You are on page 1of 8

Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 186 (2017) 61 – 68

XVIII International Conference on Water Distribution Systems Analysis, WDSA2016

Dynamic programming over a graph modeling framework for the


optimal design of pipe series in sewer systems
Natalia Duquea, Daniel Duqueb, Juan Saldarriagaa*
a
Centro de Investigaciones en Acueductos y Alcantarillados (CIACUA), Departamento de Ingeniería Civil y Ambiental, Universidad de los
Andes, Bogotá, Colombia.
b
Centro para la Optimización y Probabilidad Aplicada (COPA), Departamento de Ingeniería Industrial, Universidad de los Andes, Bogotá,
Colombia.

Abstract

The sewer network optimal hydraulic design, for a specific layout consisting of a series of pipes, is the combination of diameters
and slopes along the series that satisfy all the hydraulic, commercial, and construction constraints, while minimizing the
construction costs. This work explains an exact, exhaustive, and flexible framework to find the cost-optimal design of a series of
sewer pipes using a DP-based optimization engine [1]. The hydraulic design problem is modeled as a Shortest Path Problem [2],
where the underlying graph considers every feasible combination of diameter and slope for each pipe in the series. As a result, a
shortest path on the graph encodes the optimal hydraulic design decision for the series of pipes. The proposed methodology ensures
the global optimal solution from an economic point of view because the graph considers all possible alternatives and the Bellman-
Ford [3] algorithm implicitly explores all of them. Without having to simplify hydraulic constraints, this methodology still obtains
the optimal solution in a very short computational time using a standard desktop computer. To evaluate the performance of the
methodology, several numerical examples are presented varying the pipe material, the topography, and the number of pipes in the
series.
© 2016 The Authors. Published by Elsevier Ltd.
© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the organizing committee of the XVIII International Conference on Water Distribution
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Systems. under responsibility of the organizing committee of the XVIII International Conference on Water Distribution Systems
Peer-review

Keywords: Sewer systems; design of series of pipes; optimization; graph modelling; shortest path problem.

1. Introduction

* Corresponding author. Tel.: +57-1-3394949 ext.2805.


E-mail address: jsaldarr@uniandes.edu.co

1877-7058 © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the XVIII International Conference on Water Distribution Systems
doi:10.1016/j.proeng.2017.03.208
62 Natalia Duque et al. / Procedia Engineering 186 (2017) 61 – 68

The sewer network design problem consists of determining both the layout and the hydraulic design of the system,
where the layout is defined by the initial pipes and the flow direction in each pipe. The layout for a series of pipes is
given by a unique initial pipe where the series starts and a unique flow direction towards the last manhole (outfall).
On the other hand the hydraulic design determines the diameter and the slope of the pipe to be installed at each link
of the sewer network, and a link refers to the span between two consecutive manholes. Moreover, pipe diameters are
chosen from a discrete set of commercial diameters and the slope of a pipe is related to the elevation gap of its
extremes. The optimal hydraulic design is the minimum-cost design that accomplishes all hydraulic and construction
constraints established by the corresponding national legislation.

Due to the complexity of solving both sewer network design problems (layout and hydraulic design) simultaneously,
the literature propose a sequential process by finding a layout and then its hydraulic design. In particular, the hydraulic
design problem has been tackled using exact and heuristic methodologies. Dynamic Programing (DP) has been used
for the design of a series of pipes where the decision variables include the invert elevation at the end of each segment
and its diameter [4, 5]. However, this methodology had computational limitations due to the well-known curse of
dimensionality, which limits their ability to solve large-scale instances. Likewise, Nonlinear Programing (NLP) has
been implemented to establish both topographic and hydraulic factors such as flow rate, size, and gradient of pipes
[6]. Finally, Linear Programing (LP) was used to obtain an optimal design for a series of pipes without linearizing the
objective function nor the constraints, by fixing the lengths and diameters of each pipe to the commercially available
[7].

In this problem, the input information of the problem includes topographic information (horizontal length of the links
and ground elevation of the manholes), commercial characteristics (available pipe materials and diameters), physical
characteristics of the fluid (water density and viscosity), and hydraulic characteristics (flow resistance formula and
pipe’s internal roughness). Moreover, it is assumed that the design flow rate for each pipe is known beforehand and
corresponds to the inflow at the upstream manhole of the pipe plus the flow rate coming from the upstream pipes.
Additionally, the hydraulic constraints that must be fulfilled are: a minimum pipe diameter, a maximum filling ratio,
a minimum wall shear stress, a minimum and maximum velocity, and a minimum and maximum slope [1].

