07

.7-z
 Unclassif led
SECURITY CLASSIFICATION OF THIS PAGE

77
REPORT DOCUMENTATION PAGEOn 7

Ma.REPORT SECURITY CLASSIFICATION lb. RESTRICTIVE MARKINGS

Unclassified
2a. SECURITY CLASSIFICATION AUTHORITY 3. DISTRIBUTION/AVAILABILITY OF REPORT
Approved for public release; distribution
2b. DECLASSIFICATION I DOWNGRADING SCHEDULE unlimited.

4. PERFORMING ORGANIZATION REPORT NUMBER(S) S. MONITORING ORGANIZATION REPORT NUMBER(S)

Miscellaneous Paper EL-88-2

Ga. NAME OF PERFORMING ORGANIZATION 6b. OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATION
USAEWES (if appitcable)
Environmental Laboratory CEWES-EE-R _
6c. AnDRESS (City, State, and ZIP Code) 7b. ADDRESS (City. State, and ZIP Code)

PO Box 631
Vicksburg, MS 39180-0631
$a. NAME OF FUNDING/SPONSORING IBb. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER
ORGANIZATION (If applicable)
US Army CERL Reimbrrsable Order No. CIAO-87-109
Sc. ADDRESS(City, State, and ZIP Code) 10. SOURCE OF FUNDING NUMBERS
PROGRAM PROJECT TASK LWORK UNIT
PO Box 4005 ELEMENT NO. NNO. NO.
Champaign, IL 61820-1305 rc CCESSION NO.

11. TITLE (Incld, Security Oassification)

Predicting Internal Roughness in Water Mains

J2. PERSONAL AUTHOR(S)
Walski, Thomas M.; Sharp, Wayne W.; Shields, F. Douglas, Jr.
13a. TYPE OF REPORT 113b. TIME COVERED 14. DATE OF REPORT (Year, Month, Day) 1S. PAGE COUNT
Final report FROM TO February 1988 21
15. SUPPLEMENTARY NOTATION
Available from National Technical Information Service, 5285 Port Royal Road, Springfield,
S~ VA 22161.
17. COSATI'CODES 18. SUBJECT TERMS (Continue on reverse if necesary and identify by block number)
FIELD GROUP SUB-GROUP Water mains Carrying capacity Pipe networks
C-factors Pipes Pipe flow
Pipe cleaning Water distribution systems Roughness
19, ABSTRACT (Continue on reverue if necessary and identify by block number)

'-A method is presented for predicting the Hazen-Williams C-factor for unlined metal
water mains as a function of pipe age. The method has two steps: (a) finding the
growth/rate of internal roughness, alpha, for the water main using either historical
C-factor data or water quality data; and (b) using predictive equations for an estimate of
a future C-factor.
The predictive equations presentedheteieen were derived using linear regression of
.some 319 data points from seven utilities as well as values from the 1920 text entitled
Hydraulic Tables, by G. S. Williams and A. Hazen.
The regression
confidence interval of-+15 and for
eqtations a coeffici~nt function of pipl
C-factor vsof a determination of have
(r ) age 0.87.-.a 95-percent

20. DISTRIBUTION /AVAILABIILITY OF ABSTRACT I . ABSTRACT SECURITY' CLAJSSIFICATION"

r'UNCLASSIFIED/UNLIMITED 0 SAME AS RPT.

22a. NAME OF RESPONSIBLE INDIVIDUAL
0 DTIC USF.RS Unclassified
I22b TELEPHONE (include"AreaConJ, T 22. OFFICE SYMBOL

DD Form 1473, JUN 86 Previouseditions are obsolete. SECURITY CLASSIFICATION OF THIS PAGE

Unclassified
 PREFACE

This report was prepared by the Environmental Laboratory (EL), US Army

Engineer Waterways Experiment Station (WES), in fulfillment of Reimbursable
Order No. CIAO-87-109 from the US Army Construction Engineering Research Labo-
ratory (CERL). Dr. Ashok Kumar of the CERL was the primary point of contact
for the work. The report was written and prepared by Dr. Thomas M. Walski,
Mr. Wayne W. Sharp, and Dr. F. Douglas Shields, Jr., of the Water Resources
Engineering Group (WREG), Environmental Engineering Division (EED), EL.
Ms. Cheryl Lloyd of the WREG proofread the report.
The work was accomplished under the direct supervision of Dr. Shields
and Dr. Paul R. Schroeder, who served as Acting Chief, WREG, and Dr. R. L.
Montgomery, Chief, EED, and under the general supervision of Dr. John Keeley,
Assistant Chief, EL, and Dr. John Harrison, Chief, EL.
COL Dwayne G. Lee, CE, was the Commander and Director of WES.
Dr. Robert W. Whalin was Technical Director.

