Professional Documents
Culture Documents
Economic Design Optimization of Foundations: Yu Wang, M.ASCE and Fred H. Kulhawy, Dist.M.ASCE
Economic Design Optimization of Foundations: Yu Wang, M.ASCE and Fred H. Kulhawy, Dist.M.ASCE
Abstract: A geotechnical foundation design should address at least three basic requirements: ultimate limit state 共ULS兲, serviceability
limit state 共SLS兲, and economics. Most conventional design approaches focus on ULS and/or SLS optimization, with economics being
evaluated afterwards. As an alternative, this paper develops a design approach that explicitly considers the construction economics and
results in a foundation that has the minimum construction cost. This design approach is expressed as an optimization process, in which the
objective is to minimize construction cost, with the design parameters and design requirements as the optimization variables and
constraints, respectively. This design approach is illustrated using a spread footing example. Because construction costs vary by locale, the
economically optimized designs differ regionally. Sensitivity studies on soil properties and design requirements show that, for typical
spread footing designs in cohesionless soils, Young’s modulus 共E兲 and the effective friction angle 共⬘兲 are the key parameters. A
quantitative assessment illustrates the importance of soil property variability on cost. It is also found that, for typical spread footing
designs, a relatively stringent ULS requirement generally ensures fulfillment of the SLS requirement.
DOI: 10.1061/共ASCE兲1090-0241共2008兲134:8共1097兲
CE Database subject headings: Economic factors; Cost estimates; Optimization; Spread foundations; Limit states; Serviceability.
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
sions and materials used. The performance requirements include, There are several different levels 共or types兲 of cost estimates,
but are not limited to, safety and reliability or, more specifically, each of which has a different purpose and place in the construc-
the ULS and SLS requirements, which are given by multiple per- tion estimating process. Means 共1990兲 divides cost estimates into
formance indices, such as factor of safety, reliability index, and four different levels: 共1兲 order of magnitude estimates; 共2兲 square
allowable deformation. Successful designs must satisfy all perfor- foot and cubic foot estimates 共or SI equivalents兲; 共3兲 assembly
mance criteria. estimates; and 共4兲 unit price estimates. The order of magnitude
Every product 共or geo-structure兲 has a cost, which varies as the estimate could be described loosely as an educated guess and can
design parameters 共e.g., materials and dimensions兲 change. The be completed when only minimal information is available. One
economics requirement of foundation design is to satisfy the man- important purpose of this type of estimate is to provide informa-
datory performance requirements with minimum construction tion to an owner or client so that a budget-level decision can be
cost, and it can be addressed explicitly in design by optimization, made to proceed with the project.
in which minimizing the construction cost 共Z兲 of the foundation The square foot and cubic foot estimate is an advanced pre-
is the optimization objective. Note that Z is a function of liminary estimate based on historical data. It is effective when
the design parameters 共x1 , x2 , x3 , . . . , xn兲 and other constants features of the proposed project are known, but not yet designed
共c1 , c2 , c3 , . . . , cn兲 that define the unit costs of construction mate- 共i.e., type of building or facility and proposed number of square
rials, labor, and so on. The design parameters 共x1 , x2 , x3 , . . . , xn兲 feet or cubic feet兲.
are considered to be optimization variables, and the perfor- The assembly estimate is an intermediate one prepared during
mance requirements, measured by performance indices the early stage of project design. An assembly refers to a collec-
共PI1 , PI2 , PI3 , . . . , PIn兲, are treated as optimization constraints. tion of the individual tasks required to construct a composite
The optimization is to obtain a set of design parameters work item. For example, consider a spread footing assembly,
共x1 , x2 , x3 , . . . , xn兲 that result in a set of performance values which usually includes five tasks: excavation, formwork, rein-
共PI1 , PI2 , PI3 , . . . , PIn兲 that are equal to or exceed the required forcement, concrete, and compacted backfill. In these estimates,
performance indices 共PI1r , PI2r , PI3r , . . . , PInr兲 and also give the an assembly is treated as a cost unit, and the total cost is the
minimum Z. The optimization process can be expressed in a product of the assembly unit cost and number of assemblies in the
standard nonlinear programming format 共Castillo et al. 2002兲 as project.