This article presents a brief explanation of the mathematical model and methodology used for obtaining the optimal
hydraulic design [1] and some numerical examples to evaluate its performance in large series of pipes.

Nomenclature

࣡ Graph.
ࣨ Set of nodes ࣨ ൌ ሼ‫ݒ‬଴ ǡ Ǥ Ǥ Ǥ ǡ ‫ݒ‬௡ ሽ that belong to ࣡.
ࣛ Set of arcs ࣛ ൌ ൛൫‫ݒ‬௜ ǡ ‫ݒ‬௝ ൯ȁ‫ݒ‬௜ ‫ࣨ א‬ǡ ‫ݒ‬௝ ‫ࣨ א‬ǡ ‫ݒ‬௜ ് ‫ݒ‬௝ ൟ that belong to ࣡.
‫ݒ‬௦ Initial node that belongs to ࣨ in a Shortest Path Problem.
‫ݒ‬௧ Final node that belongs to ࣨ in a Shortest Path Problem.
ܿ௜௝ Cost of the arc ൫‫ݒ‬௜ ǡ ‫ݒ‬௝ ൯ ‫ࣛ א‬.
‫ݔ‬௜௝ Binary decision variable.
ܲ Set of manholes ܲ ൌ ሼ݉଴ ǡ ݉ଵ ǡ ǥ ǡ ݉௄ ሽ.
‫ܦ‬ Set of commercial diameters ‫ ܦ‬ൌ ሼ݀ଵ ǡ ݀ଶ ǡ ǥ ǡ ݀ௗ ሽ.
݉௞ Manhole ݇ ௧௛ that belongs toܲ.

ࣨ௞ Subset of nodes ࣨ௞ ൌ ሼ‫ݒ‬ଵ௞ ǡ ‫ݒ‬ଶ௞ ǡ ǥ ǡ ‫ݒ‬ȁࣨ ೖȁ
ሽ that belongs to manhole ݉௞ ‫ܲ א‬.

‫ݒ‬௜ ݅‫ ݄ݐ‬node at manhole ݉௞ ‫ܲ א‬.
‫׏‬୩ Ground elevation at the manhole ݉௞ ‫ܲ א‬.
‫׏‬ሺ‫ݒ‬௜௞ ሻ Invert elevation at the node ‫ݒ‬௜௞ ‫ࣨ א‬௞ .

‫׏‬ Invert elevation change.
ܳௗ ሺ݇ǡ ݇ ൅ ͳሻ Design flow rate for a pipe between manholes ݉௞ and ݉௞ାଵ ‫ܲ א‬.
ߜሺ‫ݒ‬௜௞ ሻ Diameter at the node ‫ݒ‬௜௞ ‫ࣨ א‬௞ .
‫ݏ‬ሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ Slope of the arc ሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ ‫ࣛ א‬.
Natalia Duque et al. / Procedia Engineering 186 (2017) 61 – 68 63

2. Methodology
The optimal hydraulic design methodology is proposed as an exact, exhaustive and flexible framework, modeled as a
Shortest Path Problem which is a DP-based problem. The Shortest Path Problem is commonly used in transportation
and routing problems to find a minimum-cost path (e.g., shortest distance or minimum travel time) from a specific
initial node ‫ݒ‬௦ ‫ ࣨ א‬to a destination node ‫ݒ‬௧ ‫[ ࣨ א‬3]. This type of problem is defined over a graph࣡ ൌ ሺࣨǡ ࣛሻ,
where ࣨ ൌ ሼ‫ݒ‬଴ ǡ Ǥ Ǥ Ǥ ǡ ‫ݒ‬௡ ሽ is a set of nodes, ࣛ ൌ ൛൫‫ݒ‬௜ ǡ ‫ݒ‬௝ ൯ȁ‫ݒ‬௜ ‫ࣨ א‬ǡ ‫ݒ‬௝ ‫ࣨ א‬ǡ ‫ݒ‬௜ ് ‫ݒ‬௝ ൟ is a set of arcs, and ܿ௜௝ is the cost
(e.g. distance, time, etc) of a traversing arc ൫‫ݒ‬௜ ǡ ‫ݒ‬௝ ൯ ‫ࣛ א‬.