This report should be cited as follows:

Walski, Thomas M., Sharp, Wayne W., and Shields, F. Douglas, Jr.
1988. "Predicting Internal Roughness in Water Mains," Miscellaneous
Paper EL-88-2, US Army Engineer Waterways Experiment Station,
Vicksburg, Miss.

K Accesioni For
NTIS CRA&I
OTIC TAB F]

SJ r +, .... .. . ,... .. . . . ... . .. . . . . . .

D!! 01:C 'jO

cop
CYS
1OETEO
 CONTENTS

PREFACE................. ,............ ,..................................... I

PART I: INTRODUCTION ................................................ 3

Estimation of C t qu .O .. .o o...
.3........ .... • ............. 3

PART II:
Use ofSOLUTION OF HEAD LOSS
Hazen-Williams EQUATIONS
Equation FOR C
for Rough .........................
Flov......................
Use of Darcy-Weisbach Equation for Head Los...........
5
5
6
I
Relationship Between Roughness andCF to.. .. ...... .. 6
PART III: PREDICTING GROWTH OF ROUGHNESS IN PPS...........
Historical Data onRoughness Growth~ae........... . 10
Formulas Based onWater Quality....................... 11
Examination of Linear Hyohss................... 13

PART IV: APPLICATION OF PRDICTIVE EQUATIONS ........................... 16I

Example ro em........*....... ....... ... 18

Rate of Change of C aco. ..... .. .. ... ....... 19
S16
PART V: SUMMARY AND CONCLUSIONS ... . ....... ..... . .**...* o***.***..**9 23

REFERENCES.2............... 24

2
 PREDICTING INTERNAL ROUGHNESS IN WATER MAINS

PART I: INTRODUCTION

Background

1. Some measurement of the internal roughness of pipes is an important

parameter entering into virtually all calculations involving the sizing and
analysis of water distribution systems. In water utility practice in tae
United States, the Hazen-Williams C-factor is the most commonly used parameter
to represent the carrying capacity (and internal roughness) of water mains.
2. In modern cement mortar lined pipes and plastic pipes, the pipe
roughness changes very slowly over the life of a pipe (with the exception of
waters with significant scaling potential or poor removal of aluminum hydrox-
ide flocs). However, there exist many miles of old, bare metal water mains in
which the pipe roughness has changed and continues to change significantly
over time.__ _ _ _ _a

Estimation of C-Factors

3. Knowledge of the pipe roughness (or C-factor) of in-plac:e water

mains is critical for pipe sizing calculations, and there are several ways of
obtaining this information. The first isto use "typical literature" values
for pipe roughness. As will be shown later in this paper, the only thing
"typical" about such values is that they vary widely from water System to
water system.
4. A second method is to determine the internal roughness of existing 0
mains by calculating the value of pipe roughness that will make a computer
model of the pipe network appear to be calibrated over a wide range of condi-
tions. Some examples of such a procedure are given by Walski (1986) and
Ormsbee and Wood (1986). These methods essentially give an "effective" 0
C-factor because many small pipes and minor losses are eliminated when the
distribution system is simplified (skeletonized) to make it more workable.

3
 IMUWIIWWW M MW MWU
j W UWWW UVUWWW
" jW%#PWWWU am

5. A third method is to actually measure the pipe roughness in the

field by conducting head loss tests (Walski 1984). This method is quite
expensive but gives the most credible results.
6. None of the above methods for determining C-factors, however, pro-
vide a means for extrapolating pipe roughness into the future. This paper
will review pipe reighness data collected from a wide variety of sources over
many years and provide a theoretically based procedure for predicting future
C-factor values. The procedure will require information on historical

C-factors or water quality.

