MA&E 18.4Science Development . 39.6Probe System 56.4Carrier System 144.0RTG's 30.0Launch + 30 Days Operations 18.4
Subtotal 306.8APA/Reserve (@ 15%) 46.0Total 352.9
50
6. Conclusions and Further Development
This report has discussed the effort involved in updating and
revising the SAI Planetary Program Cost Model. The Model was updated
to include data from the most recent U.S. planetary missions. It was
functionally revised to increase its flexibility in application to the
broader scope of mission/program scenarios under consideration in
NASA's long-range plans. The Model can be directly applied to cost
such diverse systems as inertial or spin stabilized spacecraft,
atmospheric entry probes and highly autonomous soft landers, and
therefore can more readily extrapolate to systems such as surface
penetrators and rovers.
A detailed error analysis of the Model against the historical
database indicated that, on average, it had captured the information
in the database with an error of less than 10%. In its basic form,
the Model was found to be highly sensitive to the number of hardware
flight units and an adjustment procedure was developed to reduce thissensitivity. This procedure, however, raises the average error of theModel against the database to just less than 20%. Based on this result,
application of 20% contingency is recommended for cost forecasts ofproject definitions which generally fall within the scope of those in
the Model's database. For project definitions that are significantly
different from those in the database, it must be left to the user toassess the Model's validity in generating a cost estimate and to apply
an appropriate contingnecy to the estimate.
In addition to continuing collection and analysis of cost data
from on-going projects, three major development efforts have been
identified to complete the Model and further enhance its capabilities.
The first task is to complete the model revision effort by developingnew algorithms for estimating costs of mission operations and data
analysis. Although historical costs will be used as guidelines, these
new algorithms must take into account the expected effects of planned
51
cost reducing procedures such as the multi-mission end-to-end information
system and reduced cruise phase activity.
The second effort would involve updating the inheritance algorithm
with a more systematic determination of the numerical weighting factors
used in the algorithm. The factors currently in use represent best
estimates of appropriate values. The update will be accomplished by
analyzing cost data from past projects which are known to have benefittedfrom hardware design heritage.
The final task, related to capabilities enhancement, would be to
•develop a new, analytical model to transform a point cost estimate into
annual funding levels. Such a model should account separately for the
different phases of a project, e.g. hardware development versus flight
operations. It must also be capable of dealing with the wide variety
of project implementation scenarios, which can range from relativelysimple Pioneer-class projects to highly complex Viking-class projects.
52
References
1. Pekar, P., A. Friedlander and D. Roberts, "Manpower/Cost Estimation
Model for Automated Planetary Projects", Science Applications, Inc.,September 1973.
2. Kitchen, L. D., "Manpower/Cost Estimation Model for Automated
Planetary Programs -2", Report No. SAI 1-120-399-C2, Science
Applications, Inc., April 1976.
3. Jet Propulsion Laboratory, "Galileo Project Management Report",JPL 1625-92, October 1980.
4. Jet Propulsion Laboratory, "Venus Orbital Imaging Radar: Cost Review",JPL 720-18, June 1978.
5. Jet Propulsion Laboratory, "Comet Rendezvous Development Project:OSS Program Review", JPL 720-32, April 1979.
53
RPPENDIX R
DETfllLED COST DRTRBRSE
Appendix A
Detailed Cost Database
Due to the proprietary nature of the data, this appendix is not in-
cluded in copies of this report intended for distribution external to the
National Aeronautics and Space Administration.
RPPENDIX B
DETRILED DESCRIPTION OFCOST MODEL RLGORITHMS
Appendix B
Detailed Description of
Cost Model Algorithms
The individual algorithms which comprise the SAI Planetary Program
Cost Model are presented in this appendix in both graphical and functional
forms. The function plots also exhibit the data from which the functions
were fitted. Statistics associated with the curve fitting process are alsoshown.
For the hardware-related cost categories, the total labor (DLH)
function is presented with the respective recurring labor (RLH) function
on the facing page. Labor to cost conversion factors are shown for the
total labor functions only.The test statistics associated with the fitted functions include
the correlation coefficient, the t statistics of the estimated coeffi-
cients, and the mean percentage error and mean absolute percentage error
as defined in Section 3.3 of the report.The axes of the function plots are not labeled because of the
proprietary nature of the sample data. Note, however, that the scaling
of each recurring labor plot is the same as its respective total laborplot.
Bl
Page intentionally left blank
Page intentionally left blank
Index to Algorithms
Cost Category Page
Structure & Devices B4
Thermal Control, Cabling & Pyrotechnics B6
Propulsion B8
Attitude & Articulation Control BIOTelecommunications B12Antennas B14
Command & Data Handling B16Solar/Battery Power B18
RTG Power B20
Aerodeceleration Module B22Landing Radar/Altimeter B24
Line-Scan Imaging B26
Vidicon Imaging B28
Particle & Field Instruments . B30Remote Sensing Instruments B32
Direct Sensing & Sampling Instruments B34System Support & Ground Equipment B36
Launch + 30 Days Operations & Ground Software B37
Image Data Development B38
Science Data Development B39Program Management/MA&E B40
B3
COLULJ
LUQ
CJ
ccoooa
oUJcc
ccj—o
GO Or- CO 31 COcn -H . . 3-. . oj o •
cn o — — i co
X,in3'
Oen
xz
05t\J(O
CJ•—iI—CO»—II—cr»—CO
I I
ccocccc
CDOto
•
CO
CJ ~ CO— en =r cc oU. t\J O O CCU. CO CD CC UJUJ • • CC Q.O — O UJCJ — — UJ
t— I—Z O CJ Z 3
•-!•—! t— t— CJ CDcn t— en en cc cncc i—i •—i uj 00
UJ _l t- \— Q_ CC_J UJ CC CCQ- CC t- t— Z ZE cc en en cc cccc o LU ujen o i- i— s: z:
COo
COcco1—CJcrLU
o
COccLU>
OCJ
0
ccoCOcc_l
o1—
cncc13OX
ccoCOu:_j
i—cnoCJ
_icc1 —o1—
oU~Bf^
t—enoo
ccoCOex_i
UJCCID
S§
:Dcc N
COS
X
enencc
\
cc»—o
*- ccg °O LJ
XX a.