The mathematical formulation is defined by a binary decision variable ‫ݔ‬௜௝ valued as one if the arc ൫‫ݒ‬௜ ǡ ‫ݒ‬௝ ൯ ‫ࣛ א‬
belongs to the shortest path and valued at zero otherwise. The objective function (1) minimizes the costs of the path,
and is subject to Equation (2) that establishes the flow balance constraints that guarantee the structural properties of a
path, and Equation (3) that restricts the variables to be binary-valued [2].

‹ ෍ ܿ௜௝ ‫ݔ‬௜௝ (1)


൫௩೔ ǡ௩ೕ ൯‫ࣛא‬
‫ݏ‬Ǥ ‫ݐ‬Ǥ
ͳ ‫ݒ‬௜ ൌ ‫ݒ‬௦ 
෍ ‫ݔ‬௜௝ െ ෍ ‫ݔ‬௝௜ ൌ ൝ Ͳ ‫ݒ‬௜ ‫̳ࣨ א‬ሼ‫ݒ‬௦ ǡ ‫ݒ‬௧ ሽ Ǣ (2)
ቄ݆ቚ൫‫ݒ‬௜ǡ ‫ݒ‬௝ ൯ ‫ ࣛ א‬ቅ ቄ݆ቚ൫‫ݒ‬௝ ǡ ‫ݒ‬௜ ൯ ‫ ࣛ א‬ቅ
െͳ ‫ݒ‬௜ ൌ ‫ݒ‬௧ 

‫ݔ‬௜௝ ‫ א‬ሼͲǡͳሽ‫ݒ׊‬௜ ‫ࣨ א‬ǡ ‫ݒ‬௝ ‫ࣨ א‬Ǣ (3)

In this case, the Bellman-Ford algorithm was selected to solve the Shortest Path Problem due to the particular structure
of the underlying graph, i.e., a directed-acyclic graph that is ordered in a topological fashion (݅ ൏ ݆ for every arc
൫‫ݒ‬௜ ǡ ‫ݒ‬௝ ൯ ‫)ࣛ א‬. This algorithm is a label-correcting method that starts setting a cumulative cost (label) ܸሺ‫ݒ‬௦ ሻ ൌ Ͳ for
the initial node ‫ݒ‬௦ ‫ࣨ א‬and ܸሺ‫ݒ‬௜ ሻ ൌ λ for the rest of nodes of the graph. Then, each node ‫ݒ‬௜ ‫ ࣨ א‬is evaluated by
scanning every outgoing arc ൫‫ݒ‬௜ ǡ ‫ݒ‬௝ ൯ ‫ࣛ א‬. If ܸ൫‫ݒ‬௝ ൯ ൐ ܸሺ‫ݒ‬௜ ሻ ൅ ܿ௜௝ , then the label for ‫ݒ‬௝ ‫ ࣨ א‬is updated with the new
minimum cumulative cost ܸ൫‫ݒ‬௝ ൯ ൌ ܸሺ‫ݒ‬௜ ሻ ൅ ܿ௜௝ and its parent (predecessor) node ‫݌‬൫‫ݒ‬௝ ൯ ൌ ‫ݒ‬௜ . When all the nodes of
the graph are evaluated, the cost labels are optimal [2].

On these terms, the hydraulic design problem was modelled in a graph where the input parameters are the set of
manholes comprised in the series of pipes ܲ ൌ ሼ݉଴ ǡ ݉ଵ ǡ ǥ ǡ ݉௄ ሽ, where manhole ݉଴ ‫ ܲ א‬is the starting point and
manhole ݉௄ ‫ ܲ א‬is the outfall; a set of commercial diameters ‫ ܦ‬ൌ ሼ݀ଵ ǡ ݀ଶ ǡ ǥ ǡ ݀ௗ ሽ ; the length of each pipe ݈; the
inflow ܳ௞ and ground elevation ‫׏‬௞ at manhole ݉௞ ‫ܲ א‬.The design flow rate for a pipe between manholes ݉௞ and
݉௞ାଵ is precomputed as ܳௗ ሺ݇ǡ ݇ ൅ ͳሻ ൌ σ௞௜ୀ଴ ܳ௜ due to the sequential layout of the series of pipes.


