Australian Journal of Basic and Applied Sciences, 5(1): 38-47, 2011

ISSN 1991-8178

Application of Analytical and Numerical Methods to the Sequent Depths Problem in

Civil Engineering
1
Prabhata K. Swamee, 1Pushpa N. Rathie and 2Luan Carlos de S.M. Ozelim
1
Department of Civil Engineering National Institute of Technology Jalandhar, India
2
University of Brasília, UnB Brasília, Brazil

Abstract: A classical problem on hydraulic research is to provide explicit equations for evaluating
the sequent depths of open channels whose geometry is not rectangular since barring for this section
configuration, other solutions are nearly inexistent. Exponential section is the category which includes
triangular, parabolic and rectangular channel shapes. For rectangular channel section, as pointed before,
it is possible to express the sequent depths analytically by radicals. For other sections, the sequent
depths are presently obtained by computational methods. In this paper, in order to find the sequent
depths of channels whose sections are exponential or trapezoidal, we apply two different methods: The
Lagrange’s Inversion Theorem, which is analytical and provides an exact solution by means of an
infinite series; and the Householder’s Methods, which are numerical and provide approximations of
the solutions by using an iterative algorithm. In general, the series obtained from Lagrange’s theorem
have fast convergence. Otherwise, if the convergence rate is low, we use the Householder’s methods.
Practical examples are also included.

Key words: Exponential section, Trapezoidal section, Lagrange expansion, Householder’s methods,
sequent depths.

INTRODUCTION

The governing equation of sequent depths is the specific momentum equation given by (Bakhmeteff, 1932;
Chow, 1959).

Q2
M  Az (1)
gA
Where Q = discharge, g = gravitational acceleration; and A = flow area, M = specific momentum; and z =
depth of center of gravity of flow area below the free surface. For majority of cross sections (1) has two
positive roots that are called sequent depths Henderson, (1966).
The present effort is carried out for finding out explicit solution for sequent depths for exponential
sections, which include triangular, rectangular and parabolic channel shapes, besides trapezoidal section. The
equations for each of those channels are, in general, high degree polynomials, such as cubics, quartics and
quintics. It’s known that the first two cases are quite simple to manipulate and give analytical solutions. On
the other hand, quintics are far harder to work with, what makes necessary the use of some other methods.
Mathematicians have known, for a long time, a theorem that fits perfectly to our needs: The Lagrange’s
inversion theorem. By giving an infinite series of implicitly defined functions, this theorem provides an exact
solution to any of our equations. The major concern of such series expansion is the rate of convergence, which
is in general high. Notwithstanding, in a few occasions one shall get slow convergence, what makes this
method a bad alternative. To solve this issue, we apply a numerical solution, the Householder’s methods. By
starting from an initial guess, an iterative formula provides approximations of the solution. One shall choose
the order of convergence of such method. By the way, the quadratic convergence expression is the well-known
Newton-Rhapson algorithm. Both methods will be stated clearly below:

Corresponding Author: Pushpa N. Rathie, Department of Civil Engineering National Institute of Technology Jalandhar,
India
Email: pushpanrathie@yahoo.com
38
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Lagrange’s Inversion Theorem:

The theorem is stated as: Let y be defined as the following function of constant a, function φ, and a
parameter θ

y  a  θφ( y ) (2)

Then any function f(y) is expressed as the following power series in θ Whittaker and Watson, (1990):


θ n  d n 1 
f ( y )  f (a)    n 1  f '( x)φ ( x)  
n
(3)
n 1 n !  dx x a
It can be noticed that the right hand side of (3) contains y through θ defined in (2).

Householder’s Methods:
Consider a function f(x) whose roots need to be estimated. By choosing an initial point, x0, close enough
to the value of the root, Householder, (1970) provided the general iteration root-finding recurrence formula:

 1/ f ( a ) 
xn 1  xn  (a  1)   (4)
 1/ f ( a 1) 
  xn
Where a+2 is the desired order of convergence of the iteration and (1/ f )( a ) is the a-th derivative of the
inverse of the function f(x).

Channel Sections:
In this section, exponential and trapezoidal sections are discussed.

Exponential Section:
An exponential section is described by
p
Y  kX (5)

where X and Y are horizontal and vertical coordinate axes respectively; k = coefficient; and p = exponent. For
p = 1, the exponential section is a triangle of side slope k vertical to 1 horizontal, for p = 2, it is a parabola
of latus rectum k-2. For p = 4, Y = 0 when kX is numerically less than unity, and Y = 4 when kX numerically
just exceeds unity. Thus a rectangular section of bed width 2/k is obtained. The area of the section for depth
y is given by

2 py  
p 1 / p
A (6)
k  p  1

The first moment of area about the free surface, Az is given by

2 p 2 y  2 p 1 / p
Az  (7a)
k  p  1 2 p  1

where M = specific momentum; and z = depth of center of gravity below free surface.