U. D
* 60
XX
X
(H1Q) SUnOH UOGtn 133UIQ IBlDi
B4
COLUCJ
LUO
ccoCDCE
OUJCC
CD
CD
CCCCIDCJ
in
inm CM co r- coen 3* c\j • •
• • • «•* t nto o o in i —
DCOccccUJ
inLJ
noccc
I
cj ~ tn uj.•-• en =r cc ou_ en 3- o ccu_ en r- cc ujuj • . cc a.o — o tu
UJ Z CJ O Z 3fvl O •—•>—• UJ —1_ _ I— >_ <_) o
tn H- tn tn cc tnIT i—t «—i n i m
UJ _J t— »— O_ CE_J UJ CE CEQ. CC I- I— Z Zz: cc tn tn CE CECE O UJ UJtn cj »— t— z: z:
LU
CC
CJIDCC
o
tntntr2:
cci—o
1 5
* 6o
synoH uoayi IDBUIQ
B5
CDLJ
LJLU
OCC
ot—,\—
*~z.IDu___
'*—CJ03rri
H-C_)1 1 1cct~\
icc0h-
C^
r-.
3C
XzX
^*CO0oo+
r»e\i
cn2 o SJrt ^^ ^x i - ouj cn
i— i UII ,_cn 3_i *— o
Q cn o
00 CO OCD 31 aCD .
3 . — C" o in -
1 11 111 11 1
i—zUJ^~* *-0•-• ^ uu_ r- cu_ e^ cUJ •o a- c<_> •- » •-
J Z CJ C.Nj ' 3 l—« *•
i •— • 1— 1-~> t— cn o
cn •— -J _J 1— Hj uj cc a. CC 1— t-i cc cn oC 0•j tj t— t-
33 C.3 -
^^
3 a3 e3 a• a3 U
»-J "2" U- C.•> a-• u- ac- i-> a
u- 2
3 CD. .i ID
-«
11
CCOccccUJ
*~ trui O
: o (—3 CC /j: uj g:_ CL._1 ^^»
UJ- •- 2
53 2j i ^; cn cnj CO CC- CC li 1
^>
- cc ^J UJ ^^I Z C_)
ccXV,COe\j r>-
o en•—i •«» en
•• ••i—cn t-o cn0 0t_>cc0 _lCO (XCC 1—_l 0
»—o1- O
t—cntc •—z> cn0 OX CJ
cc cco oCO CDcc ex_J — 1
CD
CDcnLJ
oai
oLJ
cc
ccUJ
S g
\cncnexz:
ccI—o
(H1Q) SUnOH HQ8bn 1D3UIQ IdlOi
B6
o-•CJ"2L
en £^ cc^ ^™^ ^^^—r CD^- CT
CJ H-LU a1— ccO o0= o)— 2
Q_ orcc33
«<jr <->LJCC
OJO (D -- 3« enCO 31 — • •
v . . 3- OCO O CM LO 1 CM
(M00
So
X
EX
-. *°r»- l~~l
CO COI—In 1—
jC«J "^CC CO
H2u_
p
J
LJ — > CM
1
JoccccLU
t—zLU
"— • »— • CO CC <_)u_ en o o ccu. r ID cc LULU • « CC Q_o en o LU<_) — *- uj
LU 2
•-i •-
»— 1—S LJ 0 Z Z)3 »— • I-M 111 i- 1— H- LJ O
V) r- <n in cc tncc — • — • LU CD
UJ _J t— t-- Q_ CC— 1 LU CC CCQ. CC t— t— Z Zz: cc tn en cc ccCC oin cJ 1— 1-
LU LU
CD
CDCECJ
Occ
oCJ
crccLU
Xz
enencc
cct—o
(H1U) HOGbl 133UIQ
B7
CCoCDcr
oUJCC
(O
xZ
00CO
CDin
r- cnr- =r o 3-CD • co • r*-
• If) • O •in o — r- i r-
CO"
1 '1—CO•— 11
cri —CO
LUfvl•— •cnUJiQ_
z:crcn
111
i—
i— iCJi— iu_U-LUOCJ
•z.o1-1—CC_lUJCCCC0CJ
0000_l
•COLO•-•
(J«— 1
—cni —cr»—cnt—
COCO~131.0
—0t— 11—cnt—CC1—cni—
CCoCCCCLU
t—•z.UJoCCLUa.•z.CCLUz
11or0CCCCLU
•z.LUCJCCLUQ_
LU
K—iOcnCDcr•z.crLU
3-in CO
(Om
cn
cno
COCCOf—CJcru_
oCOCCLU
O
CJ
CC0mCC
ot—
cnCC
o
CC0CDcr
h-cn0oi
cri—oi—
o
1—cnoCJ
CCoCDcr
CO
IDQ_OCCQ_
\'\\
\\\\\\\ss
:S
cncncr
cr»—o
o OO
synoH yoaei
B8
o2Z)
oco00cr
CJLUoct—Io
CD
ococ
UJoc
oO)•
o
£
x
O
X_lcc
CO
CO
CEt—(D
oo inoo en c\i c\jCD • « •
• CO O •••r- o en i —•
I
ococcoc
O — UJ— • — OC CJu_ o o ccu_ en cc ujUJ • CC Q_O O UJCJ ~ UJ
UJ "Z O "Z. 3M O «— « UJ _l— — • » - < _ ) OtO h- (D OC V>
CE — • UJ CDui _i t— QL a:_l LU CCQ- OC I— Z Z3T 2C CO CE CCCC O UJ UJwi LJ i— z: z
o
IDQ_ooca_
COtncc
CEI—o
cc t-t- ccg °
OO UJ
K £
O
o\i
synoH uoaui
B9
oI—
oDC
g
o cc5
o
I—cc
a- r a- tn -Hco •** c\j • .
• • • (\j coCD o r- r- i -*
o(O
o
Xz
X
00
CO•
CM (O
en
I I
ccccUJ
O CO O UJ•-• c\j en cc oU. CO C\J O CCu_ • r~ cc ujuj «—« • cc QLO C\J O LU
—• «-• I- I— 0 Oin H- tn tn cc tn
cc — •— uj coUJ _J I- t- 0. CC_l UJ CC CCQ_ CC I— t- Z Z
CC O UJ UJ<n t-) t- t— 32 x:
Oi—<_)CE
DC
O(_>
ccXmCD•
ov-l
<A
tno
ccoCDcc
COCC
r-
00•
CO
• •"
I—tn
o
_jccI—oI—
o
tnoo
cc cco oCQ COcc cc
X
occc
XX
XX
XX
XX
LUOID
XX
X enencc
CE
I—o
CCX
(H1Q) U08b1 133010 10101
BIO
Occ
oCJ
oI—I
I—CE_J
CJ
o>—I
CJ
ccoCOCC
CJLUccoCD
ccccCJLUcc
o r-CD CO • COr- co co •. . -H co
CO O CO I LO
Xz
men
nx_iCE
COCJ
CO
CO
i !