In the underlying graph, each manhole ݉௞ ‫ ܲ א‬is represented by a subset of nodes and ࣨ௞ ൌ ሼ‫ݒ‬ଵ௞ ǡ ‫ݒ‬ଶ௞ ǡ ǥ ǡ ‫ݒ‬ȁࣨೖȁ
ሽ, while
௞ ௞
each node ‫ݒ‬௜ ‫ࣨ א‬௞ encodes two attributes. The first one is the invert elevation on the node ‫׏‬ሺ‫ݒ‬௜ ሻ at manhole ݉௞ ‫א‬
ܲ . The second attribute is a commercial diameter ߜ൫‫ݒ‬௜௞ ൯ ‫ ܦ א‬representing the diameter of an incoming pipe to
manhole ݉௞ ‫ ܲ א‬that starts at manhole ݉௞ିଵ ‫ܲ א‬. In order to evaluate all the possible elevations where a pipe may
෩ is given. It establishes the numeric tolerance of the designs as a lower elevation
be place, an elevation change ߘ
change gives a higher exactness in the design.

Additionally, there is a set of arcs ࣛ ൌ ൛൫‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ൯ห‫ݒ‬௜௞ ‫ࣨ א‬௞ ǡ ‫ݒ‬௝௞ାଵ ‫ࣨ א‬௞ାଵ ൟ which connect two nodes of
consecutive manholes ‫ݒ‬௜௞  ‫ࣨ א‬௞ and ‫ݒ‬௝௞ାଵ ‫ࣨ א‬௞ାଵ , to represent a pipe with a particular diameter and slope. Each arc
has a cost attribute ܿሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ, which corresponds to the total construction costs of the pipe, including the cost of the
pipe and the excavation cost. Equation (4) presents the construction cost equation per arc ൫‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ൯ ‫ࣛ א‬, estimated
for sewer systems in Colombia [8]. This equation is in terms of the diameter of the pipe ݀ሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ in ݉݉ (Equation
64 Natalia Duque et al. / Procedia Engineering 186 (2017) 61 – 68

5) and the excavation volume required for placing the pipe ܸ in ݉ଷ , as a function of the slope between both extremes
of the pipes (Equation 6).

ܿሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ ൌ ͳǤ͵ʹ ‫ כ‬൫ͻͷ͹ͻǤ͵ͳ ‫݀ כ‬ሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ଴Ǥହ଻ଷ଻ ൅ ͳͳ͸͵Ǥ͹͹ ‫ ܸ כ‬ଵǤଷଵ ൯ (4)
݀ሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ ൌ ߜ൫‫ݒ‬௝௞ାଵ ൯ (5)
‫׏‬൫‫ݒ‬௜௞ ൯ െ ‫׏‬൫‫ݒ‬௝௞ାଵ ൯
‫ݏ‬ሺ‫ݒ‬௜௞ ǡ ‫ݒ‬௝௞ାଵ ሻ ൌ (6)
݈

On the other hand, basic hydraulic constraints are taken into account as the graph is built. For instance, the hydraulics
design must satisfy that ߜ൫‫ݒ‬௜௞ ൯ ൑ ߜ൫‫ݒ‬௝௞ାଵ ൯ to prevent obstructions in the system and that ‫׏‬൫‫ݒ‬௜௞ ൯ ൐ ‫׏‬൫‫ݒ‬௝௞ାଵ ൯ to avoid
adverse slopes. Other hydraulic constraints that ensure a proper operation of the sewer network are also taken into
account. Table 1 shows the hydraulic constraints that must be considered according with the Colombian regulation.
The minimum allowed diameter prevents obstructions in the system. The maximum filling ratio prevents overflow
problems and ensures ventilation inside the system to avoid environmental difficulties. The minimum wall shear stress
and velocity ensure a cleaning process inside the pipes; while the maximum velocity prevents problems such as
erosion, cavitation, air entrapment, hydraulic jumps, among others. The minimum and maximum slope constraints are
limited by the minimum and maximum velocities, respectively.

Table 1. Hydraulic design constraints.