Swamee, (1993) gave the following equation for critical depth in an exponential channel:
p
 p  1 3 Q 2 k 2  3 p  2
yc     (7b)
 p  4 g 

39
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Fig. 1: Canal Sections (a) Triangular (b) Rectangular

Fig. 2: Parabolic Sections:

Trapezoidal Section:
Due to stability questions, trapezoidal channels are preferred for carrying large discharges. The main issue
is the side slope thickness, since vertical side walls require large thickness to resist the earth pressure. On the
other hand, sloping side walls require less thickness. For a trapezoidal section of bed width b and side slope
m horizontal to 1 vertical (See Fig. 3), the flow area is

A  y  b  my  (8a)

where m = side slope m horizontal to 1 vertical. See Fig. 3. The moment of the flow area about the free
surface is
1 1
Az  by 2  my 3 (8b)
2 3
Analytical Considerations- Sequent Depths:
Combining (1), (6) and (7a), one gets

Q 2 k  p  1 2 p2 y 
2 p 1 / p
M  (9)
g 2 py  p 1 / p k  p  1 2 p  1

40
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Fig. 3: Trapezoidal Section

Dividing through by yc3 and using (7b), following equation is obtained:

2
2 p2 z 
2 p 1 / p
2  p 
mc     (10)
z  p  1   p  1 2 p  1
p 1 / p

where
Mk
mc   2 p 1 / p
(11)
y c

Triangular Section:
Putting p = 1 and k = 1/m for a triangular channel, (10) is converted to the following quintic equation:
1 z3
mc   (12)
2z2 3

Where z = y/yc and

M
mc  (13)
myc3

Rewriting (11) for obtaining z1, one gets

1 1 5
z12   z (14)
2mc 3mc

Taking z12  w one finds

1 1 5/ 2
w  w (15)
2mc 3mc

Using Lagrange series solution for f(w) = w1/2, (15) gives

 3mc 
n
  d n 1 x 5 n 1 / 2 
z1   2mc  
1/ 2
 n 1  (16)
n 1 n!  dx 2  x (2 m )1
c

41
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Equation (16) is evaluated as

1  2 3 mc5 n 1 / 2   5n  1 / 2 

  3 n 1 / 2  n

z1   2mc   
1/ 2
(17)
2 n 1 n !  3n  3 / 2 

Equation (17) is simplified as

mc3   0.5880  
5/ 2 5 15 / 2 10
 0.6274   0.6538   0.6731 
z1   2mc 
1/ 2
 1       .     ... (18)
24   mc   mc   mc   mc  

In order to get z2, (12) is rewritten as
 3 
z23  3mc    z22 (19)
 2 

Taking r  z23 one get

3 2 / 3
r  3mc  r (20)
2
Using Lagrange series solution for f(r) = r1/3, (20) gives

 3 / 2   d n 1 x 2 n 1 / 3 
n

z2   3mc  
1/ 3

n!  dx n 1 3  (21)
n 1   x 3mc
Equation (21) is simplified to

 3 2 n  2 / 32 n mc 5 n 1 / 3  5n  1 / 3

z2   3mc  
1/ 3
(22)
n 1 n !  2n  2  / 3

Equation (22) is further evaluated as

mc4 / 3   0.4251 5/ 3  0.4922 10 / 3  0.5368 5  0.5692  20 / 3 

z2   3mc 
1/ 3
 4/3 1       .     ... (23)
2.3   mc   m c   m c   m c  
Parabolic Section:
Putting p = 2 for a parabolic channel, (10) is converted to
8 8 z 5/ 2
mc  3/ 2  (24)
9z 15
where
Mk
mc  (25)
yc5/ 2
wherein k-2 is latus rectum of the parabola.

For obtaining z1, rewriting (24) as z13/ 2  8 /(9mc ) 8 z14 /(15mc ) and further denoting q  z13/ 2 , (24)
becomes
8 8 8/3
q  q (26)
9mc 15mc

42
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Using Lagrange’s series expansion for f(q) = q2/3 ,(26) provides

2/3 n
2  1  8   d n 1 x   
8 n 1 / 3
 8 
z1          (27)
 9mc  3 n 1 n !  15mc   dx n 1  x 8 /(9 m )
c
Equation (27) is simplified to

 8 
2/3

3
13 n  7  / 3 8 n  3  n
2 5 mc 
 8n  2 / 3   8n  2  / 3
z1     x x (28)
 9mc  n 1 n!   5n  5  / 3
By evaluating (28) one gets

 8 
2/3
 0.6753 
10 / 3
  1.0349 8 / 3  1.0967 16 / 3  1.1383 8  1.1687 32 / 3 
z1      1       .     ... (29)
 9mc   mc    mc   mc   mc   mc  

To obtain z2, (24) will be rearranged as

15mc 5 3/ 5
s  s (30)
8 3

where s  z2
5/ 2
.