LUI
occccLU
•- ~ I0 r-_ _^ ccu_ co oLL. en ccLU • ccO —• LUCJ —
I I-
CJocLUQ_
LU
LU Z CJ Z =3hg O »- LU _J•-••-• I— CJ OCO I— CO CC (D
CC •—• LU CDLU _J •— Q_ CC_1 LU <XQ_ OC I— Z Z2: oc co cr crCE O LU LUco cj i— a: z
01cr
LUO
cocoCE
CEt—o
cr
2 \
(Him sunoH uoaui IDBHIQ
Bll
cno
CJ
OCO00cc
LJUJOC
crK-O
too<n
x•z.
r-31
•=rnX_la
enLJ
eni—ccI—in
CM Or- n* co en coCD O • • •
O . . — CO C\J—« o =r « i c\j
I I
ccocccc
UJ |
•-1 — v | ZCJ .-^ tO n i—i — O CC CJu. r co o ccu_ =r —« cc ujLU • • oc OLo a* — LU
LU Z CJ CJ Z =>
*-• •— I— I— CJ OCO h- CO CO CC CO
CE —• •-• UJ COLU _J I- I- Q. CE—1 LU CE CEQ. OC I— I— Z Zz: cc co to CE CECE O UJ UJto cj i— »— z: a:
CCXVO)O) CM• in
0) CO•
V* CO
COo
inoco(__, jccU-
"
inocLU
±^of i
occoCOCE-1
O1—
COoc13OX
ccoCO<r_i
H-COOCJ
iCEt—O1—
O
*~t—tooo
ccoCDCE_]
CELJ
OCJUJ
XX
XX
X
X
COCOCE
CE
I—O
x
Xx
miai sdfiQH aoatn IGBUIQ 10101
B12
CO"Z.o
CJ
ocooocc
CJLiJCC
O2•—iCCCCIDCJUJcc
or- ooen o
• •
CO O -H
to en=r
ienCM
in00oo
X
CO
CD
X_Jcc
CDCJ
eni—it—a:i—to
ococcccLU
cj — in••-• CD OO CCU_ C\l OO OLL. CD - CCLU • • CCO — — LU(_) — —
uj<_>CCLU
LU
LU Z O O Z 15M O •-• •— LU _l«—• •—i I— I— CJ Oto t— en to cc in
cc •-• —• LU caLU _J I— I— Q_ d_l LU CC CEQ_ CC t- I— Z 2n cc en en cc CECX O LU LUv> o *— i— s: 2:
CECJ
\\ o
oCJLU
LU
\\
\\
\\
O
inina:
\
CEt—O
\
O
\ "-
\ *.
sunoH uoaoi ioama
B13
DCOCOCC
oct—1
o—Icch-o
=r oco —• co t\icn cr> • • c\j
in oin o co —«
ooCO
z:xz.
COCOO
to
X_Jo
cno
cn
cci—cn
ini I
ccoccccLU
t_> _ CO LU•— CO =T CC OU_ CO CO O CCU_ O « DC UJLU • • £C Q_O CO —« UJ(_)*-"••• ii i
UJ Z C_) O Z =}M O — •-• UJ _l
—• •— 1— I— O Oto i— to co oc toCC —i —i UJ 03
UJ _l I— I— Q_ CC-J UJ CC OCQ. CC t— t- -Z Zz: cc to co cc ccCC O UJ LUto <_> i— i— 2: z:
cnoco
o1—4
cnDCLU
OLJ
ccX
toco to
CDCO 3-
W CO
to i—o toCJ O
CDcc
COCC H-r3 coo cX C_>
CE CCo oCD CC3cc cc
COcn
UJI—^cn
totocc
ccI—o
X S O
X
O XX
X
mien sunoH uoatn io 10101
B14
Li-
CCOOQ<X
CJLJCC
o
OCCCIDLJLJJOC
CDCOen
•CO
CDCJ
CO
J—(XJ—(D
CCoCCCC
00CT> — C\J 00CD 3* • •. • CD UD
in o cr> i oj
LU
O ~•— CD CC OU. CO O CCU_ CO CC LUUJ . CC Q_O CO LUCJ — LU
LU Z O Z =)M O "-• LU _J•-• —• I— CJ Otf> t— in cc en
CC *—• LU CDLU _i i— o_ a:_l LU CEa. cc H- z zs: cc <n a: ccCC O LU LUCO O I- 21 S
COen
LUI—:z:CC
enCOCC
CCI—o
sunoH yoGui IDBUIQ
B15
o
CD
Q
CE
ccCQd
e_>LJoc
cri—o
to oin o cn inen . in . .
• CD • CD COCD O — CO I CM
I I
CCocccc
=rCMO
coC\J
_Jo
ene_>
en
tn
iI
UJ,
UJ
LJ —•• UJ—• —. IT CC LJu. co r\j o ccu. r\j o cc ujUJ • • CC Q_O 3" O UJO — — uj
LJ LJ Z ID
«—i «-• t— I— O Oco i— to cn cc to
CE —• — UJ CDUJ _l t— t— Q_ CE_j LLJ CE CEQ. CC t- t— Z Z2: cc tn tn CE CECE O UJ UJto LJ »— i— s: z:
cc
COco cn. to
CD —V* CO
tno
enccoi__,jexu_
0^ 1
enccUJ
i L0
LJ
CCOCOCE
-1O1—
tncc=3OX
ccoCOCE
t—cnoo_jCE1—o1—
ot™«^
1—to0LJ
CCOCOCE
CE
CEQ
Q
CE
OCJ
\\
\\
totoCE
CCI—O
sunoH yosoi loayia 10101
B16
ocr
zo
(_JzZD
ccoCDCC
UJDC
ao
QCQC
LUQC
O =rco 3« r* • inCD CO CO CM •. . . — CO
CO O CO CO I CO
COCO
CMo
COCOCO•
fM
Q-XLU
Hw.