follows: The fourth level is the unit price estimate, in which the work is
Objective divided into the smallest possible work items, and a unit price is
estimated for each item. The unit price then is multiplied by the
Minimize cost Z = f共x1,x2,x3, . . . ,xn,c1,c2,c3, . . . ,cn兲
quantity required to obtain the cost for the items, and the esti-
Constraints mated cost is the sum of the costs for all items. The result is the
most accurate and detailed estimate. All decisions regarding the
Performance Index 1, PI1共x1,x2,x3, . . . ,xn兲, is equal to or exceeds PI1r construction materials and methods must have been made before
Performance Index 2, PI2共x1,x2,x3, . . . ,xn兲, is equal to or exceeds PI2r the unit price estimates, because detailed working drawings,
Performance Index 3, PI3共x1,x2,x3, . . . ,xn兲, is equal to or exceeds PI3r specifications, details on construction equipment, and productiv-
] ity data are needed for determining the quantities of materials,
Performance Index n, PIn共x1,x2,x3, . . . ,xn兲, is equal to or exceeds PInr
equipment, and labor.
Clearly, each estimate level requires different amounts of time
Practical Constraints, xil 艋 xi 艋 xiu,i = 1,2,3, . . . ,n
to complete and has different levels of accuracy. As the estimates
in which xil and xiu⫽lower and upper bounds, respectively, for progress from the least to the most accurate 共i.e., from 1 to 4兲, the
design parameter xi and define the practical ranges of the design time needed for a modest building project might increase signifi-
parameters. For example, consider a design parameter that defines cantly from several minutes to several weeks. The required input
the dimensions of a spread footing. These dimensions must be data also vary significantly with estimate level. The estimator’s
greater than zero, and therefore xil must be greater than zero. experience on similar previous projects might be the major input
The economically optimized design then can be obtained by data for the order of magnitude estimates, as opposed to the other
solving the optimization problem described above. There is sub- three levels of estimates, which more frequently use published,
stantial literature available on optimization, and well-established annually updated, unit cost data, such as Means Square Foot
methodologies are applied widely for solving optimizations in Costs 共Means 2006c兲, Means Assemblies Cost Data 共Means
various disciplines. Optimization software and toolkits also have 2006a兲, and Means Building Construction Cost Data 共Means
been developed, such as the “Solver” function in Microsoft Excel. 2006b兲. Note that unit cost data are location dependent and vary
Background information on optimization is beyond the scope of from region to region. Costs for the same materials vary from
this paper, and the reader is referred to Pedregal 共2003兲 and region to region, or even from city to city, and labor rates and
Castillo et al. 共2002兲 for details. productivity also vary geographically. Therefore, it is not surpris-
ing to find that the unit costs in major cities 共e.g., New York City兲
are much higher than those in rural areas.
Cost Estimates For the best optimization, it would be desirable to adopt the
level of cost estimates that are the most accurate, i.e., the unit
A key component in the optimization is calculating Z 共cost esti- price estimates. However, it is difficult to apply these estimates
mate兲, which consists of two basic tasks: 共1兲 determining material directly because they require detailed working drawings, specifi-
quantities and construction labor required 共i.e., quantity takeoff兲; cations, details on construction equipment, and productivity data,
and 共2兲 estimating reasonable costs for them 共Means 1990兲. Es- none of which obviously are available during design. In this
sentials of cost estimates are reviewed in this section, while the paper, a simplified unit price estimate is developed that uses the
reader is referred to Bledsoe 共1992兲, Means 共1990兲, and Waier unit prices developed for assembly cost data 共Means 2006a兲.
共1993兲 for details. Table 1 shows an assembly example for constructing a
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
Table 1. Calculation of Spread Footing Assembly Unit Price 共after Means 2006a兲
Cost per assembly
United States dollars 共USD兲
Spread footing Unit price
共0.91 m ⫻ 0.91 m ⫻ 0.30 m deep兲 Quantity Units Material Labor Total 共USD兲
Excavation 0.451 m3 N/Aa 11.35 11.35 25.16
Formwork 1.115 m2 7.80 50.16 57.96 51.97
Reinforcement 5.443 kg 5.61 6.15 11.76 2.16
Concrete 0.253 m3 37.62 6.27 43.89 173.96
Compacted backfill 0.199 m3 N/Aa 0.79 0.79 3.97
a
Total N/A N/Aa 51.03 74.72 125.75 N/Aa
Note: Cost data represent United States national average for 2007.
a
N/A⫽not applicable.
