16.885 Aircraft Systems Engineering
Cost Analysis
Karen Willcox MIT Aerospace Computational Design Laboratory
AEROSPACE COMPUTATIONAL DESIGN LABORATORY Outline
• Lifecycle cost • Operating cost • Development cost • Manufacturing cost • Revenue • Valuation techniques
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Lifecycle Cost
Lifecycle : Design - Manufacture - Operation - Disposal Lifecycle cost : Total cost of program over lifecycle
85% of Total LCC is locked in by the end of preliminary design.
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Lifecycle Cost
100 95% 85% 80
65% 60
40 Disposal Operation and support Impact on LCC (%) 20 design Detailed design Manufacturing and acquisition Preliminary design, system integration Conceptual 0 Time
(From Roskam, Figure 2.3)
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Operating Cost
¾ Airplane Related Operating Cost SROC (AROC) 10% CROC ¾ Passenger Related Operating Cost 2% (PROC) PROC Cargo Related Operating Cost ¾ 18% AROC (CROC) 70% ¾ Systems Related Operating Cost (SROC)
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Airplane Related Operating Costs
CAPITAL COSTS: CASH AIRPLANE RELATED Financing OPERATING COSTS: Insurance Capital Crew Depreciation Fuel Costs CAROC Maintenance 40% 60% Landing Ground Handling GPE Depreciation GPE Maintenance Control & Communications
CAROC is only 60% - ownership costs are significant!
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 CAROC Breakdown per Trip
Other Control & Comm 3%
Ground Fuel is roughly 20% of Handling 9% 60% of 70% of Total 7% Operating Cost Landing Crew i.e. 8% 6% 40%
Maintenance 15% Fuel 20%
typical data for a large commercial jet
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Non-Recurring Cost
Cost incurred one time only: Engineering - airframe design/analysis - configuration control
- systems engineering Engineering Tooling - design of tools and fixtures - fabrication of tools and fixtures
Other Tooling - development support - flight testing Other
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Recurring Cost
Cost incurred per unit: Labor - fabrication - assembly - integration Labor Material to manufacture -raw material - purchased outside production - purchased equipment Production support Material -QA - production tooling support - engineering support Support
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Learning Curve As more units are made, the recurring cost per unit decreases. This is the learning curve effect. e.g. Fabrication is done more quickly, less material is wasted. n Y x Y0 x
Yx = number of hours to produce unit x n = log b/log 2 b = learning curve factor (~80-100%)
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Learning Curve
1 Every time 0.8 production 0.6 doubles, cost is reduced 0.4 b=0.9 0.55 by a factor
Cost of unit unit of Cost 0.2 of 0.9
0
0 1020 3040 50 Unit number
Typical LC slopes: Fab 90%, Assembly 75%, Material 98%
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Elements of a Cost Model
Engineering Data Build Schedule & Performance 120 lanes lanes 80
40
0
Number of p of Number 2000 2010 2020 2030 Year Recurring Cost Component Breakdown Plane
Wing Fuselage COST
Winglet Skin Ribs MODEL
RC/lb NRC/lb Weight Subparts/lb Non-Recurring Learning Curve NRC Distribution Cost 1 2 0.8
0.4 1 Cost of unit
0 NRC ($B) 0 1020 30 40 50 Unit number 0 2003 2007 2011 2015 Year
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Typical Cost Modeling
1. Take empirical data from past programs. 2. Perform regression to get variation with selected parameters, e.g. cost vs. weight. 3. Apply “judgment factors” for your case. e.g. configuration factors, complexity factors, composite factors. There is widespread belief that aircraft manufacturers do not know what it actually costs to turn out their current products.
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Cost Modeling • Aircraft is broken down into modules – Inner wing, outer wing, … – Modules are classified by type • Wing, Empennage, Fuselage, … • Cost per pound specified for each module type – Calibrated from existing cost models – Modified by other factors • Learning effects • Commonality effects • Assembly & Integration: a separate “module” • 2 cost categories: development & manufacturing Production run: a collection of modules
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Cost Modeling
Plane
Centerbody Wing Landing Propulsion Systems Payloads Final Gear Assembly At this level, the degree of detail can Winglet Outer Inner … range from e.g. “wing” Wing Wing to “rivet”.