UJ "•-• COo ~
11CM
i-" co r -U_ COU. CO
CM0
1 UJ . •1 o
LJ"1 '
1—CDi— i(_
CCJ—to
CM 0
ccoCCCCLU
1 CJfc__
LU Z CJ
11
CC0ccccLU
1—ZUJCJccUJa.LU_
CJ Z =)rvi o •-• •—•— 1 1-•4 1-CO 1— CO
(-CO
cc — • «UJ _j—1 UJa. ccs: cccc o
(-CC1—CO
1-
LUCJCCLUQ_
i0COCOcc
cch-COZccLU
zccLU
CO CJ 1— 1— Z 21
CEo
Q
CE
oCJ
\\
(H1U) SUHOH HD8U1 133UIQ
\ _r»-
\
\012 1
xz
COCOex
ccI—o
B17
ocLU
O0_
o»— 1•
0
ZDu_OCOCDCC
f—0LUOC
O
_Jcc1p^CDf—
~~ rz:X"Z.
X
«— 1^-*CD0O
0
+
COCDCD
inCO (_j U^^ frQ. r* -x •— oLU CO
•—i U
fT ^X 3_i *— aa in v
j4^
T
JJ
:•>
f\jCOCD
•
O
11111
1—
sCJ^*U.U.LUOCJ
•z.o1—CE_lLUCCCCOCJ
O
P
.
PClc
0«_
c.^1-c^hG1-0
1-
r•)•
rD
r>DD•
•»mf
J•i
r)•A
C
D
— -<CO
•
^H
«_l
~
»-
CDOO
•
O*-*
CJI—It—CO
»—CE1—CO
1—
—H
•— 1
1
CCoococLU
t—•z.LUCJCCLUQ_
•z.crLUz
r>-•
OJ«— •
11oc0ococLU
,LUCJCCLUa.
LUH-
Z)_JoCOCOCE
•z.CELUZ
OC01—
CCu_
2;ED
inocUJ
^^^DCJ
CCX"V-^3-•
CD
-^w,«
1—CO0CJ
CC0COCE_J
01—
COCCIDOX
CCoCOcr_j
CO
•r-t
m
CO
.»
t—COoCJ
—Icri —ot—
Dl—
t—COooCCoCDCC_J
CCUJ
CEOQ
OCcroCO
XN
\
I
COCOcr
CE
o
o
tHIQ) SUnOH yOQBT 103010 IblQi
B18
o
CCUJ
oQ_
ccoCQCE
LJUJCC
O
O
CCcc
LUOC
Xz
oCM
o
o
CO
in•
CM
a.X
nxcc
tOO' 'H-toi— i^___'•'-i~"to
0
uhfr>
0
u
Q2GV
3
Jg•4
T
JJ.
Er>
COin
•
o
i—UJ»— 1
oI-*11U-LUO0
zol«4
1—cc_lUJccccoo
in3-
„
—
^^CO.—1LO
•
CM"~
O
(—cni— ii —ccI—CO
*—
fl
0
^
cccc«•
cfK-u1»
t-01-0
h
T
•
D
^
r3U1
D•
D
J•
rj•H
:T
C
aeaau
i-*iUcaua^cu2
U
r
F—
5rEJ
JJCJL
£
Ej
cn.JCM
11
CCoccccLU
1—ZLU(_>rrLUQ_
UJ1—
_lOcnCDcc
zccUJ2=
ccLU
CECQ
CCCT
OCO
s.
\
COencc
cci—o
(Him synoH ao9bi loayio
B19
ro tor- o -^ — «CD • c\j • on
o
QCOCOcr
CJUJcc
crH-O
CD O -« 3- CO
ininOT
xZ
oCO
inton
xa
COCJ•— •1 —COl_ll_
crh—CO
u11
^tlU
a3a(j
H
2U*•»L^uuucc
J 2•1 C~t I-
•> t-G
J _J U_ a: ac eT (.
»
JH
*J
-1
J3J
£3-i
CJJ
5j
cpu0
c.*•1-0h-1-01-0
1-
•
Dr>m
•>3_<•
J•4
1
•H
C
D
-*
ininCO
•
o—
0
V—COI-H
1—cc1—CO
1—
aeaaut-zLt.auaicua
c3
j
rjJCJL
rcj—
II
ccoccccUJ
t—zoccUJa.
UJi—^j0COCDCC
CCUJ
in r-r-
CD «-*
w en
too
COcco1—CJcru_
oCOccUJ
oCJ
u
ccoCDcc_J
o1—
tocc=)oX
ccoCOcc_J
1—tooCJ
_Jcc1—o1—
o
1—to00
cc0COcc_l
ccUJ
oQ_
CJDI—DC
\\\\\\\
Xz
totocc
cc»—o
\N
(HID) synoH yoaoi i33«io
B20
LJ
13U.
CCoCOCE
LJUJCC
O
O
CCCC
LUCC
in
*:•x.
x
CC
•
I—nCO
CO
r
uril~
U
u
Q2ay
u
j-4•*
•)
JJ
:•>
0•
— •
H~*•z.LU«— «CJ«— 1u.u.LUCDCJ
Zo<— tt—CC_JLUCCCCoCJ
-X
^
00
r
c»•
1-01"
I-G1-u
t-
^
DD•
IM
J•*
•)-1
C
-)
*
^^
LD^H
r-•
o~-
o1—11—in*—*t—CC1—to
t—
cc
QeQQLI
1-
*UCQUQ
CLI2
D•
D
CDCCJ
FJJCJL
ZLJ
0•
0
11CCoCCCCLU
1—ZLUCJCCLUQ_
LU1—r>ioCOCOCC
•z.CCLUf
O)+J
c
OJ
OJ
c•1—
QJs_
to<->
•r^4_>
in
•i-jre
-4-J
I—
*
DC
LJ
O
Q_
CDJ—DC
\\\
X o
s\
\
\X
XX
COtoenr:
cci—o
XX
o
(H1U) U09H1 133010 O N I U U n 3 3 U
B21
LU
zol_l1 —CJzD
Lu
CCOCDcr_J
1 —CJll 1
DC1 'UJ
_lCEJ—^D1—
n
(O
00
d
3ExZ•"»
XCO
— « f 1 LlCC ~~ K3* " *•
K- oen CO
^ , L
ll ,_
E 1_j *— cQ CO 0
_*3* OT oo en men r>- r~ • •
• • • 3* in•> o o e\i i e\j
1 1 1 1 t1 11 cc1 01 cc
1 cct- 1 LUz 1 rnLU 1— U)
— • ~ z CC(_) ~ CD LU O« — — cc ej H-u. CD =r o cc f ju_ =r <o cc LU jLU • • CC Q. ~o en o LU Lue_) — ~- LU
•- •— zj z e_) e_) z ^ IT-)j O «-• «-• UJ _J *— '•« •— ' H- h— O O ^"^•> i— en en cc en CO
CC •— • •— • LU CD CCJ _l t— t— D_ CC UJj LU cc er ->- CC t- t- Z Z ^i cc en en cc CE -c-CO LU LU C5T o i— i- z: 2: CJ
ccXN.cnp- eo
• cnO CO»— i •w on
t* ••i—en i—o en(_) O
cc0 _lCD CCCC 1—_l 0
1—oI— O
1—encc t-3 en0 0^ occ cc0 0CD CDCC CL_J _l
o
crccUJ
CJLUQOCCLUCE
x
\s
XX
XX
XX
XX
XX
X
enenccs:
txi—o
oo
xxo. X
X
(Him sunoH yosbi i33yia 10101
B22
LUi— i
••M—
1
QO"5"
o»—i•0
Z)u.