41
 PART II: SOLUTION OF HEAD LOSS EQUATIONS FOR C

Use of Hazen-Williams Equation for Rough Flow

7. The Hazen-Williams equation, which is traditionally used in the

United States for calculating head loss and flow, is, strictly speaking, only
applicable to smooth flow (i.e., flow in which roughness elements in the pipe
wall do not penetrate through the laminar bourdary layer). The equation
becomes somewhat inaccurete for transition flow and even less accurate for
completely rough flow which is usually the case in old unlined water mains.
The equation can be written as:

V - 0.55 C D0.63 S054)

where
V - velocity, ft/sec
C - Hazen-Williams, C-factor
D - diameter, ft
S - hydraulic gradient, ft/ft
8. The Hazen-Williams equation, however, is still used even in rough
flow because the error in predicting head loss is not significant except for
long pipes with very high velocity. In mcst water distribution system prob-
lems, the Hazen-Williams equation is sufficiently accurate.
9. In rough flow the C-factor becomes a function of the velocity and
can be corrected by the equation (Walski 1984) given below

C -C (Vo/V)O0081 (2)
0 0

vhere
C - C-factor at velocity V

C 0 - measured C-factor at velocity V0

V - actual velocity, L/T
V - velocity at which C-factor measured, L/T
The velocity at which the C-factor is measured can vary widely, but a typical
value of roughly 3 ft/sec (0.9 m/sec) can be used for most calculations.

5I
I -- _MM
Equation 2 shows that the error associated with assuming a constant but
typical velocity is minor.

Use of Darcy-Weisbach Equation for Head Loss.

10. The more theoretically correct equation for head loss is the Darcy-

Weisbach equation which can be solved for velocity as

V- (3)

where
2
g - acceleration due to gravity, ft/sec
f - Darcy-Weisbach friction factor, dimensionless
The friction factor can be calculated based on properties of the flow and the
wall roughness.

Relationship Between Roughness and C-Factor

11. Because of the fact that the C-factor is a function of velocity,

the initial discussion below will center on pipe roughness (e), which repre-
sents the height of equivalent sand grain roughness. When C-factors are dis-
cussed, they will be in regard to some known velocity (say 3 ft/sec
(0.9 m/sec)).
12. By solving both the Hazen-Williams and Darcy-Weisbach Equations for
head loss and eliminating terms, the C-factor can be related to the Darcy-
Weisbach friction factcr by

C- 17.25 (4)
f 0.54(VD) 0.081

The friction factor, f , can be related to internal pipe roughness over a

wide range of flows by the Colebrnok-White transition equation (Colebrook
1939). However, this equation cannot be solved explicitly for f . Since
this paper concerns pipes with significant tuberculation and scale, a simpli-
fled form of the Colebrook-White equation, attributed to von Karman and

6I

ýIý
NZN%'NYKhIM2VM
Prandtl (von Karman 1930), that is applicable only to fully rough flow, can be
used. This equation can be solved explicitly for f as given below.

f (
(1.14 - 2 log (e/D)]- (5)

where e is the roughness height (same units as D). (All logarithms in this

paper are base 10.) Equation 5 can be substituted into Equation 4 to give a
formula relating roughness height to C-factor for a velocity of 3 ft/ssc

(0.9 m/sec) ae shown below (neglecting the D0.081 term)

C - [14.6 - 25.6 log (e/D)] 1.08 (6)

Equation 6 is used below to make conversions between C-factors and roughness

height.
13. Equation 6 appears as a nearly straight line on a semi-log plot
over a typical range of values for e/D . The exponent, 1.08, makes the equa-
tion somewhat difficult to use and can be eliminated by determining the
straight line that most closely approximates Equation 6 over the range of
C-factors of interest. This line can be given by

C - 18.0 - 37.2 log (e/D) (7)

Equation 7 will be used in the methods presented later to predict C-factors as

a function of age of pipes and water quality.

14. While the complete Colebrook-White transition formula cannot be

solved explicitly for f over a wide range of Reynolds numbers and pipe
roughness, other formulas, most notably one by Swamee and Jain (1976) can be
solved explicitly for friction factor for smooth, transition and rough flow.
The formula of Swamee and Jain is given below

f 0.25 (8)

where N is the Reynolds number.

71

na VAVaA .A t.AnfLrAr6PMW WUND.%JWWVU%ýJ ~~V~

VV WUWWV .WJ YU~ VVUW% 'W.M V
wqy%,
 S
15. Equation 8 can be inserted into Equation 4, and using a velocity of
3 ft/sec (0.9 m/sec) and eliminating the Reynolds number term which is negli-
gible for rough flow, the C-factor can be given by

C- 33.3 [log (e/3.7D)2 ]54 M

The above equation can be simplified by realizing that the log term (even
though it i negative) is always raised to the second power so that the abso-M

lute value of the log can be taken and the exponents can be combined to give

C - 33.3 1 log (e/3.TD) 1 1.08 (10)

Equat'ons 6, 7, 9, and 10 give virtually the same value for C-factor for any
value of pipe roughness greater than 0.001 ft (0.305 mm) and are reasonably
close for values to 0.0001 ft (0.0305 mm).