Identifier Weight Area RC per Subparts NRC per NRC time pound per pound pound distribution
Labor Material & Support Tooling Engineering Other Equipment
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Development Cost Data
Baseline QA Baseline Dev. Labs Baseline QA QA
Baseline Dev. Dev. Labs Labs
Baseline Tool Tool Fab. Fab Baseline ToolFab.Tool Fab. Baseline Tool Tool Design Design
Baseline M.E. M.E.
Baseline Engr. Engr.. Baseline Tool Design
Baseline M.E.
Baseline Engr. non-dimensional labor hours
non-dimensional time
Boeing data for large commercial jet
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Development Cost Model • Cashflow profiles based on beta curve: t c (t ) Kt D 1 (1 t )E 1 • Typical development time ~6 years • Learning effects captured – span, cost 0.06
Support 0.05 Tool Fab
0.04 Tool Design
ME
0.02 Engineering (from Markish) normalized cost 0.01
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19normalized 20 21 22 23 24 25 26 27 28 29 time 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Development Cost Model
Payloads 8% Wing 20% Systems 17%
Empennage 9% Installed Engines 8% Landing Gear 1%
Fuselage 37% Representative non-recurring cost breakdown by parts for large commercial jet (from Markish).
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Development Cost Data
For your reference: $/lb assembled from public domain weight and total cost estimates plus representative NRC breakdown by aircraft part (see Markish). Tool Engineering ME Design Tool Fab Support Totals
40.0% 10.0% 10.5% 34.8% 4.7% 100.0%
Wing $7,093 $1,773 $1,862 $6,171 $833 $17,731
Empennage $20,862 $5,216 $5,476 $18,150 $2,451 $52,156
Fuselage $12,837 $3,209 $3,370 $11,169 $1,508 $32,093
Landing Gear $999 $250 $262 $869 $117 $2,499
Installed Engines $3,477 $869 $913 $3,025 $408 $8,691
Systems $13,723 $3,431 $3,602 $11,939 $1,612 $34,307
Payloads $4,305 $1,076 $1,130 $3,746 $506 $10,763
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Manufacturing Cost Model
• Aircraft built Æ modules required • Modules database – Records quantities, marginal costs – Apply learning curve effect by module, not by aircraft
Labor Materials Support 85% 95% 95%
time AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Manufacturing Cost Model
Final Assembly 6% Wing Payloads 27% 11% Systems 6%
Installed Engines 9% Empennage Landing Gear 10% 3%
Fuselage 28% Representative recurring cost breakdown by parts for large commercial jet (from Markish).
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Manufacturing Cost Data
For your reference: $/lb values assembled from public domain weight and total cost estimates plus representative RC breakdown by aircraft part (see Markish).
Labor Materials Other Total
Wing $609 $204 $88 $900
Empennage $1,614 $484 $233 $2,331
Fuselage $679 $190 $98 $967
Landing Gear $107 $98 $16 $221
Installed Engines $248 $91 $36 $374
Systems $315 $91 $46 $452
Payloads $405 $100 $59 $564
Final Assembly $58 $4 $3 $65
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 NASA Cost Models
Online cost models available at http://www.jsc.nasa.gov/bu2/airframe.html
e.g. Airframe Cost Model - simple model for estimating the development and production costs of aircraft airframes - based on military jet data - correlation with empty weight, max. speed, number of flight test vehicles, and production quantity
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Revenue Model
Revenue model must predict market price and demand quantity.
180 90 160 80 140 767-200ER 70 A300 767-300ER 120 A310 777-200 60 A330 777-300 100 50 A340 747-400 747 80 MD-11 40 A300 767 60 30 A330 777 price ($M) 20 40 A340 deliveries MD-11 10 20 0 0 1980 1985 1990 1995 2000 1985 1990 1995 2000 year year
Historical wide body data from Markish. No correlation found between price and quantity.
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Aircraft Pricing
Cost-Based Pricing Market-Based Pricing
Performance Cost + Profit = Price Operating Cost Market Competition Value Passenger Appeal Personal aircraft Commercial transport Business jets? Military aircraft
Source: Schaufele
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Commercial Aircraft Pricing
• Total Airplane Related Operating Costs are fairly CAROC constant. • Aircraft price must balance CAROC.
PRICE Total AROC COST/WEIGHT (Capital costs) TRADE-OFF
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Business Jet Empirical Data
Figure A7 in Roskam:
-1 AMP1989 = log {0.6570 + 1.4133 log WTO}
AMP1989 is predicted airplane market price in 1989 dollars
Take-off weight: 10,000 lb < WTO < 60,000 lb
BUT Gulfstream GIV and 737 BJ versions do not fit the linear trend.