cc0COCE
1—u1 1 1oc0
o1— 1
ccccCJLUcc
CO3« =T CO C\Jo> r- • •
• • O COCO O CD 1 -^
1 1 1 11 11
1 j cc
*—Z-X
—t—ccotoX
1 cccc
1— LU•z.LU 1—
CJ ~ LU•-• C\J CC CJLU CO 0 CCLU CO CC LULU • CC Q_O J LUCJ — LU
CO »- H-OJ /j LU 2 CJ Z Z)tO •-' 1x4 O «-" LU _JtO ^ •— — • 1- CJ 0
• T— co i— co cc co3- CO CC — • LU CO
r— i UJ _l 1— Q_ CCH 1 _J LU CE
J-j- to. cc t- -z. -z.i H- r: cc co CE cc_J »~~ CC O LU LUtc CO co cj i— z: s
O
CEOCLU
LUCJLUQOCC
CE
totoCE
CEt—O
synoH aoaui IDBUIQ
B23
DC
LU
crccCLoenDC
en en —«en —« in t\jO)
Ot—•I—O
ocoCQCE
O
CC
co —• oco o — ru i
• CO
eni>-inen
OJ0
D
X
O
en
i—ini—CE
in
i ii ii i
cccc
LJ en co uj—• o r^ cc <_>-u_ 3- in o ccU_ • CD CC LUuj -^ • cc CLO —« O UJ<_) — — UJ
UJ Z CJ O Z =3NJ O 1-1 1-1 UJ _l_ _. I— I— (_, Oco i— en en cc to
CC _ _ LJ CQUJ _1 I— I— Q_ CC_l UJ CC CCCL. CC I— H- Z Zz cc en en cc cccc o LU ujco e_> h- i— z: 2:
OCOI—LJ
COOC
(_)
CCX
COO CO• in
o —
v» co
en i—o tno o
cco _jCD CCCC H-
enCC
cooCJ
cc cco oCQ CDCi. CL
x
N
CD
I—I
a
CE
XX
X
XX
XX
XX
XX
X
COtoa:
<ri—o
XX
XX
X
x°
(H1Q) d09U1 133UIQ IdiOl
B24
CT)
QCUJ
LU
CE
CCcroCEQC
CD
enCD O C\J
o -* en o or>
u_
DCDCOCE_J
1—LJUJCC^™*
ao2
DCCC
LJUJCC
(S.
(Om<M
£X
5X
C\J
•
u
I_1cc
1
H
2
•-L^
UI uI u
9
J•4
J ~-• o_ r_ 0j
^
^ p•> ua nj r•
~•i.
D aT CJ 0. a
11
ccoccccUJ
k—p ™
ZUJ
c c_>D CCI UJc a.
cn
o _ «i <_)
UJ Z O (_) Z 13rvi o >—' >—• uj _i•-< •_ i— i— o oco i— co co cc co
CE •—"•-. LU COuj _i t— t— a. CE_J UJ CC CEa. oc i— t— z zz: cc co co CE crCE O UJ UJCO O I— I— SI S
\\\\.
O
CE
\
COCOCE
CEt—O
*(H1U) SUnOH U09U1 133BIQ
B25
CD
CDcr
LJ
ccoCOCE
oUJcc
CEI—D
inCM
entoo
toLJ
(DI—I»—CEH-(O
r^ coCD r- r- 3-CD en • • co
• • c\j o •co o co —• i r^
I
ccocccc
e_> en o uj•—i eo I*- cc eju. o in o cca. • oj cc LUuj o • cc o_o ~* — mio >- — uj
uj z e_> e_> z =>1x4 o «-»•—• uj _i•-• •—• i— i— ej oen i— en en cc en
rr" i—i i-^ 111 mUJ _l I— t— Q_ CE_l LU CE CEQ_ CC I— I- Z Zz: cc en en er erCE O Hi ni<r> <-> t- t- ^ z:
ccDC
r-in
oeo
•
en
en0
enccoi—LJCEU_
O
COCCUJ>^»1oLJ
0
CC0COCE_J
0t—
enccZ)0:cCC0COCE_l
^—en0o
_iCE(—OI—
Ofc»^^
1—enoe_>
ccoCOCE—i
crCJeni
LU
X
XX
XX
XX
XX
XX
X
enen
CEi—o
X
\\
(H1Q) SUnOH UOSbl 133UIQ 10101
B26
CD
CDCE
LJ
OCOCOcr
LUOC»—IQ
O
ccOC
UJOC
OT =rCDo
co o. en
CM —oi CO
o>ooo
CD00or
X_1cc
COCJ
CO
I—CE•—CO
i- Iz |UJ I
—• enu_ oou. enUJ •O — «o —
UJ
CO t— <DCC — •
LU _J t—_l UJ CEQ_ CC I—z cc tnCC OCO LJ I—
I I I
ccoccccUJ
en LUco cc <_>o o cc3- cc LU
• CC Q_— UJ— uj
t— t—O Z ZD•— UJ _lt— (_) OID CC ID— • UJ 03I— Q_ <ICCt— Z Zcn cc cr
UJ UJ
crCJen
COCOCE
CEI—o
sunoH yoaui IDBUIQ
B27
OCo03CC
LJOZ
crv—o
entotoo
CDCD3=
COf \_1 'H—COi— it—crH—CO
n
uM^0
u
Q2au
Gaa
1 C
i-2u
C.»•
uLiuCC
J 2J C-• 1-•> H
GJ _J U. a: ac c•> c
Dr>n•D
rJ
J•H
J
DJ
W
5-«
CJJ
5j
C
u^
^"
0C!