8
 PART III: PREDICTING GROWTH OF ROUGHNESS IN PIPES

Early Observations

16. The growth of pipe roughness with time has been observed for some
time. Williams and Hazen (1920) in their original hydraulic tables give sug-
gested values for C-factor for each diameter pipe for various ages. They do
not quantitatively describe the effect of water qualIty on pipe roughness.
Nevertheless, their suggested values have been used as virtually "gospel" by
engineers since they were first published.
17. The work of von Karman (1930), Prandtl (1933), and Nikuradse (1932)
showed that friction factors in the Darcy-Weipbach equation could be related
to the pipe wall roughness and the Reynolds number of the flow. Colebrook
(1938) developed an analytical relationship to describe this.
18. More importantly for this paper, Colebrook and White (1937)
addreseed the problem of changing pipe roughness with time. Using their data
and data prepared by a Committee of the New England Water Works Association
(1935), Colebrook and White advanced the hypothesis that pipe roughness grows
roughly linearly with time and that the rate of growth depends most highly on
the pH of the water. They assumed that roughness was virtually zero when the
pipe was new. A slightly modified form of their equation to predict roughness
is given in Streeter (1971) as

e - e + at (11)
0

whereI
e - absolute roughness
height, L
e - roughness height at time zero, L
a growth rate in roughness height, L/T
t time, T
19. The hypothesis that e is a linear function of t is examined in
detail below. Equation 11 can now be substituted into Equations 6, 7, 9, and
10 to give a method of prodicting C-factors as a function of time. This is
done for Equations 7 and 10 below:

-- - - - -- - - -
 1W¶.IWInnm
iqr~i WIW¶WI
owMlnWI WInmowmww I anW
I WIW
109m wrlI WJIUpa1WnMWI

Colebrook-White

C - 18.0 - 37.2 log (X) (12)

Swamne-Ja in

(13)
(0.27X) I 1.08
+ at)D .C - 33.3 1 log
where X - (e
20. Equations 12 and 13 contain two constants, a° , the initial rough-
ness, and a , the roughness growth rate, which must be determined empiri-
cally. The initial roughness depends on the pipe material, but a typical
value of 0.0006 ft (0.18 m) gives reasonable results for new metal pipes and
is reasonably close to values reported by Lamont (1981). Methods to predict
the roughness growth rate must take into account water quality. Data on the
relationship of roughness growth to water quality are prevented below.

Historical Data on Roughness Growth Rate

21. Colebrook and White (1937) reported values of roughness growth rate
ranging from 0.000018 ft/yr (0.066 uun/yr) to 0.00017 ft/yr (0.63 im/yr).
Based on data they obtained from the New England Water Works Association
(1935), they proposed the following equation to predict the growth rate,
although they admitted rLat it was "little better than a guess."

a - 0.0833 exp (1.9 - 0.5 pH) (14)

22.
Another significant finding of Colebrook and White was that the
loss of carrying capacity was due much more to the increase in pipe roughness
rather than the decrease in pipe diameter due to the space occupied by the
roughness elements.
23. The roughness growth rates that could be back calculated from the

tables of Williams and Hazen (1920) varied considerably depending on the diam-
eter but generally corresponded to 0.002 ft/yr (0.6 mm/yr). Their values were
based on a fairly limited number of observations and they noted, "In general
it may be stated that rather large deviations from the indicated rates of

10
reduction ih carrying capacity are found in individual cases, but that, in the
experience of the authors, the variations are about as often in one direction
as in the other." They recommend field head loss tests wheu a high degree of
accuracy is required. 0

Formulas Based on Water Quality

24. As the years went by, more investigators reported data on growth of
internal pipe roughness versus time. Lamont (1981) presented the most thor-
ough compendium of data on pipe roughness and gave the most rational guidance
on the effects of water quality on roughness growth. His approach consisted
of identifying four "trends" in roughness growth and determined a linear
growth rate for each trend as given in Table 1. Lamont used the Langelier
Index, which is essentially an indicator of the saturation of the water with
respect to calcium narbonate, to characterize water quality. Numerical values
for growth rate as compared with Langelier Index are shown in Table 1 and can

be predicted by Equation 13 below

a - 10-(4.08+0.38 LI) for LI < 0 (15)

where LI is the Langelier Index. Note that Equation 15 is developed only

for Langmller Indexes less than zero (i.e. corrosive water) and sW)ould not be

Table 1
Roughness Growth Rate for Varying Water Quality

Growt__ rate
_.