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Commercial Jet Empirical Data Figure A9 in Roskam:
-1 AMP1989 = log {3.3191+ 0.8043 log WTO}
AMP1989 is predicted airplane market price in 1989 dollars
Take-off weight: 60,000 lb < WTO < 1,000,000 lb
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Military Aircraft Empirical Data Figure A10 in Roskam:
-1 AMP1989 = log {2.3341+ 1.0586 log WTO}
AMP1989 is predicted airplane market price in 1989 dollars
Take-off weight: 2,500 lb < WTO < 1,000,000 lb
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Revenue Model: Price
• Assumption: market price based on 1. Range 2. Payload 3. Cash Airplane-Related Operating Cost (CAROC)
• Regression model: D P k1(Seats) k2 (Range) f (CAROC )
• Note that speed does not appear. No significant statistical relationship between price and speed was found in available data.
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Revenue Model: Price
Narrow bodies: P 0.735(Seats )1.91 0.427(Range ) f ( CAROC)
Narrow bodies 80
70
60
50
40
30
20 y=x Airbus
Estimated price ($M) 10 Boeing
Estimated price ($M) 0 0 1020304050607080 ActualActual price price ($M)($M)
Model from Markish, price data from Aircraft Value News, The Airline Monitor, 2001. AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Revenue Model: Price
Wide bodies:
P 0.508(Seats)2.76 0.697(Range ) f ( CAROC )
Wide bodies 160
140
120
100
80
60 y=x 40 Airbus
Estimated price ($M) 20 Boeing
0 Estimated price ($M) 0 20 40 60 80 100 120 140 160 ActualActual price price ($M)
Model from Markish, price data from Aircraft Value News, The Airline Monitor, 2001. AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Revenue Model: Quantity
• Demand forecasts – 3 sources: Airbus; Boeing; Airline Monitor – Expected deliveries over 20 years – Arranged by airplane seat category • Given a new aircraft design: – Assign to a 4000 Airbus seat category 3500 Airli ne Mon i tor – Assume a 3000 Boeing market share 2500
– Demand forecast Æ 2000
20-year production Quantity 1500 Quantity
potential 1000
500
0 100 125 150 175+ 200 250 300 350 400 500+ SeatSeat Category Category
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Revenue Model: Dynamics
• Expected aircraft deliveries: forecasted • Actual deliveries: unpredictable • Observe historical trends: growth rate, volatility
400
350
300
250
200 narrow 0.0228x units units y = 94.571e 150 R2 = 0.3724
100
50 y = 52.776e 0.0167x wide R2 = 0.4092 0 1 4 7 10 13 25 28 31 34 37 40 16 19 22 6-mo. period
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Valuation Techniques
The top 5 investor questions: • How much will I need to invest? • How much will I get back? • When will I get my money back? • How much is this going to cost me? • How are you handling risk & uncertainty?
Investment Criteria • Net present value • Payback • Discounted payback • Internal rate of return
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Net Present Value (NPV)
• Measure of present value of various cash flows in different periods in the future • Cash flow in any given period discounted by the value of a dollar today at that point in the future – “Time is money” – A dollar tomorrow is worth less today since if properly invested, a dollar today would be worth more tomorrow • Rate at which future cash flows are discounted is determined by the “discount rate” or “hurdle rate” – Discount rate is equal to the amount of interest the investor could earn in a single time period (usually a year) if s/he were to invest in a “safer” investment
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Calculating NPV
• Forecast the cash flows of the project over Its economic life – Treat investments as negative cash flow • Determine the appropriate opportunity cost of capital (i.e. determine the discount rate r) • Use opportunity cost of capital to discount the future cash flow of the project • Sum the discounted cash flows to get the net present value (NPV)
C1 C2 CT NPV C0 2 ! T 1 r 1 r 1 r
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 NPV example
Period Discount Factor Cash Flow Present Value
0 1 -150,000 -150,000 1 0.935 -100,000 -93,500 2 0.873 +300000 +261,000
Discount rate = 7% NPV = $18,400
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Discount Rate
• One of the problems with NPV: what discount rate should we use? • The discount rate is often used to reflect the risk associated with a project: the riskier the project, use a higher discount rate • Typical discount rates for commercial aircraft programs: 12-20% • The higher the discount rate, the more it does not matter what you do in the future...