*•
L^1-0^Hai-0
i-
r3•
->M
^
T3r•
ri*
j^
O•4
C
T
^H
in•
31
31
COCOCOo
•
—*-"
o1-1—to1— 1
*—cc1—CO
t—
C
C
QeQau
»-2uC.QUQ
ZGU2
3•
3
3
J
rJJ
J
»
Ej
LO•
O
11
ccoccccLU
t—
OccLUa.
LUf—
1oCOCOcczccLU3C
CDcr
N
oCJ
XX
cc3:v.OJin
•
CO
CDCOin
•
en
COo
CO01Oh-
CCu_
oCOQZLU
O
CCoCOcc-Jo1—
COcc13031
CC0COa.
H-CO0o
1cc1—0I—
0
1—to0CJ
cc0COcc
oX
X
X
XX
X
X
totocc
ccI—o
XX
X
oX
XX
XX
(mm sunoH yoayn io3yio 10101
B28
LJ
OCOCOcr
UJcc
GCccrDCJUJcc
CO
CO O
c\jc\j
3- O O
in
z:x
x
o•
•4
D
X_lcc
CO
•—it—COI—It—exI—CO
I II I
Z IUJ I
ccoCCccUJ
o ~« CM cc uu. in o ccu_ — cc ujUJ • CC Q_O — UJO — UJ
UJ Z (_> Z =>O —• UJ _l
i— (_) O'f) CC tO
CC —• UJ CDH- a. cc
—I UJ CCQ. CC t- Z Zz cc to cc ccCC O UJ UJto o K- z: 2:
to
UJ
CE
LJ
(0COcc
ccI—o
synoH uoayi I D E H I Q
B29
01I—
LU
IDDCf—CO
Q_lUJ
LU_JCJ
DCCCQ_
o•—it—CJ
QCOOQ
CJ
tn(\i
X•z
x
CD
in
D
tn
tn
a
ur«»•
u
u
Q3QU
T
Jsi
^T
JJ
:•>
rcc
c
H2U*•
C»-
uuucc
*e•-i-G
UQQec
Dnm
3
V
j-I
j•4
_
J3J
»
5-t
CJJ
3J
aa
u^
c
C
Ur
cM
1-
o1-
»-ai-0
h
D•)•0^
3rD•
•>U
J•4
T-1
C
r)
onCO
•CO-*
X-.ini
C\Jr-
•o^^
CJ«— 11—tn«— ii—cc1—tn
i—
c
c
acQau
2uLGLC
*CL2
3•D
CDCCJ
F
LIJCJL
*Ej—
to•C\J
11cc0ccccLU
1—ZLU(_)CCLUa.
UJ^^
_jotnCOtr
2ccII 1s:
\\
x
X
XX
XX
XX
XX
CC
OJto
O
QCLU
CJ
CO
en
tn H-o tnCJ O
CJccO _1CD CCCC I—_l O
tncc i—13 tno on: cjcc cco oCD COcc cc
XX
X
If.•z.
tntncc
cct—o
X
sunoH uoaoT i33yia 10101
B30
en
UJ
IDDCI—cn
CJ
ocoCDCE
CJUJOC
D
O
OCOCCDCJUJOC
CDCDCD
CD 3"-H P-
• CM GOCO O I
(Or-O)
oCDr-
o
n
(0CJ
CO
CE
UJi—iCJ
UJoCJ
O 3-• •
O —•
! l
CC
CCOCUJ
— CD UJo r- cc cjCD CD O OCr~ on cc uj. . OC 0.
O — UJ— — UJ
UJo —• —•
to t—CC
UJ —1—J UJQ- CCz: ccCE OCO CJ
I— I— O OCO CO CC CO•— •— UJ CD*— »— O_ CECC CCI— I— "Z. 2CO CO CE CE
UJ UJ
O_JUJ
UJ_JCJ
DCcrQ_
\
\
\N
\\\\
COCOCC
CCI—o
V
sunOH uoatn i03uio 9Niyyno3u
B31
cn
LU
GCI—cn
oi—oZD
OCoCOcr
LJLUOC
CE1—O
oO)o>
x•z.
X
CO
oCOCO
cn
>—ii—cn
H-cri—cn
o CDr*- =r a1 c\j LOcn • cn • •
• CM . — c\jCD o — r- i —
I I
CCoOCOCLU
•z.CJ CO O LUi—i — cn oc LJu_ o cn o ocU. • LO OC LULU CO • 0= Q_O CD O LUO — ~- LU
LU -Z. LJ LJ Z =>HJ O •— —• LU _l*-• i V- I— LJ OCO »— CO CO OC CO
CE —• •— LU COLU _J t— t— Q_ CE—I LU CE CEQ_ CC I- I— -Z. -Z.ZI CC CO CO CE CECE O LU LUCO LJ I— +- 2C Z
LOCD CO
COo r\jW> CO
CO0
enOCoh-
crLl_
O"
cnCCLU
^^^_oCJ
oCC0COCE_J
CD1—
COOCCDO3:
CCoCOCC_l
1—COo
-JCEt—01—
o^—
1—COooCCoCOCC_J
cn
LUCO
\\\\\\\
LUF—O
LUQC
\
X\\
xX
COCOcrz:
CEK—O
XXX
X
synoH uoaoi 133010 10101
B32
CO1—~2LLU^~— \)CC1 —CO
OCoCDCE-J
1—
LJOC
0
ozOCOC—\
LJUJOC
<nO)<n00
O
Xz
X
=r
orJ
D
X
OC
COLJ
inCO CD O) —* OCD n CP • •• • • •-•* 3"
o in r^ i —«I I
ccoOCccUJ
O .-* CO UJ•— =r co cc ou_ r- co o ocU_ O CO CC LUUJ • . OC D.o r~ o ujO " ~ UJ
UJ Z O (_> Z =7r«j o •— •— uj _i•->•-• I— i— (_) OCO t- CO CO OC (O
CE •— —• LU OQUJ _l I— t— Q_ CE—I UJ CE CEQ_ OC I— I— Z ZZ OC CO CO CE CECE O UJ UJto o i— i— a: z:
CD
I—I
CO\ o
CO
LUI—O
LUCC
\ o\
\
\\\
COCOCE
CE
I—O
\
c s.jk 6
\\\
(Hid) synoH uosbi 102010
B33
CO11 —
J*r
UJ1 1 1.