0.91 m ⫻ 0.91 m ⫻ 0.30 m deep spread footing. The first column ULS and SLS Designs
contains the five individual tasks cited previously. The second and
The ULS and SLS requirements for the spread footing are given
third columns show the quantities and associated units for these
by the required factor of safety 共FSr兲 against soil failure and the
tasks. The United States national average costs for materials and
allowable settlement 共␦r兲, respectively. In this example, FSr = 3
labor are given in the fourth and fifth columns, respectively, and
and ␦r = 25 mm are used as the design criteria. In addition, the
the sixth column shows the total costs. The unit prices are calcu-
foundation depth 共D兲 should be greater than a minimum depth
lated by dividing the sixth column by the second. The quantities
共D1兲 = 0.5 m to prevent frost damage and should be limited to a
of the five individual tasks are functions of the design parameters
maximum depth 共Du兲 = 2.0 m to minimize disturbance to adjacent
共e.g., dimensions兲 and can be estimated. The following example
structures.
illustrates this process for a spread footing.
The ultimate bearing capacity 共qult兲 of a footing on a cohesion-
less, horizontal, sand deposit can be calculated from Eq. 共1兲, for
general shear and a concentrically applied vertical load 共Vesic
Spread Footing Design 1975兲
Spreading footing design is arguably among the simplest in geo- qult = 0.5B␥⬘N␥␥s␥d + q⬘Nqqsqd 共1兲
technical engineering, and therefore it is used to illustrate the
in which ␥⬘⫽effective unit weight of soil below footing;
optimization framework.
q⬘⫽effective overburden stress at foundation level; N␥ and
Nq⫽bearing capacity factors, and ␥s, ␥d, qs, qd⫽shape and
Design Example depth factors, as given below
Fig. 1 illustrates a spread footing in dry sand with a unit weight N␥ ⬇ 2共Nq + 1兲tan ⬘ 共2兲
共␥兲 = 18.5 kN/ m3 and an effective friction angle 共⬘兲 = 35°. The
sand has a Young’s modulus 共E兲 = 50 MPa and a Poisson’s ratio
共兲 = 0.3. The groundwater is at considerable depth, so the spread Nq = 共e tan ⬘兲tan2共45 + ⬘/2兲 共3兲
footing is not affected by groundwater. The footing is designed to
support a column that transfers a vertical load 共F兲 = 3,000 kN to ␥s = 1 − 0.4共B/L兲 共4兲
the footing. Three key design parameters for the spread footing
are the depth 共D兲, width 共B兲, and length 共L兲, as shown in Fig. 1. ␥d = 1 共5兲
qult
FS = 共8兲
F/BL
For this example, the settlement 共␦兲 is calculated using the
following elastic solution 共Poulos and Davis 1974兲
F共1 − 2兲
␦= 共9兲
zE冑BL
in which z⫽shape factor given in Fig. 2 共Whitman and Richart
1967兲. Although this example uses an elastic solution to calculate
Fig. 1. Spread footing design example ␦, the economically optimized design framework is independent
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
the concrete volume 共Qc兲 in cost estimates. Qr can be expressed
as
Qr = kQc 共13兲
in which k⫽proportionality coefficient that is taken as
29.67 kg/ m3 共Means 1990兲 herein. Compacted backfill is mea-
sured by its volume, and this quantity 共Qb兲 is the difference be-
tween Qe and Qc, expressed as
Qb = Qe − Qc 共14兲
Then the construction cost 共Z兲 of the spread footing can be
estimated as
Z = Q ec e + Q f c f + Q cc c + Q rc r + Q bc b 共15兲
in which ce, c f , cc, cr, and cb⫽unit prices for excavation, form-
Fig. 2. Shape factor z
work, concrete, reinforcement, and compacted backfill, respec-
tively, as summarized in the last column of Table 1.