mm/yr Langeliet
Index
Trend Name (ft/yr)

1 Slight attack 0.025 0.0

0.000082

2Moderate attack 0.076 -1.3

0.00025

3 Apprecia;.e attack 0.25 -2.6

0.00082

4 Severe attack 0.76 -3.9

0.0025
extrapolated to waters with scale forming tendencies (i.e. positive Langelier
Index) or waters with other postprecipitation problems. The range of values
for growth rate of roughness reported by Lamont is similar to that reported by
Colebrook and White.
25. While it is possible to predict the roughness growth rate for
waters with a negative Langelier Index, there is considerably less data avail-
able to predict this rate for positive Langelier Index values. Natural waters
usually do not have a positive Langelier Index, and therefore, the problems
with supersaturated water usually only exist in hot water lines or for utili-
ties which soften water but do not recarbonate. Larson and Sollo (1967)
describe a city In which 2 in. (50.8 mm) deposits of magnesium hydroxide were
found in 8 in. (203.2 mm) pipes and another in which the C-factor dropped from
135 to 100 in 15 years due to calcium carbonate and magnesium hydroxide depo-
sits. They were able to reproduce this phenomenon in stainless steel pipes in
the laboratory.
26. Another problem is post precipitation of aluminum hydroxide in dis-

tribution systems where the water is treated using alum. No quantitative

model is available to relate loss of carrying capacity to pH or aluminum con-
centration, but loss of carrying capacity has been observed by Costello
(1982), Emery (1980) and Larson and Sollo (1967). Emery reported that
C-factors dropped from 140 to a range of 109 to 100 over a 10- to 23-year
period due to such deposits. This corresponds to a roughness growth rate of
0.00025 ft/yr (0.076 mm/yr).
27. A committee on Loss of Carrying Capacity of Water Mains of the
California Section American Water Works Association (1962) presented data on
changes in C-factor for the Los Angeles water system. They presented the
results of over 70 tests for pipes with diameters between 4 and 16 in. (101.6
and 406.4 mm). Roughness growth rates were calculated for each diameter and
were found to range from 0.0020 to 0.0011 ft/yr (0.61 to 0.34 mm/yr).
28. Hudson (1966, 1966, 1973) reported the results of numerous head
loss tests in nine cities. While the data were reported in terms of C-factor
instead of roughness, and there was a considerable amount of scatter among the
data, it was still possible to back calculate roughness growth rates, using
Equation 6 for each city as shown in Table 2.
29. The above data show that most surface waters tend to be corrosive.
However, the limestone well water in San Antonio proved to be much less

12
 Table 2
Roughness Growth Rates Calculated from Hudson's Data

Growth Rate
mm/yr
City (ft/yr) Description of Water Quality
Atlanta 0.61 Soft river water
0.0020

Fort Worth 0.55 Carrying capacity lost even in concrete

0.0018 and cement mortar lined mains

Denver 0.18 Mountain reservoirs

0.0006

New Orleans 0.16 River water

0.00052

Cincinnati 0.14 River water

0.00043

Chicago 0.10 Lake water, alum treated

(south) 0.00033

St. Paul 0.045 Unsoftened surface water

0.00015

Chicago 0.027 Lake water, no alum

(north) 0.0009

San Antonio 0.015 Wells in limestone

0.00005

corrosive because it is near saturation with respect to calcium carbonate.

Unfortunately, data on the pH, calcium, and alkalinity of the waters in the
above table were not available so a quantitative method to predict growth rate
as a function of water quality could not be developed. The higher growth rate

for the Chicago south treatment plant is attributed to postprecipitation of

aluminum hydroxide.