3
DCI1 —CO-^^
1— — 1
CD^^
o|_-(
1—
ll
CCoCQCC
H™oLJnr**—*""*
CD
_Jcc1—OH~
iv<O
(VI
^^z:X•z.
X
2(**••-J-* *"" ~*. H-
to CO*— <" «—
5 £a en
t»
Ut»»-u
U
azau
enenen•
•) o
111
i—•z.LU
(_)>—iU.u_LUOe_>
J Zg O
•) t-cc
J _JJ LU_ CC: ccc o•j e_>
3«=r.en^^
^^enr^-•"
CD
eji— •^~enH-CC1—en
h-
3"CD•
r-en
1
|1
1
r^enr-evj^M
^^
,_,
>—>
t—eni—ert—eni—
o•o
1111
CC0CCCCLU
f—•zLUejccLUa.
•z.CCLUz:
en•
e\i
11
ccoocccLU
1—•z.LUoocLUQ_
LU1—
_JOencocc
•z.ccLUz:
COQCO1—
CCu_
aV— »
1 'COOCUJ
c(_)
CCXX,LnLO.
en
V*
••i—enoej
oco00ccJ
o1—
encc~^oX
oco00cc_J
2yin3<
•en••
i—enooi
cci—o1—
ot—
1—enooccoCOccj
Q_
CECO
O
CO
QJCO
CJLUDC
X
XX
X
XX
enencr
oci—o
XX
\\ i
\\
sunoH yoatn 10101
B34
CO11 —_ -
^—LU^>z_ .^^_J
OC11 —CO~^r_
1 — 1
CD^^
ot f,f~"CJ
IDu_OCoCOcc
1—CJ
OC' 1
ao"Z.*••*
OCOC
CJLUOC
c
o>=*—,^^
Xz
to* CJ L
C ' *"""* ^>— c•—i L
n ._V~ £
31 :™^ I_J f~ CCC CO I
co r-r^ o to LOCD • • •
• LO CO OO O CO 1 CO
1 1 11 1ccocc
1 cc1— 1 UJ"Z. jUJ 1—
LJ P"* 1 1 1I »-• CD CC O| u. -~ O CC1 U. 31 CC UJ1 UJ • CC Q_1 O -~ UJ1 O ~ UJ
1— 1—U Z (J Z IDtj r~i i— i n i i- •-• 1— CJ Oo i— tn cc to
CC — UJ CDU _J 1— Q_ CC-1 UJ CC1- CC 1- Z Zc cc en cc ccCO UJ UJ0 CJ 1— 3C 2:
cxCO
CD
CO
UJCO
CJUJDC
tocc3=
CCI—o
O
sunoH yoGbi ioauia
B35
en en ccto co CM in xen • ^
_ . oZl ~1
h—
ZID
1 L_1
~2L °cLU m• fT•^— 1
Q_ i —i — i CJ
=> £C o r *
LU iorC«)
^^ I—
«M to o i rvrt
COo>d
UJ2aX
x_lal<*J
x
r- COD! <->vD{if/
0 CO,_,
" H-1 £_l >—o to
H2ii
=1
*— • >-cj r
-t . 3- V.• o • in- i co in to
• r-I I 0 0I I —I W CO1 cc1 P" > •• ••
I CC 1-CC tO H-LU O CO
! i- CO. _ ' z cc~ CM LU O
•— CM -- CC CJ 1—U. tO CO O CC (_ju_ en en cc LU i::LU • • CC Q_ u-O o o LU u-c
LU 2rsi ei- »-
J - «• LUi— i— y
= 0 CJ Z =) O
D *•-• »-
p^ *•- »-
^ *1
' ._'
- CJ 0 *"*to i— to co cc co tn
CE —• —• UJ CD DCUJ _J 1-- 1-- Q- CE LJ_J LU CE CE ->Q. CC t— t— Z Z ^Z CC tO CO CE CE ^CE OtO CJ »— t-
LU LU C_>- H 31 t_>
CJ O
CC0 _lCD CECE 1——1 0
1—O1— O
1—tocc »—=3 COO OX CJ
cc cco oCO COCE CC_l _J
OCCLD
DCOQ_Q_IDCD
LUh-CO
CO
UJ2a
X
NN
N
"- XX
XX
XX
XX
XX
<nccDO
CDCE
UJCCCE2accCE
CEt—o
sanoH uoatn 133010 IBIOI
B36
UJcccr. m
_ ^h
I
1I iU—oCO
o*r
V-JoccCD
o!__•i —LJ"Z.IDU.
rr*^^>
OCOcr__ j
\—oLUCC
D
_|croi—
r\jenen
•LO O
11
LU |3 i<=> 1X
1—X Z-1 LUa ~
ol sj i—i— U.
u_X LU
00 CJ0 COCO LJ LU Zen ^ t*t o° i- lo H^0 CO CC
i— i LU _JII i_ _J LU
rr Q- ccx "- z cc_i I— cc oa CO co o
COLO
•
=pCO
•
COoCOeno
•
o
CJ
1—CO
1—ac»—CO
H-
3-•
O1
ccoccccLU
t—ZLU
CCLUa.zccLU3:
—•
CD
11cc0ccccLU _
H~•JF fV~
LU O<-> 1—CC /_j
" P1u_
LUt— ^r• ^D__l
CO COm CCCC LU
•Z ^cc -i-n i C3z C_)
ccX••x^—ir^-.
0—«w
i—CO0CJ
cc0CDCC
-1o1—
COcc.ID0X
cc0COcc_l
— 1«^_CM•
CO
~t—COoCJ
1a:i—o
ot—
t—COoCJ
cc0cacc
CDQ_
cr
oCO
CJ
:Dcr
XX
XX
XX
X
XX
XX
XX
X oX
X
XX
x 6 XX
XX
COcc3o
ccoCOcc
cccc2oocccX
ccI—o
(Him synoH yoetn
B37
ccoCO(X
CJLUCC
LU
o —LO cn r^ =rCD • *"^ • CO
• « "* • t"^ •co o i co i r»
oa>CM(O
O_a.
o•
o
u
X_la
(O
CO
I II l
Ii !
ccccLU
UJ —«-• 3" ~U C\J O LU
•— — CD CC OU. O CM O CCU- O CO CC LULU • • CC Q_O O -* UJO — ~ LU
LU Z O (_> Z =)tvl O •— •— LU _l«—• •— t— I— U Otn i— to to cc to
CC •— •— LU CDLU _j i— i— o_ cc_l LU CC CC0- CC I— I— Z Zs: cc to to cc ccCC O UJ LU<r> <-) t- t- x: s:
to=r o
CO
CO
too
IDfv"Oh-0
L-
O1 *
COccLU
^•C-JC_)
0
cc0CDCC
O
I—
tocc3oX
ccocca._i
i—too0
_Jcc»—0t—
o11
r—toooccoCDCC
1
LU>LUQ
CE
CUQ
CDcn
Q-0.