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
Table 2. Comparison of Three Spread Footing Designs
Width, B Length, L Depth, D Cost Difference
Design option 共m兲 共m兲 共m兲 共USD兲 共%兲
Optimized 1.86 2.30 1.38 1086 N/Aa
Example 1 2.60 2.60 0.60 1421 31
Example 2 1.70 2.40 2.00 1143 5
a
N/A⫽not applicable.
It should be noted that the construction cost also is affected cludes the unit prices for Beijing, the capital of China, which are
by other factors, such as constructability, construction schedule, extracted from construction cost data in China 共DLS 2006; CMC
local contractor’s preference, and availability of standard or local 1995兲.
materials 共e.g., formwork sizes兲. Although these factors are The comparable, economically optimized designs of spread
not included in this example, the optimization framework can be footings for the United States average, New York City, and
extended readily to incorporate these factors as optimization Beijing are summarized in Table 5. The optimized design for New
constraints. York City is a square footing with B = L = 2.10 m and D = 1.25 m.
This design is slightly different from the one for the United States
national average, with shallower depth 共1.25 m versus 1.38 m兲,
Sensitivity Study slightly larger footprint 共2.10⫻ 2.10= 4.41 m2 versus 1.86⫻ 2.30
= 4.28 m2兲, and a construction cost that is about 1.4 times higher
than the United States national average cost. The optimized de-
Effect of Locale
sign for Beijing has B = 1.18 m, L = 3.15 m, and D = 2.00 m, which
Construction costs vary from region to region, or even from city is significantly different from the one for the United States na-
to city, as materials costs, labor rates, and productivity vary geo- tional average. The depth 共D = 2.00 m兲 is the upper bound value
graphically. Because the construction costs are considered explic- and is significantly larger than the depth 共D = 1.38 m兲 for the
itly in the design, the location of the designed geo-structures and United States national average. On the other hand, the footprint
the geographic variation of construction costs can result in differ- 共1.18⫻ 3.15= 3.72 m2兲 is smaller than that for the United States
ent optimized designs. national average 共1.86⫻ 2.30= 4.28 m2兲.
In cost estimates, the location effect is addressed using the city All designs result in FS= 3 and ␦ = 25 mm and reflect different
cost index. This index for the United States national average is strategies in fulfilling the design requirements. The footing might
taken as 100, and the indices for 316 major cities throughout the be either relatively shallow with a relatively large size or rela-
United States and Canada are given by Means 共2006a兲. For ex- tively deep with a relatively small size. The variation of founda-
ample, consider the cost indices for New York City, which are tion depth leads to different quantities of excavation and backfill,
110.5 and 164.1 for materials and labor for foundation work, the construction costs of which are dominated by labor costs, as
respectively. The materials and labor unit prices in New York City shown in Tables 1 and 3. In contrast, the variation of footing size
are adjusted by multiplying the ratio of the city cost indices results in different quantities of formwork, concrete, and rein-
共1.105 and 1.641兲 to the national average unit prices, such as the forcement, the construction costs of which are much more af-
data in Table 1. Table 3 summarizes the calculation of spread fected by materials costs, as shown in Tables 1 and 3. Since
footing unit prices in New York City. The fourth and fifth col- construction cost is a product of quantity and unit price, the eco-
umns in Table 3 are obtained from the comparable columns in nomically optimized deign, which explicitly considers the con-
Table 1 and the city cost index ratios 共1.105 and 1.641兲. The sixth struction cost, depends on the unit price of each individual work
column shows the total costs, and the seventh column summarizes task and, particularly, the relative ratios between them. As the
the unit prices. Table 4 summarizes the unit prices of spread foot- relative ratio changes, the economically optimized design might
ings for the United States national average and New York City, also change. Examination of Table 4 shows that, although all unit
which shows higher prices in New York City. Table 4 also in- prices in Beijing are substantially smaller than those for the
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
Table 4. Summary of Unit Prices of Spread Footings for Three Locations
Cost per unit
共USD兲
United States
national New York City, Beijing,
Work task Units average United States Chinaa
Excavation m3 25.16 41.30 1.38
Formwork m2 51.97 81.55 4.00
Reinforcement kg 2.16 2.99 0.54
Concrete m3 173.96 204.98 47.50
Compacted backfill m3 3.97 6.51 0.22
a
Extracted from DLS 共2006兲 and CMC 共1995兲, assuming that the
currency exchange rate between USD and Chinese currency, RMB is 1:8.
