Title: NAFCOM Probabilistic Cost Analysis Module presented by Sharon Winn NASA Cost Symposium March 10, 200
1NAFCOMProbabilistic Cost Analysis
Modulepresented bySharon WinnNASA Cost
Symposium March 10, 2004
2NAFCOM Description
- NAFCOM is a parametric estimating tool for space
hardware. - It is intended to be used in the very early
phases of a development project. - NAFCOM can be used at the subsystem or component
levels and estimates development and production
costs. - NAFCOM is applicable to various types of missions
(manned spacecraft, unmanned spacecraft, and
launch vehicles). - There are two versions of the model a government
version that is restricted and a contractor
releasable version.
3NAFCOM Evolution
- Since 1990, eight versions of NAFCOM have been
developed and distributed across NASA and other
government agencies.
NASCOM Books
NASCOM Automated DB
NASCOM Ver. 3.0
NASCOM Ver. 4.0
NAFCOM Ver. 2002
NAFCOM Ver. 2004
NAFCOM96
NAFCOM99
1990
1992
1994
1996
1998
2000
2002
2004
- NASCOM database in hardcopy
- Estimators hand-entered data into spreadsheets
- Database contained 65 data points
- Allowed online searching copying of data
- Cost estimates developed in spreadsheets with
CERs created by individuals - Database contained 70 data points
- Fully functional cost model with user defined WBS
and data access - CERs built within NASCOM using 1st Pound method
- Database contained 91 data points
- Combined NASA and Air Force data
- Enhanced search and filtering of data
- Standardized WBS elements created
- Database contained 102 data points
- First non-weight based CERs for five subsystems
(Complexity Generators) - Gov and contractor versions distributed
- Database contained 114 data points
- Total re-write of all NAFCOM program code
- Complexity Generators for all subsystems
- Major user interface improvements
- Database contains 122 data points
- Complexity Generators reworked
- Fully integrated version of SOCM
- Probabilistic Cost Risk Analysis module
4NAFCOM Probabilistic Cost AnalysisRationale
- NAFCOM provided a point estimate as opposed to a
probabilistic range estimate. - The addition of a probabilistic cost risk
analysis module - provides the ability to identify, analyze, and
quantify risk/uncertainty - Expands NAFCOM from a static analysis tool to one
that provides probabilistic analysis of uncertain
values
5NAFCOM Probabilistic Cost AnalysisDevelopment
- SAIC was tasked with developing a Probabilistic
Cost Analysis capability for NAFCOM. - Experts in the risk field, including
representatives from MCR, Aerospace, NASA, IPAO,
and Mitre participated in the methodology
development. - A meeting was held at NASA HQ to discuss issues
including sampling approaches, incorporation of
correlation and model implementation. - Dr. Steve Book has been working directly with
SAIC and NASA to ensure that the best possible
approaches are considered for integration into
NAFCOM and to consult on the methodology
implementation.
6NAFCOM Probabilistic Cost AnalysisModel Selection
- Choice between Analytic approach (FRISK) or Monte
Carlo-based sampling approach - Analytic Approach
- Pros Simple for summing WBS elements,
instantaneous results, simple development - Cons Difficult to apply risk to algebras beyond
summation (i.e., inflation, phasing) - Monte Carlo Approach
- Pros Allows probability analysis on different
algebras - Cons Difficult to develop or interface w/COTS,
time intensive to run (I.e., 5,000 to 10,000
samples) - Decision Use Analytic Approach
7NAFCOM Probabilistic Cost AnalysisImplementation
At Estimating Level
B
Best-Estimate Cost
0
SEE
Distribution of Estimating Error
L
H
B
Cost Implications of Complexity Drivers
(Technical, Management and New Design)
8NAFCOM Probabilistic Cost AnalysisImplementation
At Summation Levels
WBS Element 1
Total Cost
- Approximate by lognormal distribution
- Compute analytically
- Mean
- Mode (Most likely)
- Confidence percentiles
WBS Element 2
.
.
.
9NAFCOM Probabilistic Cost AnalysisModel Interface
- Users can elect to run NAFCOM in Risk On or Risk
Off mode via toolbar
10NAFCOM Probabilistic Cost AnalysisModel Interface
- NAFCOMs Risk On Mode provides capability to
define triangular distributions for all cost
driver inputs and for cost thruputs - These distributions are combined with CER errors
to produce distributions for nonrecurring and
recurring cost for each subsystem
11NAFCOM Probabilistic Cost AnalysisModel Interface
- The distributions for each subsystem are then
aggregated, taking correlation and System
Integration into account - Both intra- and inter-subsystem correlation is
taken into account - Users can elect to modify the default correlation
matrix - The default correlation matrix was developed
based on Dr. Steve Books recommendation of 0.2
correlation across all elements (knee of the
curve)
12NAFCOM Probabilistic Cost AnalysisModel Interface
- The final result is uncertainty distributions for
DDTE, Flight Unit, Production and Total Cost. - Result data includes summary statistics,
probability densities, and cumulative
distributions for each major estimating element
(i.e. stage, bus)
13NAFCOM Probabilistic Cost AnalysisSummary
- Uses well-documented analytical method to
calculate risk - User can input low, most likely, and high values
for all NAFCOM complexity generator and
conventional CER inputs - Incorporates both technical and estimating
uncertainties - Incorporates correlation between subsystem costs
- Results displayed to the user are summary
statistics, probability densities, and cumulative
distributions for DDTE, Flight Unit, Production,
and Total Costs for each major estimating element
(i.e. stage, bus, etc.) - User can select either the Normal distribution or
the Lognormal distribution to approximate the
final results
14NAFCOM Probabilistic Cost AnalysisCurrent Status
- Model is in final test and debug phase
- A beta version of the model was delivered to MSFC
last week - Roll out of model to cost estimating community is
planned for early April - NAFCOM can be downloaded from the NAFCOM home
page at http//nafcom.saic.com - winns_at_saic.com
- 256-971-7275 256-705-8539
15NAFCOM Probabilistic Cost AnalysisPlanned
Enhancements
- Initial delivery of NAFCOM probabilistic cost
analysis module does not address engine CERs, the
allocation of risk dollars, or an assessment of
the accuracy of the sampling method used. - The the cost risk analysis capability will be
expanded to include all engine CERs including
liquid, combined cycle, and air breathing
estimating algorithms allowing the user to
perform risk analysis on all current NAFCOM CER
types. - The ability to determine which elements have the
most cost risk associated with them and to
allocate risk dollars back to those WBS elements
will be added. - User will be allowed to select a point estimate
(most likely,mean, median, mode) and a level of
confidence above the point estimate to determine
risk dollars. - There is no unique optimal way of allocating
the risk dollars, but NAFCOM will implement a
technique acceptable to the NASA estimating
community that takes into account inter-element
correlations and other features of the estimate. - The accuracy of the analytical versus Monte Carlo
approaches to sampling will be studied to
determine the most cost effective capability. - Determine if there is any value added in using
the Monte Carlo sampling approach over the
statistical method. - Document findings to support/refute the need to
use one approach over the other.