Using Lagrange’s theorem for f(s)=s2/5, (30) gives:

 15mc 
2/5
1  5 
 n
 d n 1 2 x 3 n 1 / 5 
z2        n 1  (31)
 8  n 1 n !  3   dx 5  x 15 mc / 8
Equation (31) is simplified to

mc    8n  2  / 5
2/5  13 n  2  / 5 3( n 1) / 5 (24 n 1) / 5  8 n  2 / 5
 15mc  3 5 2
z2     xx (32)
 8  n 1 n!   3n  3 / 5
Equation (32) is evaluated as

 15mc 
2/5
 0.0080 
6/5
  0.6872 8 / 5  0.8080 16 / 5  0.8883 24 / 5  0.9467 32 / 5 
z1      1        .       ... (33)
 8   mc    mc   mc   mc   mc  

Rectangular Section:
Putting p = ¥ for a rectangular, (10) is converted to the following cubic equation:
2 2
mc  z (34)
z
where
2M
mc  (35)
byc2

where k = 2/b has been substituted. Equation (34) is rewritten as

z 3  mc z  2  0 (36)

43
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Solving (36) as cubic Tignol (2007), the sequent depths are obtained as

1/ 2
m   δ  2π 
z1  2  c  cos   (37)
 3   3 
1/ 2
m  δ
z2  2  c  cos   (38)
 3  3
Where

  3 3/ 2 
δ  arccos      mc  3 (39)
  mc  
As an alternate solution, using Lagrange’s theorem for z1, (36) is rewritten as
mc 1 2
z11   z1 (40)
2 2
1
Taking z1  l one gets
mc 1 2
l  l (41)
2 2
Using Lagrange series solution for f(l)=l-1,(41) gives

2 
2 n (1) n 1  d n 1 2 n 1 
z1    n 1 x  (42)
mc n 1 n!  dx  xm c /2

Equation (42) is simplified to

 2 n 1
mc    3n  1 
 2  3 n 1
2
z1   (43)
mc n 1 n !  2n  2  

Equation (43) is further evaluated as

8   2.2894   2.4019 
3 6 9
2  2.4778   2.6840 
z1   1       .    (44)
mc mc4   mc   mc   mc   mc 

For obtaining z2, (36) is written as

z22  mc  2 z21 (45)

Taking z22  u one gets

u  mc  2u 1/ 2 (46)

Using Lagrange series solution for f(u)=u1/2, (46) gives


(2) n  d n 1 x  n 1 / 2 
z2   mc  
1/ 2
  (47)
n 1 n !  dx n 1 2  x  m
c

44
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Equation (47) is simplified to

2n 1 mc    3n  1 / 2 
  3 n 1 / 2

z2   mc  
1/ 2
(48)
n 1 n !  n  1 / 2 

Equation (48) is further evaluated as

1   1.3104  
3/ 2 3 9/ 2 6
 1.5874   1.7720   1.9064 
z2  m 1/ 2
c  1        .       ... (49)
mc   mc   mc   mc   mc  
 
Trapezoidal channel:
Using (8a) and (8b), (1) reduces to

Gb2 1 1
mb   α 2  α3 (50)
α 1  α  2 3

where mb = m2M/b3. In order to get α1, (50) is rewritten as

Gb2 1   5 5 4  3 
1       mb 2  (51)
mb mb  3 6 2 
Applying Lagrange’s inverse expansion given by (3) to (51), one gets

Gb2  mb n d n 1  x5 5 x 4 x3 
n
2
α1       mb x   (52)
mb n 1 n! dx n 1  3 6 2   x  Gb2

mb
Equation (52) does not provide a good value of α1 due to the slow rate of convergence of the series. In this

case, one shall use Householder’s methods of iteration by taking  *  Gb2 mb as the first estimative, due to

its closeness to the exact value of α1. Thus, by rearranging (51) and denoting:
5 4 3 3
l ( )   5    3mb 2  3mb  3Gb2 (53)
2 2
The derivatives are:
l ( ) 9 2
l ( )   5 4  10 3   6mb  3mb (54)
 2

