You are on page 1of 32

UNSTEADY FRICTION IN

TRANSIENT FLOW
DR. WALTER F. SILVA-ARAYA
ERDC VICKSBURG
JULY, 2014

UNSTEADY PIPE FLOW EQUATIONS


equations of one-dimensional unsteady pipe flow:
Governing

Continuity:
Momentum equation:
Js and Ju are the steady and unsteady friction terms.
The steady friction can be expressed by the Darcy-Weisbach relation as
(several equations available to estimate f)
Solution by the method of characteristics (MOC) for H and V (most
efficient)

THE PROBLEM
Energy dissipation in the modeling of transient flow in pipe
systems is not adequately accounted for by using steady state
formulas such as Darcy-Weisbach or Hazen-Williams
70
computed
experimental

60

P re s s u re (m )

50
40
30
20
10
0
-10
-20

0.2

0.4

0.6

0.8

1
Time (s)

1.2

1.4

1.6

1.8

UNSTEADY FRICTION MODELS


1.

Instantaneous Acceleration Models (Brunone et al. 1991,


Ramos et al. 2003, Vitkovsky et al. 2006)

2.

Weighted Function Based models (Zielke, 1968, Vardy and


Hwang, 1991, Vardy and Brown 2004, 2007)

3.

Two dimensional models (Silva-Araya and Chaudhry 1997,


Pezzinga, 1999)

INSTANTANEOUS ACCELERATION BASED MODELS


The IAB models assume that the wall shear stress is a function of
the instantaneous local accelerations and an empirical coefficient
which may be a function of the fluid properties.

One coefficient formulation (Brunone et al. 1991 ):

Two coefficients formulation (Vtkovsky et al., 2001) :

INSTANTANEOUS ACCELERATION BASED MODELS

is the sign of the convective acceleration term and was


introduced to produce damping during acceleration and have no
effect during decelerations.

Kut and Kux are two decay coefficients; related to the local and
convective accelerations respectively.

It has been verified numerically that the term KutV/t affects the
phase shift of transient pressure waves and KuxV/x the damping.
(Ramos et al. 2003).

ESTIMATION OF DECAY COEFFICIENTS


The number of parameters could be one or two depending on the chosen model.
The empirical coefficients must be supplied for application of the IAB unsteady
friction models.

Several formulas have been proposed to estimate the coefficient for oneparameter formulations

H. Prashanth Reddy,Walter F. Silva-Araya, andM. Hanif


Chaudhry(2012).Estimation of Decay Coefficients for Unsteady Friction for
Instantaneous, Acceleration-Based Models.J. Hydraul. Eng.,138(3), 260271:
Used Genetic Algorithms to estimate Decay Coefficients

GENETIC ALGORITHMS
A genetic algorithm (GA) is a search technique used in computing to find
exact or approximate solutions to optimization and search problems.
The GA is a random search algorithm based on the concept of natural
selection inherent in natural genetics, presents a robust method for
search for the optimum solution to the complex problems.

The artificial survival of better solution in GA search technique is


achieved with genetic operators: inheritance, mutation, selection and
crossover based on evolutionary biology.

GA: OBJECTIVE FUNCTION


The least squares (LS) are commonly used as the objective function in optimization
problems.

The normalized root mean-squared deviation or error (NRMSE) is a scaled value of the LS
objective function.

NRMSE is the root mean-squared deviation (RMSD) divided by the range of measured
values.

NRMSD provides a measure of the goodness of fit between the experimental and computed
data allowing comparison between different data sets because is non dimensional.

Differences between the values predicted by model and the measured values are used in
this parameter which is minimized by GA

Complete time-history of pressure measurements at the valve location are used to estimate the IAB coefficients in
this research using NRMSE as the objective function.

N = total number of time levels of data


J = time level Hj* & Hj = measured and computed pressures at time level j
Hmax* & Hmin* = maximum and minimum values of measured pressures.