Examination of Linear Hypothesis

30. If sequential measurements of C-factor spanning several years are

available for the pipe sizes and water source of interest, and if water qual-
ity has been fairly unchanged over this period of record, then the engineer

13
can predict C-factors as a function of pipe age using Equation 12 or 13. The
likely error inherent in such an approach is probably acceptable if the hypo-
thesis that pipe roughness is a linear function of pipe age.
31. The predictive equations for C (Equations 12 and 13) are based on
the hypothesis that pipe roughness, e , is a linear function of pipe age,
t . This hypothesis was examined using all available data. The data set used
consisted of sequential measurements of C for pipes of known diameter and
age. Some 319 data points from seven utilities plus values from Williams and
Hazen (1920) and average values from Lamont (1981) were used. Fi-st, all
C-factors were converted to pipe roughness values (e) using Equation 16.
Then, for each utility, pipe roughness growth rates (a) were determined for
each pipe diameter by linear regression. The values of e computed with
Equation 16 were input to the regression as dependent variable values, and
pipe ages (t) were the dependent variable values. The a values determined
from regression were then combined with published or measured values of e0
and the known values of D ani t to compute X .

log (X)
t is
32. The amount of scatter present in the entire data set, and thus the
validity of the linear hypothesis, was examined by plotting
as shown in Figure 1. Figure I indicates that a linear function of
a good predictor of pipe roughness and therefore Equations 12 or 13 may
C versus
'
be used to predict C-factors as a function of pipe age. The 95-percent confi-
dence interval about the regression line is also shown in Figure 1. Predicted
C-factors are within an interval of ±15 of observed values with 95-percent
confidence. The coefficient of determination for both equations is 0.87,
which means that the equations describe 87 percent of the variation in C.
This is especially good considering the fact that the data were collected by a
wide variety of investigators for a wide range of conditions. For example,
different investigators use different methods for determining minor head

losses.
33. In addition to the fact that all of the observations fall reason-
ably close to both predicted curvres, the two equations (12 and 13) give nearly
equal results over the range 'iý C normally encountered. Therefore, either
can be used, depending on the preference of the engineer. Examples of their
use follow.

14

k 1ý11
11
 %**

00

0 4i
0 cc Q 0
0 a
05.1

0 do

04.

t2-

*1

*- S, VI
 PART IV: APPLICATION OF PREDICTIVE EQUATIONS

Procedure

34. To apply the predictive equation to a water system, the engineer

needs to first determine the growth rate of roughness.
This can be done in
two ways. First, if
C-factor and pipe age data are available, the engineer
can plot pipe age versus roughness (e). Roughness (e) can be calculated from

C-factor by solving Equation 6 or 10 for e :

e-D (io0[ (18-C) /37.2] (16)

e - 3.7 (D 10- (C *)(17)

The slope of the e versus t plot will be the growth rate which can be used
for predicting future C-factors. The intercept with the vertical axis gives
an indication of initial pipe roughness. However, since new pipe C-factors
usually are not measured, since there is some error in C-factor measurements,
and since the predictive equations are not very accurate for roughness
approaching zero, the estimates of initial roughness may appear unreasonable
in some cases (e.g. negative). In such cases, the engineer should use typical
initial roughness values of approximately 0.0006 ft (0.18 mm). In any case,
the value of initial roughness will usually have very little effect on
C-factors after the first few years of a pipe's life.
35. if the transformed data do not fall on a straight line with reason-
able error, the engineer must consider why this has occurred. The usual
explanation is that water quality has changed during the life of some of the
pipes. The engineer must then try to determine the aging rate given the cur-
rent water quality since that water quality will influence future changes in

C-factor.
36. If historical C-factor values are not available or water quality is
expected to change in the future, then the engineer must predict roughness
growth based on water quality. Equation 15 should be used for this purpose
since the Langelier Index is based on more than simply pH, as given in Equa-
tior 14. Corrosion inhibitors will also have an effect on growth rate.

16
 37. Once the roughness growth rate is known, the C-factor in a given
year can be predicted using Equations 12 and 13, knowing the age of the pipe
for that year. This is an acceptable procedure if the precise age of instal-
lation of each pipe segment is known. Determining the year laid for each pipe
segment can be a fairly tedious process and is one that is currently not
required in most pipe network models. An alternative procedure is to extrapo-
late the C-factor given the current value. The current value can be deter-
mined from field tests or the value needed to ca?.ibrate a pipe network model
as discussed earlier.
38. To modify the predictive equations so that the year laid ueed not
be known, it is necessary to first substitute for pipe age in the equations as
shown below

t - TO

0
(18) I
where
T - Year of interest (e.g. 1995)
T - year of installation
It is possible to reorganize Equations 12 and 13 to eliminate T , which
results in an equation to predict C-factor in any year T , given the C-factor
in some other year, T

a(T-Tl) + D 10 (18-C 1 )/37.2

C = 18.0- 37.2 log D (19)

6 /25.7
[r0.c92 1.08
0. 27a (T- T1 ) +DD 10-C 1 (20)
C - 33.3 Ilog I

where
C1 M C-factor in year
T1 (known)
TI = year in which C-factor is known
T - year in which C-factor is predicted
39. The above equations can be easily incorporated into computer models
of pipe networks so that an engineer can simulate future system behavior.