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
Fig. 5. Effects of design requirements FSr and ␦r
Fig. 4. Effect of Young’s modulus E
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
relatively inexpensive labor for excavation and backfill, resulting Appendix. Optimization Model in Microsoft Excel
in a relatively deep footing with a relatively small size. In con-
trast, the designed footing in New York City is relatively shallow Fig. 6 shows the optimization model for spread footing design in
with a relatively large size. a Microsoft Excel spreadsheet. The spreadsheet is divided into
A sensitivity study was performed: 共1兲 to identify the key soil three zones: 共1兲 an input data zone from Rows 1 to 7; 共2兲 a design
properties that most affect the design and construction costs; and parameter and calculation zone from Rows 9 to 18; and 共3兲 a cost
共2兲 to quantitatively assess the benefit of improving soil property estimate zone starting from Row 20.
characterization. It was found that E and ⬘ are the key param- The input data consist of soil properties, design performance
eters that significantly affect the design of spread footings in co- requirements, practical constraints, and unit price data, in this
hensionless soils and therefore deserve specific attention in site case taken from the last column of Table 1. The design parameter
investigation. However, if E is relatively large 共i.e., greater than and calculation zone include three design parameters 共D, B, and
60 MPa in this example兲, then it is no longer a key parameter. In L兲 and two assumed constants of footing design and calculations
addition, sensitivity studies show that, when ␦r is relatively small, for both ULS and SLS designs. Eqs. 共1兲–共8兲 are implemented in
it controls the final design. As ␦r increases, its effect decreases Cells E10–E18 for calculations of bearing capacity factors, shape
and becomes negligible when ␦r is larger than some threshold factors, depth factors, bearing capacity, and FS. The settlement is
value, which decreases as FSr increases. This result suggests that, calculated using Eq. 共9兲 in Cell H12 and Eq. 共16兲 in Cell H11.
for spread footing design, a relatively stringent ULS requirement The cost estimates contain the quantity takeoffs, which are calcu-
共i.e., large FSr兲 generally ensures fulfillment of the SLS require- lated in Cells B22–B26 using Eqs. 共10兲–共14兲, and the total cost
ment 共i.e., ␦r兲. estimate in Cell D22 with Eq. 共15兲.
Finally, it should be noted that the economically optimized The optimization model then is set up using the Excel toolkit
design framework is independent of the calculation models used Solver, which can be launched by going to the drop-down menu
for qult and ␦ and the types of structures under design. of “Tools” and selecting “Solver.” As shown in the lower right
corner of Fig. 6, the Solver parameters window appears that per-
mits definition of the optimization objective, variables, and con-
Acknowledgments straints. The optimization objective is defined by assigning Cell
D22 as the target cell and checking the “Min” button. Cells B10–
The work described in this paper was supported by grants from B12 are the optimization variables. Both design requirements and
City University of Hong Kong 共Project Nos. 7002072 and practical constraints are treated as constraints and implemented as
7200070兲. Cell B10艋 2, Cell B10艌 0.5, Cell B11艌 0, Cell B12艌 0, Cell
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
E18艌 Cell H3, and Cell H12艋 Cell H4. The optimization is ex- ␥⬘ ⫽ effective unit weight of soil below
ecuted by clicking the “Solve” button in the upper right corner, footing;
and the minimum construction cost and the corresponding design ␦ ⫽ calculated settlement;
parameters and calculated FS and ␦ are generated. For the ex- ␦r ⫽ required allowable settlement;
ample shown in Fig. 6, the minimum construction cost is USD ␥d , qd ⫽ depth factors in bearing capacity
1,086. equation;
Note that the Solver toolkit may not have been installed by ␥s , qs ⫽ shape factors in bearing capacity
default, but it can be activated by going to the drop-down menu of equation;
“Tools,” selecting “Add-Ins . . . ,” and checking the “Solver Add- ⫽ Poisson’s ratio; and
In” checkbox. ⬘ ⫽ effective friction angle.
Notation References
The following symbols are used in this paper: Bledsoe, J. D. 共1992兲. Successful estimating methods, R. S. Means Co.,
Kingston, Mass.