IAB ONE COEFFICIENT


Positive compatibility:

Positive characteristic
Negative compatibility:

Negative characteristic:

IAB ONE COEFFICIENT


Rearranging positive characteristic
equation:

Rearranging negative characteristic


equation:

IAB TWO COEFFICIENT MODEL


Positive compatibility:

Positive characteristic
Negative compatibility:
Negative characteristic:

DATA BASE OF EXPERIMENTAL DATA


Case

L (m)
98.11

D (m)
0.016

Vo (m/s)
0.066

a
(m/s)
1298

Pipe
material
Copper

1112

0.80

IMP-PAN1
USC1

156.00

0.025

0.057

1180

Copper

1424

0.76

USC2

156.00

0.025

0.094

1061

Copper

2348

0.46

IMP-PAN2
E4

98.11

0.016

0.340

1298

Copper

5729

0.16

37.20

0.022

0.300

1320

Copper

6593

0.65

98.11

0.016

98.11

0.016

0.631

1298

Copper

10632

0.08

1298

Copper

15838

0.06

IMP-PAN3
IMP-PAN4
R1

28.00

R2

14.00

P1
P2
E3
P3
P4

143.70
143.70
37.20
143.70
143.70

0.046

0.940
0.692

373

PVC

31800

0.22

0.046

0.903

350

PVC

41496

0.32

0.053
0.053
0.022
0.053
0.053

0.017
0.100
0.300
0.184
0.288

1363
1363
1320
1363
1363

Steel
Steel
Copper
Steel
Steel

900
5295
6593
9742
15249

7.46
1.26
0.65
0.69
0.44

Valve
location
Downstrea
m
Downstrea
m
Downstrea
m
Downstrea
m
Downstrea
m
Downstrea
m
Downstrea
m
Downstrea
m
Downstrea
m
Upstream
Upstream
Upstream
Upstream
Upstream

Reference
Adamkowski and Lewandowki, 2006
Sattar et al., 2008
Sattar et al., 2008
Adamkowski and Lewandowki, 2006
Bergant and Simpson, 1995
Adamkowski and Lewandowki, 2006
Adamkowski and Lewandowki, 2006
Ramos et al., 2004
Ramos et al., 2004
Pezzinga and Scandura,1995
Pezzinga and Scandura,1995
Bergant and Simpson, 1995
Pezzinga and Scandura,1995
Pezzinga and Scandura,1995

IAB ONE COEFFICIENT


Vardy-Hwang (1995)

Vardy-Hwang (2003)

Reddy et al. (2012)

1,400 < Re < 50,000

DECAY COEFFICIENTS (USVC AND DSVC)


0.07

K GA DSVC

K GA USVC

K VB 2003

0.06
0.05
IAB Decay Coefficients

PVC Pipe

R-S-C
-2012

Laminar
Flow

K VB 1995

0.04
0.03

VB- 2003

VB - 1995
0.02
0.01
0
100

1000

10000
Reynolds Number

100000

IAB TWO DECAY COEFFICIENTS MODEL


Within the range of Reynolds numbers included in this analysis, the decay coefficients
for the two-coefficient models for turbulent flow in elastic pipes are relatively constant.

The following values are recommended for preliminary analysis:


Downstream Valve Closures-TwoCoef/MOC: Kux = 0.046, Kut = 0.034 (5,000 < Re < 30,000)
Downstream Valve Closures - TwoCoef/FD: Kux = 0.033, Kut = 0.034

(5,000 < Re < 30,000)

Upstream Valve Closures - TwoCoef/MOC: Kux = 0.031, Kut =0.011 (5,000 < Re < 15,000)
Upstream Valve Closures - TwoCoef/FD:

Kux = 0.028, Kut =0.025 (5,000 < Re < 15,000)

More data is needed to establish dependence of the decay coefficients for laminar
flows.

The number of upstream valve closures experiments found in the literature is limited
and do not cover a wide range of Reynolds numbers.

WEIGHTING FUNCTION BASED MODELS


Zielke (1968) presented a weighting function model of
transient laminar pipe friction.

This model accounts for the influence of the local velocity


history at any point along a pipe when deducing the
instantaneous magnitude of the local skin friction.

The next stage in complexity is to introduce rates of change


of acceleration and higher order derivatives, leading to
formulae of the type

WEIGHTING FUNCTION BASED MODELS

Following Zielkes idea the unsteady shear stress can be

expressed as a weighted integral of accelerations at all preceding

times:

t = elapsed
timeapplications:
since calculations began
computer
For

t* = T t is measured backwards in time from the current instant, t = T.


The weighting function W is a unique function of t*
dU/dt is a function of t.

i = t*/t
N = T/t
t is the time interval between successive calculations.
In practice, the weighting function is small when t*is large so the numerical integration
can be truncated. However, the number of terms to be retained is inversely proportional
to the magnitude of the integration time step and is sometimes very large.