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Payback Period
• How long it takes before entire initial investment is recovered through revenue • Insensitive to time value of money, i.e. no discounting • Gives equal weight to cash flows before cut-off date & no weight to cash flows after cut-off date • Cannot distinguish between projects with different NPV • Difficult to decide on appropriate cut-off date
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Discounted payback
• Payback criterion modified to account for the time value of money – Cash flows before cut-off date are discounted • Surmounts objection that equal weight is given to all flows before cut-off date • Cash flows after cut-off date still not given any weight
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Internal rate of return (IRR)
• Investment criterion is “rate of return must be greater than the opportunity cost of capital” • Internal rate of return is equal to the discount rate for which the NPV is equal to zero
C1 C2 CT NPV C0 2 ! T 0 1 IRR 1 IRR 1 IRR • IRR solution is not unique – Multiple rates of return for same project • IRR doesn’t always correlate with NPV – NPV does not always decrease as discount rate increases
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Decision Tree Analysis
• NPV analysis with different future scenarios • Weighted by probability of event occurring
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Dynamic Programming
• A way of including uncertainty and flexibility in the program valuation • Key features: • Certain aspects of the system may be uncertain, e.g. the demand quantity for a given aircraft = UNCERTAINTY • In reality, the decision-maker (aircraft manufacturer) has the ability to make decisions in real-time according to how the uncertainty evolves = FLEXIBILITY
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Dynamic Programming: Problem Formulation • The firm: – Portfolio of designs – Sequential development phases – Decision making • The market: – Sale price is steady – Quantity demanded is unpredictable – Units built = units demanded • Problem objective: – Which aircraft to design? – Which aircraft to produce? – When?
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Dynamic Programming: Problem Elements
1. State variables st
2. Control variables ut 3. Randomness
4. Profit function
5. Dynamics 1 ½ • Solution: Ft ( st ) max®S t (st , ut ) Et >Ft 1( st 1 ) @¾ ut ¯ 1 r ¿ • Solve iteratively.
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Dynamic Programming: Operating Modes How to model decision making?
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Example: BWB
• Blended-Wing-Body (BWB): – Proposed new jet transport concept
• 250-seat, long range Image taken from NASA's web • Part of a larger family sharing site: http://www.nasa.gov. common centerbody bays, wings, ...
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Example: BWB Simulation Run
16 120 15 14 mode 13 demand 100 12 11 80 10 9 8 60 7 6 5 40 per year 4 3 20 2 operating mode 1 demanded quantity 0 0 0 2 4 6 8 1012 14 16 18 20 22 24 26 28 30 time (years)
3,000,000
2,000,000
1,000,000
0 0 2 4 6 8 1012 14 1618 2022 24 26 28 30 -1,000,000
-2,000,000
-3,000,000 cash flow ($K) -4,000,000 time (years)
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 Example: BWB Importance of Flexibility
25 dynami c programmi ng
20 Net Present Value
15
10
5
0 program value ($B) ($B) value program 11 18 28 44 69108 171 270 -5 3 5 7
-10 initial annual demand forecast At baseline of 28 aircraft, DP value is $2.26B versus NPV value of $325M AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885 References
Airbus Global Market Forecast, 2000-2019. Appendix G, Detailed passenger fleet results, p. 74. Aircraft Value News Aviation Newsletter www.aviationtoday.com/catalog.htm The Airline Monitor, ESG Aviation Services. Boeing Current Market Outlook, 2000. Appendix B, p. 42. Jane's All the World's Aircraft. London : Sampson Low, Marston & Co., 2001. Markish, J. Valuation Techniques for Commercial Aircraft Program Design, S.M. Thesis, MIT, June 2002. Markish, J. and Willcox, K. “Valuation Techniques for Commercial Aircraft Program Design,” AIAA Paper 2002-5612, presented at 9th Multidisciplinary Analysis and Optimization Symposium, Atlanta, GA, September 2002. Markish, J. and Willcox, K., “A Value-Based Approach for Commercial Aircraft Conceptual Design,” in Proceedings of the ICAS2002 Congress, Toronto, September 2002. NASA Cost Estimating website, http://www.jsc.nasa.gov/bu2/airframe.html Roskam, J., Airplane Design Part VIII, 1990. Raymer, D., Aircraft Design: A Conceptual Approach, 3rd edition, 1999. Schaufele, R., The Elements of Aircraft Preliminary Design, 2000.
AEROSPACE COMPUTATIONAL DESIGN LABORATORY 9/19/2004 16.885