173
 Example Problems

40. Two example problems to illustrate the predictive equations are

presented below. The first covers the case in which no C-factor data are

available, but the year the pipe was laid is known. The second covers the
case in which there are some historical C-factor data.
Example 1

U
41. It is necessary to predict the C-factors in 1995 for the pipes

described in the first three columns of Table 3. The pipes are unlined cast
iron and Langelier Index for the water is -1.5. Determine C-factors for sev-
eral individual pipes plus plots of C-factor versus age for 6, 12, 24 and
.c in. (152.4, 304.8, 635.0 and 914.4 mm) pipes.

Table 3

Data for
T Example

Diameter
Pipe (ft) Year Laid C in 1995

1 0.5 1925 68
2 0.5 1905 64

3 0.5 1940 72
4 2.0 1925 91

42. The rate of aging can be estimated from Equation 15 as

a = 10-(4.080.38(-1.5)] - 0.00031 ft/yr (0.095 mm/yr)

The age of each pipe in 1995 can be determined by subtracting the year laid S
from 1995. Then, using an initial
tion 12 can be used to generate the C-factor in
roughness of 0.0006 ft (0.18 nu), Equa-
1995 as I
C - 18.0 - 37.2 flog [0.0006 + 0.00031 (1995 - To)]/DJ

A plot of C-factors versus age for several diameter pipes is shown in Fig-
ure 2. Note that C-factor decreases more rapidly in small diametcr pipes.

18

0N
 150

120

0
(90/,
LL

60 EXAMPLEI

30 1I I
0 20 40 60 80
AGE. YR

Figure 2. C-factor versus age

Example 2
43. In this problem, several C-factor tests were conducted and the
results are shown in the first three columns of Table 4. The data were then

transformed as shown in Equation 17, and the res .:ts are shown in Figure 3.

The roughness growth rate was found to be 0.0019 ft/yr (0.58 mm/yr) using lin-
ear regression. Equation 20 was then used to predict the C-factors in 2005:

10_(926 /57_11..0
C - 33.3 1 log [0.0103 + D 1 0 -(CD1 25.7) 111

Rate of Change of C-Factor

44. While pipe roughness increases linearly with time, C-factors ini-
tially decrease rapidly and then change more gradually as shown in Figure 2.

19
 0.12

0.10 1
0.08

UI 0.06

0.04

0.0,0.001f9
00

o I I I , I
0 10 20 t, YR 30 40 50 •

Figure 3. Growth rate versus time for Example 2

120 200
5'
LS
 Table 4
Data for Example 2

Diameter C-factor Age in e C factor

pip (ft) in 1985 1985 (ft) in 2005
1 2.0 s0 20 0.041 69
2 1.0 61 35 0.065 54

I
3 1.0 55 45 0.095 50
4 0.5 48 45 0.073 41
5 0.5 42 50 0.107 37

This explains why results of pipe cleaning jobs, when not accompanied by
cement mortar lining or a change in water quality, are usually short lived as
described by Dutting (1968) and Frank and Perkins (1955). To quantify this
effect, it is possible to take the derivative of Equation 12 with respect to
time. This gives the rate of change in C-factor with time.

dC -16.1 (-
d- [o/a) + t)] 2)

The absolute value of the right side of Equation 21 is plotted in Figure 4 for
roughness growth rates, a , of 0.001 and 0.0001 ft/yr (0.305 and
0.0305 m/yr). Figure 4 shows tb'.t the C-factor drops off initially,
decreases rapidly and then changes very slowly.