WEIGHTING FUNCTION BASED MODELS


A turbulence model is required for the shear stress
distribution
used eddy viscosity (several
Vardy and Brown (1995)

versions have been proposed):

= rate
of change of kinematic viscosity from

at the wall to

at the core/shear-layer

EQUATIONS OF FLOW

Layer
Shear

Solve by Laplace Transform.

Obtain transformed of the weighting function, .


Finally inverse transform to obtain . (Vardy and Brown, 1995)

Several variations of this method have been proposed.

R = pipe radius

CASE STUDY
Measured pressure history at the valve
70
60
50
40
30
20
10
0
0.000
-10
-20

0.500

1.000

1.500

2.000

2.500

CASE STUDY

Steady State Friction


Computed

Experimental

70
60
50
40
30
Pressure (m)
20
10
0
0

0.2

0.4

0.6

0.8

-10
-20
Time (s)

1.2

1.4

1.6

1.8

CASE STUDY

OneCoef/MOC

80

Experimental

Computed

70
60
50
40
Pressure
30 (m)
20
10
0
0.0

0.2

0.4

0.6

0.8

1.0

-10
-20
Time (s)

1.2

1.4

1.6

1.8

2.0

CASE STUDY

TwoCoef/FD

Experimental

80

Computed

70
60
50
40
30 (m)
Pressure
20
10
0
0.0

0.2

0.4

0.6

0.8

1.0

-10
-20
Time (s)

1.2

1.4

1.6

1.8

2.0

CASE STUDY

TwoCoef/MOC

80

Experimental

Computed

70
60
50
40
Pressure
30 (m)
20
10
0
0.0

0.2

0.4

0.6

0.8

1.0

-10
-20
Time (s)

1.2

1.4

1.6

1.8

2.0

CASE STUDY
Vardy and Brown (2002) WFB
70

Computed

Experimental

60
50
40
30
Pressure head (m)
20
10
0
0

0.2

0.4

0.6

0.8

-10
-20
Time (s)

1.2

1.4

1.6

1.8

CASE STUDY
Vardy-Brown (2004) WFB model for rough pipes
70

Computed

Experimental

60
50
40
30
Pressure head (m)
20
10
0
0

0.2

0.4

0.6

0.8

-10
-20
Time (s)

1.2

1.4

1.6

1.8

Valve Closure Parameter


0.8
0.6

CASE STUDY

Axis Title 0.4


0.2

Silva & Chaudhrys model

0
0

0.5

1.5

Time (s)

Computed and Measured Pressure Head at the Valve

Pressure Head (m)

70
60
50
40
30
20
10
0
0
-10
-20

0.5

1.5

Time (s)

Computed Pressure Head

Measured Pressure Head

2.5

2.5

CASE STUDY

Velocity Profiles

Silva & Chaudhrys model

0.5 s

0.7 s

0.9 s

3
1.1 s
2.5

1.5

0.5

0.01

0.02

0.03

0.04

0
0.05

-0.5

0.06

CONCLUSIONS
One dimensional IAB and WFB unsteady friction models have been verified under limited
conditions, mostly instantaneous valve closures, small pipe diameters, short pipe lengths
and smooth pipe walls.

Most of the conclusions about their capabilities to simulate transients for long periods and in
rough pipes have been deduced from theoretical considerations and from the experience for
the model behavior for sudden valve closures in smooth pipes. Few field examples are
available and the transients in those cases are under controlled conditions and are not
severe.

During slow control valve movements, the flow profiles have sufficient time to adapt to the
transient conditions and expand those changes into the inner turbulent region before the
movement is completed. When the valve closes completely, the flow conditions are
different from the initial steady state.

The IAB and WFB unsteady friction models should be used with caution in general-purpose
transient programs for the simulation of complex systems with fast control mechanisms.

CONCLUSIONS
Two dimensional or quasi-two-dimensional unsteady friction models
are computationally time-intensive but produce an approximation of
the velocity profiles in transient flows, including a flow reversal near
the wall.

If the energy dissipation associated with these profiles is accounted


for, then the model should provide good results for simulation over
longer periods of times.

Turbulence models for transient flows are a matter of research;


however, the development of efficient two-dimensional unsteady
friction models is a promising alternative for an accurate prediction of
energy dissipation in transient flow.

THANKS

You might also like