LU
XX\
XX\
CCLUQ_
to
X»-o_
\
sunoH uoatn IHIOI
B38
o
oZID
OCOCDCL
oc
O
CM -« *
I
o>COPO<o
(OCOCO
X_la
en
en
o o• •
"* o oI I
UJ
OJ
OCX
COr-•
CM
r>-COCO
*f» CO
ccccUJ
CJ CD CO•— CO CO CC CJU. CO CM O CCu_ • en cc ujuj r-» . cc a.O CM O UJ
UJ Z <_><_> Z 3M O •-••-• UJ _l
<_>(/)o
in enCO
uj _j t— »— a. cc_j uj CC CCQ- OC I- I- Z Zz cc en tn cc cccc o uj ujen o i- i— a: z
OCUJ>
(_)
o0
enoo
cco _iCD excc i—_j o
enOC
OCoCDcc
enooocoCDex
UJ
UJo
cro
o
UJtoencc
ff £e °g |
o
LUo
otn
\
CH1Q) SUHOH U08U1 133UIQ IbiOl
839
LU
UJ
LUCDCC
cc
CECCO
CCQ_
o en enen -
CC- CM 3" X
cn . .Z0
1 —0zIDu_
cc0COCE
1
1 —CJll 1ff~
*"""*o_1(X•r~—f 't—
• CO *•in o
11
otoo>oJ^X
aU-J*"*
X
r*"en co0 (_>o
^^
i—0 CO
1— 1II 1
X ^^_) 1—a co
i-
-
zLU —•— • r- ^
• CM X,i o • r-
^O CT) O
r in inCO
I — 101 ~-
«» CMccF^ •• ••
CC 1—OC CO H-uj o in
£ £ii i ^3
—• o r- cc LJ j—U_ O ID O CC fju_ —• cn cc uj ,?.UJ • • CC Q_ *-*-0 0 0 LU LL.l_) — — LU
1-UJ Z 00 2M o — • -•— • i-in H
" h- 1-
- *- zz =) P->
- LU _J-oo rr- in (n oc in tn
cc •-• »•LU -J K-- 1-
- LU m CC- o_ cr LU
_i LU cr cc ->O_ CC i— i— Z Z -,TC. aI 0D 01 Ci crcr o LU uj <uin <_> \— i— 2IX CO
<_) 00
cc0 _lCD crCC 1—_J O
i —o1— O
1 —<nCC 1-3 inO 0X 0
cc cc0 0CO CDCC CL_J __1
X
NX
X
X
XX
XX
XX
XX
x=-
XX o
XX
x
X
VIccZ)oX
ccocccc
LJ
occQ_
CCI—o
XX
(Him synoH uoaui IBIOI
B40
flPPENDIX C
I N H E R I T R N C E MODELj
Appendix C
Inheritance Model
The following pages define the inheritance classes and associated
cost reduction algorithm used with the SAI Planetary Program Cost Model.
Although the definitions refer to "cost", in practice the inheritancealgorithm is applied individually to the subsystem labor hour estimates
prior to conversion from hours to cost.
The underlying assumption upon which this inheritance model is founded
is that heritage in design philosophy and/or physical hardware affects
only the non-recurring portion of cost. Thus, at the highest extreme of
heritage, a project which uses unaltered residual hardware from a pre-
vious project still incurs a transfer cost exactly equal to the recurring
cost of the hardware item. At intermediate levels of heritage, a fixedpercentage of the non-recurring cost is incurred depending on the in-
heritance class.The inheritance class definitions were determined by mutual consent
of cognizant personnel at NASA, JPL, and SAI. The inheritance class
weighting factors were not derived by analytical techniques but were also
developed by consensus agreement.
SAI PLANETARY PROGRAM COST MODEL
INHERITANCE CLASS DEFINITIONS
e Class One: Off-the-Shelf/Block Buy.
The subsystem is taken off of the shelf in working conditionor ordered while the normal production line is operating asan additional unit.
e Inheritance = 100% of non-recurring cost (NRC)
o Cost = recurring cost (RC)
c Class Two: Exact Repeat of Subsystem.
The exact repeat of previous subsystem but to be used inslightly different spacecraft or after line has closed down.Only design work is needed.
e Inheritance = 80% of NRC
t Cost = 20% of NRC + 100% of RC
e Class Three: Minor Modifications of Subsystem.
A previous design is required but it requires minor modifi-cations. Thus, the spacecraft will still incur all thedesign cost and most of the test and development cost inensuring compatibility of the old design and the newminor mods with the new use of the subsystem.
e Inheritance = 25% NRCe Cost = 75% of NRC + 100% of RC
e Class Four: Major Modifications of Subsystem.
A previous design is required but major modifications have tobe made to the design. This gets very close to a new subsystemsince even new subsystems rely on previous design and experience.Some savings in development is possible.
» Inheritance = 5% NRCe Cost = 95% of NRC + 100% RC
Class Five: New Subsystem.
The subsystem is basically new design.
e Inheritance = 0% NRC» Cost = 100% NRC + 100% RC
Cost Reduction Algorithmby Inheritance Classes
Let Xj = Percent of Subsystem Off-the-ShelfX2 = Percent of Subsystem Exact RepeatX3 = Percent of Subsystem Minor ModXit = Percent of Subsystem Major ModX5 = Percent of Subsystem New Design
Thus X], + X2 + X3 + X4 + X5 = 100% of Subsystem Mass
NRC = Non-recurring cost estimate (without inheritance)RC = Recurring cost estimateTC = Total cost estimate (including inheritance effects)Z = Percent cost reduction
If Z = l.OXj + 0.8X2 + 0.25X3 + O.OSX^ + O.OX5
Then TC = (100% - Z) NRC + RC
flPPENDIX D
DETfllLED ERROR RNflLYSIS
Appendix D
Detailed Error Analysis
Due to the proprietary nature of the data, this appendix is not in-
cluded in copies of this report intended for distribution external to the
National Aeronautics and Space Administration.