Constraint Value Condition
1 Minimum diameter ʹͲͲ݉݉ Always
ͲǤ͹ ݀ ൑ ͸ͲͲܿ݉
2 Maximum filling ratio ͲǤͺ ͲǤ͹ ൑ ‫ ݎܨ‬൑ ͳǤͷ
ͲǤͺͷ Other cases
3 Minimum wall shear stress ʹܲܽ ݀ ൒ ͶͷͲܿ݉
4 Minimum velocity ͲǤ͹ͷ ݉Τ‫ݏ‬ ݀ ൏ ͶͷͲܿ݉
ͷ ݉Τ‫ݏ‬ ݇௦ ൐ ͲǤͲͲͲͳ݉
5 Maximum velocity
ͳͲ ݉Τ‫ݏ‬ ݇௦ ൏ ͲǤͲͲͲͳ݉
6 Minimum slope The one for which the minimum velocity and shear stress are obtained.
7 Maximum slope The one for which the maximum velocity is obtained.

Furthermore, there are a minimum and maximum depths to place the pipes, measure from the ground to the crown
elevation of the pipes. The minimum depth (݄௠௜௡ ) protects the structure of the pipes and ensures that domestic
discharges drain by gravity. The maximum depth (݄௠௔௫ ) limits axial loads over the pipes. These excavation bounds
are presented in Table 2 according to RAS (2000).

Table 2. Excavation limits.


Road classification ࢎ࢓࢏࢔ ሺ࢓ሻ ࢎ࢓ࢇ࢞ ሺ࢓ሻ
Pedestrian or green area 0.75 5.0
Vehicular 1.2 5.0

In sewer systems, pipes must be joined by their invert elevation in order to avoid hydraulic jumps into the system that
might cause erosion of sewer materials and release gases [9]. Also, for the first pipe the invert elevation is set at the
upper excavation limit and the diameter of the upstream node must match the diameter of downstream node. Fig 1
shows the graph that represents the hydraulic design problem for a series of three pipes, four possible elevations and
four commercial diameters.
Natalia Duque et al. / Procedia Engineering 186 (2017) 61 – 68 65

݉଴ ‫ܲ א‬

݉ଵ ‫ܲ א‬

Fig 1. Layout and hydraulic design graph for a series of three pipes of a sewer network.

After solving the shortest path problem, the solution obtained is a minimum-cost path that encodes the design decisions
(diameter and invert elevations) of the complete series. Since the solution is a path, the methodology selects a single
pipe (arc) for each link among all the possible alternatives, while verifying all the hydraulic constraints. The number
of evaluated alternatives depends on the quantity of available commercial diameters and the elevation change used
෩ ). For a vehicular road, if ߘ
(ߘ ෩ ൌ ͳܿ݉ there would be 380 possible elevations between the excavation limits, while
෩ ൌ ͳͲܿ݉ result on 38 possible elevations to evaluate. Therefore, there is a tradeoff between the precision of the
ߘ
design versus the computational effort to create the graph and solve the corresponding shortest path problem.

3. Numerical Examples

The methodology was evaluated in a large series of pipes and compared to an Artificial Intelligence (AI) based
program called Sedal Plus, which is an expert system for the design of all types of sewers (sanitary, storm and
combined) based on uniform flow, the rational method or the EPA hydrographs ( SWMM model) [10]. The proposed
methodology is coded in Java and compiled using Eclipse SDK version 4.6. The experiments are performed on a
computer with an Intel Core i7-6700T CPU @2.8GHz with 16GB of RAM allocated to the memory heap size of the
Java Virtual Machine on Windows 7. Computational time to solve the design problem measures all the steps of the
proposed methodology, and this performance metric is reported individually for each instance.

The series of pipes evaluated is located in Bogotá, Colombia, and belongs to the sewer network of the sector known
as Chicó Norte. It starts near the 7th Avenue and drains to a collector in Diagonal 92 as shown in Fig 2. It has 28 pipes
29 Manholes, a steep ground and pipe lengths between 68 m y 168 m. Table 3 and Table 4 present the input data for
the ground elevation at each manhole and the design flow rate and length for each pipe, respectively.
66 Natalia Duque et al. / Procedia Engineering 186 (2017) 61 – 68

Fig 2. Location and topology of the series of 28 pipes.

Table 3. Manholes input data.