B ⫽ footing width; Castillo, E., Conejo, A. J., Pedregal, P., Garcia, R., and Alguacil, N.
B0 ⫽ over-excavation distance along width 共2002兲. Building and solving mathematical programming models in
direction; engineering and science, Wiley, New York.
cb , cc , ce , c f , cr ⫽ unit prices of compacted backfill, Coduto, D. P. 共1994兲. Foundation design: Principles and practices,
concrete, excavation, formwork, and Prentice-Hall, Englewood Cliffs, N.J.
reinforcement, respectively; Construction Ministry of China 共CMC兲. 共1995兲. Standard construction
c1 , c2 , c3 , . . . , cn ⫽ unit costs of construction materials, cost data, No. GJD-101-95, China Planning Press, Beijing.
labor, and so on; Davis, Langdon, and Seah Ltd. 共DLS兲. 共2006兲. Construction cost hand-
D ⫽ footing depth; book: China and Hong Kong 2006, Hong Kong.
Dl ⫽ lower bound of footing depth; Federal Emergency Management Agency 共FEMA兲. 共1996兲. Performance
Du ⫽ upper bound of footing depth; base seismic design of buildings: An action plan for future studies,
E ⫽ Young’s modulus; FEMA-283, Federal Emergency Management Agency, Washington,
D.C.
F ⫽ vertical load;
Ghobarah, A. 共2001兲. “Performance-based design in earthquake engineer-
FS ⫽ factor of safety;
ing: State of development.” Eng. Struct., 23共8兲, 878–884.
FSr ⫽ Required factor of safety;
Lambe, T. W., and Whitman, R. V. 共1969兲. Soil mechanics, Wiley, New
H ⫽ footing height;
York.
k ⫽ proportionality coefficient in Eq. 共13兲; Means. 共1990兲. Means estimating handbook, Kingston, Mass.
L ⫽ footing length; Means. 共2006a兲. 2007 RS Means assemblies cost data, Kingston, Mass.
L0 ⫽ overexcavation distance along length Means. 共2006b兲. 2007 RS Means building construction cost data, King-
direction; ston, Mass.
N␥ , Nq ⫽ bearing capacity factors; Means. 共2006c兲. 2007 RS Means square foot costs, Kingston, Mass.
PI1 , PI2 , PI3 , . . . , PIn Pedregal, P. 共2003兲. Introduction to optimization, Springer, New York.
⫽ performance indexes; Phoon, K. K., and Kulhawy, F. H. 共1999兲. “Characterization of geotech-
PI1r , PI2r , PI3r , . . . , PInr nical variability.” Can. Geotech. J., 36共4兲, 612–624.
⫽ required performance indexes; Poulos, H. G., and Davis, E. H. 共1974兲. Elastic solutions for soil and rock
Qb , Qc , Qe , Q f , Qr ⫽ quantities of compacted backfill, mechanics, Wiley, New York.
concrete, excavation, formwork, Sowers, G. F. 共1979兲. Introductory soil mechanics and foundations:
and reinforcement, respectively; Geotechnical engineering, 4th Ed., Macmillan, New York.
q⬘ ⫽ effective overburden stress at foundation Vesic, A. S. 共1975兲. “Bearing capacity of shallow foundations.” Founda-
level; tion engineering handbook, H. Winterkorn and H. Y. Fang, eds.,
qult ⫽ ultimate bearing capacity; Chap. 3, Van Nostrand Reinhold, New York.
xil ⫽ lower bound of design parameter xi; Waier, P. R. 共1993兲. Means unit price estimating methods: Standards and
procedures for using unit price cost data, 2nd Ed., R. S. Means Co.,
xiu ⫽ upper bound of design parameter xi;
Kingston, Mass.
x1 , x2 , x3 , . . . , xn ⫽ design parameters; Wellington, A. M. 共1887兲. Economic theory of the location of railways,
Z ⫽ construction cost; Wiley, New York.
z ⫽ shape factor for settlement calculation Whitman, R. V., and Richart, F. E. 共1967兲. “Design procedures for dy-
given by Fig. 2; namically loaded foundations.” J. Soil Mech. and Found. Div.,
␥ ⫽ soil unit weight; 93共SM6兲, 169–193.
Downloaded 11 Jun 2012 to 180.211.192.67. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org