I2

ii21
 H
IS
10

nr8

U-

t•.
0

C, Sa
- 0.00? FTYR
4
2

U.
0

2 0.0001 FT1YR _

0 10 20 30 40 50
AGE. YR

Figure 4. Rate of changes of C-factor versus time

22
 PART V: SUMHARY AND CONCLUSIONS

45. The equations presented in this paper provide an easy-to-use but

accurate method for predicting C-factors in unlined metal pipes. The data
substantiate the claim of Colebrook and White (1937) that roughness height in
such pipes grovs roughly linearly over time.
46. The equations presented depend most heavily on the parameter, a
that describes the rate of growth of roughness over time and depends on the
quality of the water with respect to corrosion or precipitation. This param-
eter should be determined based on historical data for the given water system.
If this information is not available, it can be estimated based on the
Langelier Index or extrapolated from other utilities with similar water
quality.
47. Since the C-factor is a logarithmic function of time, the C-factor
tends to decrease most rapidly for new or recently cleaned pipes. However,
after a few decades the drop in C-factor with time is very slow.
23

23j
 REFERENCES

California
Water Section
Maine," J. AWWA, Committee,
AIIAVol. 54, No. 1962(Oct).
10, p. 1293. "Loss of Currying Capacity in

Colebrook, C. F. 1939. "Turbulent Flow in Pipes with Particular Reference to

the Transition Re-ion between the Smooth and Rough Pipe Laws," J. Inst. Civil
Eng. London, Vol. 11, p. 133.
Colebrook, C. F., and C. M. White. 1937. "The Reduction of Cerrying Capacity
of Pipes with Age," J. Inst. Civil Eng. London, Vol. 10, V. 99.
Costello, J. J. 1982. "Postprecipitation in Distribution Lines Following
Water Treatment, National Conf Proceedings, p. 1245.
Dutting, R. F. 1968. "Cleaning and Lining Water Mains in Place," presented
at New England Water Works Association Training School.
Emery, P. M. 1980. "Some Aspects of the Performance of Cement Mortar Lined
Water Mains," AWA Natl. Conf. Proceedings, p. 249.
Frank, J. A., and Perkins, A. G. 1955(Apr). "Water Main Cleaning and Lin-
ing," Water and Sewage Works.

I(
Hudson, W. D. 1966(Feb). "Studies of Distribution System Capacity in Seven
Cities," J. AWWA, Vol. 58, No. 2.
Hudson, W. D. 1966(Apr). "Loss of Water Main Capacity," J. Southeastern Sec.
AWWA, Vol. 30, No. ', p. 44.
Hudson, W. D. 1973(Fcb). "Computerizing Pipeline Design," ASCE Transporta-
tion Engineering Journal, Vol. 99, No. TEl, p. 73.
Lamont, P. A. 1981(May). "Common Pipe Flow Formulas Compared with the Theory
of Roughness," J. AWWA, Vol. 73, No. 5, p. 274.
Larson, T. E., and Sollo, F. W. 1967(Dec). "Loss in Water Main Carrying
Capacity," J. AWWA, Vol 59, No. 12, p. 1565.
New England Water Works Association Comittee on Pipeline Friction Coef-
ficients. 1935. "Summary of Report," J. New England Water Works Association,
Vol 49, p. 239.
Nikuradse, J. 1932. "Gesetzmassigkeiten der Turbulenten Stromung in Glatten
Rohren," VDI Forschungsh, Vol. 356.
Ormsbee, L. E., and Wood, D. J. 1986(Apr). "Explicit Pipe Network Calibra-
tion," J. of Water Resources Planning and Management, Vol. 112, No. 2, p. 166.
Prandtl, L. 1933. "Neuere Ergebnisse der Turbulenxforschung," Zeit. Ver.
deu. Ing., Vol. 77, p. 105.
Streeter, V. L. 1971. Fluid Mechanics, McGraw-Hill, Inc., New York.
Swamee, P. K., and Jain, A. K. 1976(May). "Explicit Equations for Pipe-Flow
Problems," ASCE Journal of Hydraulics Division, Vol. 102, No. HY5, p. 657.
von Karman, T. 1930. "Mechanisehe Ahnlichkeit und Turbulenz," Proc.
3rd International Congress for Applied Mechanics, Stockholm.
Walski, T. M. 1984. Analysis of Water Distribution Systems, Van Nostrand
Reinhold, Inc., New York.

241

| ~tLIl~,~I~EJLw uIRRkA ~ n
Walski, T. M. 1986(Apr). "Case Study: Pipe Network Model Calibration
Issues," J. of Water Resources Planning and Managemnnt, Vol 112, No. 2,
p. 238.
Williams, G. S., and Hazen, A. 1920. Hydraulic Tables, John Wiley & Sons,

I
New York.

I
I
S
I
0
0