Ground Ground Ground
Manhole elevation Manhole elevation Manhole elevation
(m) (m) (m)
0 2575.89 10 2556.9 20 2553.85
1 2573 11 2556.72 21 2553.74
2 2568.05 12 2556.62 22 2553.56
3 2565.71 13 2555.87 23 2554.13
4 2562.88 14 2555.78 24 2553.36
5 2560.93 15 2555.37 25 2553.05
6 2559.41 16 2554.93 26 2551.63
7 2558.7 17 2554.69 27 2551.46
8 2557.93 18 2554.59 28 2551.85
9 2557.33 19 2554.29

Table 4. Pipes input data.


Design Design Design
Length Length Length
Pipe flow rate Pipe flow rate Pipe flow rate
3 3 3
(m /s) (m) (m /s) (m) (m /s) (m)
1 0.0251 167.64 11 1.0559 103.55 21 1.5521 76.18
2 0.0397 71.89 12 1.1021 70.34 22 1.5658 91.6
3 0.095 72.37 13 1.1627 71.89 23 1.5804 84.26
4 0.1349 74.8 14 1.2234 71.36 24 2.005 98.39
5 0.3949 68.1 15 1.2515 71.5 25 2.0196 83.6
6 0.4363 80.84 16 1.2975 71.18 26 2.0621 97
7 0.4777 75.4 17 1.315 116.04 27 2.5971 80.98
8 0.5261 78.89 18 1.3425 92.3 28 2.611 92.36
9 0.5567 106.6 19 1.3566 93.47
10 0.6008 99.82 20 1.47 92.74
Table 5 summarizes the results obtained by using the proposed methodology, and Table 6 shows the results obtain by
using Sedal Plus. In this work, the hydraulic design uses the Manning’s resistance equation, nonetheless, the
methodology could use any other resistance equation as Darcy-Weisbach’s equation. The following tables presents
the hydraulic design and total costs for a series of 28 concrete pipes (݊ ൌ ͲǤͲͳ͵) a small elevation change ߘ ෩ ൌ
ͳܿ݉and the following set of commercial diameters (in ݉): 0.2, 0.25, 0.3, 0.35, 0.38, 0.4, 0.45, 0.5, 0.53, 0.6, 0.7,
0.80, 0.9, 1, 1.05, 1.20, 1.35, 1.4, 1.5 ,1.6 ,1.8, 2, 2.2, 2.4.
Natalia Duque et al. / Procedia Engineering 186 (2017) 61 – 68 67

Table 5. Hydraulic design obtained using the proposed methodology.


Upstream invert elevation Downstream invert elevation Slope Diameter Cost
Pipe
(m) (m) (-) (m) (COP)
1 2574.49 2571.60 0.0173 0.200 $ 45,509,081
2 2571.60 2566.64 0.0689 0.200 $ 19,397,922
3 2566.64 2564.00 0.0364 0.250 $ 22,246,305
4 2564.00 2561.37 0.0352 0.300 $ 25,555,504
5 2561.37 2559.28 0.0306 0.450 $ 29,353,419
6 2559.28 2557.71 0.0195 0.500 $ 37,149,828
7 2557.71 2556.44 0.0169 0.530 $ 36,021,700
8 2556.44 2555.60 0.0106 0.600 $ 40,779,617
9 2555.60 2555.24 0.0034 0.700 $ 60,442,167
10 2555.24 2554.83 0.0041 0.700 $ 56,384,657
11 2554.83 2554.49 0.0033 0.900 $ 67,963,316
12 2554.49 2554.24 0.0036 0.900 $ 46,086,734
13 2554.24 2553.77 0.0065 0.900 $ 47,046,985
14 2553.77 2553.30 0.0065 0.900 $ 46,749,074
15 2553.30 2552.81 0.0069 0.900 $ 47,076,468
16 2552.81 2552.61 0.0028 1.000 $ 49,802,525
17 2552.61 2552.28 0.0029 1.000 $ 81,673,320
18 2552.28 2552.00 0.0030 1.000 $ 64,940,005
19 2552.00 2551.71 0.0031 1.000 $ 65,915,262
20 2551.71 2551.38 0.0036 1.000 $ 65,295,023
21 2551.38 2550.92 0.0060 1.000 $ 53,611,051
22 2550.92 2550.35 0.0062 1.000 $ 65,291,249
23 2550.35 2549.82 0.0063 1.000 $ 61,122,515
24 2549.82 2548.84 0.0100 1.000 $ 72,993,024
25 2548.84 2547.99 0.0102 1.000 $ 62,334,597
26 2547.99 2546.96 0.0106 1.000 $ 72,856,537
27 2546.96 2545.84 0.0138 1.000 $ 60,912,659
28 2545.84 2544.56 0.0139 1.000 $ 72,505,699
Total Cost $ 1,477,016,242

Table 6. Hydraulic design obtained using Sedal Plus.