 2l ( )
l ( )   20 3  30 2  9  6mb (55)
 2

 3l ( )
l ( )   60 2  60  9 (56)
 3
Finally, considering

l*  (l / l )(* ) (57)

45
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

The value of α1 with the quadratic convergence method is

1  *  l* (58)

For cubic convergence one gets

l*
1   *  (59)
 l  l   (* ) 
1  l*
2
And, for quadric convergence
  l  l   ( * )  
l* 1   l* 

 2 
1   *  (60)
 l  l   ( * )  2
1   l  l   (* )  l*   l*
6
To get α2, (50) is written as

 Gb2 α3 
α 22  2mb  2   2  (61)
 α 2 1  α 2  3 
Taking u  α 22 , one gets

 G2 u 3/ 2 
u  2mb  2  1/ 2 b 1/ 2   (62)
 u 1  u  3 
 
Applying Lagrange’s inverse expansion given by (8) for f(u) = u1/2 to (62), one gets

 n

d n 1  1 1/ 2 
 2  x3/ 2  
n

Gb2
 2  (2mb ) 1/ 2
 x    (63)
n! dx n 1  2  x1/ 2 1  x1/ 2  3  
n 1
    x2m
b

Again, due to slow convergence, one shall use Householder’s methods. Now, following the result of (63)

the first estimative shall be taken as *   2mb 

1/ 2
. Since the sequent depths are roots of the same

polynomial, equations from (53) to (60) apply for both α1 and α2 , thus the results are already given above.

Practical Examples:
For illustrating the use of equations obtained in the foregoing sections, the following examples are
presented.

Example 1:
A 4 m wide rectangular channel carries a discharge of 10 m3/s. Find the sequent depths corresponding to
a specific momentum of 10 m3.

Solution:
In this case g = 9.79 m/s2 is adopted.
The specific momentum parameter mc  2 M /(byc2 )  20 /(4x0.8612 ) = 6.745. Using Lagrange Series,
the solution is described as:

Taking 3 terms:
y1 = 0.259 m; and y2 = 2.100 m.

46
 Aust. J. Basic & Appl. Sci., 5(1): 38-47, 2011

Taking 4 terms:
y1 = 0.259 m; and y2 = 2.096 m.

Taking 5 terms:
y1 = 0.259 m; and y2 = 2.096 m.

On the other hand, the explicit solution provides:

y1 = 0.259 m; and y2 = 2.096 m.

Example 2:
A 2 m wide trapezoidal channel having side slope m = 1 carries a discharge of 10 m3/s. Find the sequent
depths corresponding to a specific momentum of 10 m3.
Solution: In this case g = 9.79 m/s2 is adopted. For the parameter Gb:

 
Gb  m1.5Q / b 2 gb  10 /(22 9.79 x 2)  0.565
The specific momentum parameter mb = m2M/b3 = 10/8 = 1.25. Using Lagrange Series combined with
Householder’s methods, the solution is described as:

For quadratic convergence:

y1 = 0.431 m; and y2 = 2.645 m.

For cubic convergence:

y1 = 0.430 m; and y2 = 2.415 m.

For quadric convergence:

y1 = 0.430 m; and y2 = 2.317 m.

By trial and error procedures, the following was obtained:

y1 = 0.429 m; and y2 = 2.258 m

Conclusion:
Analytical and numerical methods have been applied to solve the sequent depth problems. The results
given are useful in situations where no computational resource is available and to give accurate and fast
converging algorithms for future civil engineering software. It’s known that computational methods are far
easier to use than the ones provided in the present paper, however, the methods and ideas presented provide
unexplored ways of solving civil engineering problems. The results obtained are very accurate, proving the
effectiveness of the formulas given.

REFERENCES

Bakhmeteff, B.A., 1932. Hydraulics of open channels. McGraw-Hill, New York.

Chow, V.T., 1959. Open channel hydraulics. McGraw-Hill, Inc., New York.
Henderson, F.M., 1966. Open channel flow. The MacMillan Co., New York.
Householder, A.S., 1970. The Numerical Treatment of a Single Nonlinear Equation. New York: McGraw-
Hill.
Swamee, P.K., 1993. “Critical depth equations for irrigation canals.” J. Irrig. and Drain. Eng., ASCE,
119(2): 400-409.
Tignol, J.P., 2001. Galois' theory of algebraic equations. World Scientific Publishing Company, Singapore;
2001.
Whittaker, E.T., G.N. Watson, 1990. A course of modern analysis. 4th Edition, Cambridge Univ. Press,
Cambridge, U.K., pp: 133.

47