Upstream invert elevation Downstream invert elevation Slope Diameter Cost
Pipe
(m) (m) (-) (m) (COP)
1 2574.49 2571.40 0.0184 0.200 $ 45,608,033
2 2571.38 2566.45 0.0686 0.200 $ 19,465,669
3 2566.35 2564.01 0.0323 0.300 $ 24,757,793
4 2563.99 2561.18 0.0376 0.300 $ 25,599,450
5 2561.03 2559.08 0.0286 0.450 $ 29,490,130
6 2559.03 2557.47 0.0193 0.500 $ 37,319,615
7 2557.37 2556.69 0.0091 0.600 $ 38,751,650
8 2556.67 2555.80 0.0110 0.600 $ 40,604,240
9 2555.70 2555.12 0.0055 0.700 $ 60,456,115
10 2555.02 2554.70 0.0032 0.800 $ 61,234,160
11 2554.60 2554.08 0.0050 0.900 $ 68,498,663
12 2554.06 2553.67 0.0055 0.900 $ 46,605,196
13 2553.65 2553.21 0.0061 0.900 $ 47,658,730
14 2553.11 2552.88 0.0033 1.000 $ 50,388,874
15 2552.86 2552.62 0.0034 1.000 $ 50,489,761
16 2552.60 2552.34 0.0037 1.000 $ 50,081,521
17 2552.32 2551.88 0.0038 1.000 $ 82,432,617
18 2551.86 2551.43 0.0047 1.000 $ 65,764,866
19 2551.41 2550.96 0.0048 1.000 $ 67,069,254
20 2550.94 2550.42 0.0056 1.000 $ 66,774,498
21 2550.40 2549.92 0.0063 1.000 $ 54,941,522
22 2549.90 2549.31 0.0064 1.000 $ 67,112,368
23 2549.29 2548.74 0.0065 1.000 $ 62,915,858
24 2548.54 2548.21 0.0033 1.200 $ 84,145,530
25 2548.21 2547.93 0.0034 1.200 $ 70,517,539
26 2547.93 2547.59 0.0035 1.200 $ 80,923,878
27 2547.57 2547.04 0.0066 1.200 $ 66,488,365
68 Natalia Duque et al. / Procedia Engineering 186 (2017) 61 – 68

Upstream invert elevation Downstream invert elevation Slope Diameter Cost


Pipe
(m) (m) (-) (m) (COP)
28 2546.24 2546.19 0.0005 2.000 $ 110,081,844
Total Cost $ 1,576,177,737

Costs achieved using the proposed methodology for this series of pipes is 6% cheaper than the one obtained by the
expert system Sedal Plus. Moreover, this gap increases proportional to the number of pipes in the series due to the
exponential growth of the number of alternatives that must be evaluated and that the proposed methodology selects
the minimum-cost solution among all of them.

4. Conclusions

The proposed graph modeling framework seeks to obtain the minimum-cost design of a series of pipes as a shortest
path problem. The graph model allows the representation of every possible diameter - slope combination per section,
and the resulting path will encode the hydraulic design of each pipe in the series. Also, discrete commercial diameters
can be used due to the structure of the graph avoiding the need to round the continuous diameters, which can overstate
the total costs. Since the Bellman-Ford algorithm is an exact method and implicitly explores all the possible
alternatives, this methodology ensures the global optimal solution from an economic point of view without making
any simplifications of the problem’s constraints. This methodology obtains the optimal solution in a short
computational time using a standard desktop computer. The computational time is affected by the number of
commercial diameters evaluated and the elevation change ߘ ෩ . A smaller value for the elevation change ߘ
෩ will find a
lower total construction costs but will spend more computational time.

References

[1] N. Duque, D. Duque, & J. Saldarriaga “A new methodology for the optimal design of series of pipes on sewer systems”. Journal of
Hydroinformatics, 2016. DOI: 10.2166/hydro.2016.105.
[2] R. K. Ahuja, T. L. Magnati & J. B. Orlin Network flows: theory, algorithms, and applications. Prentice-Hall. New Jersey, 1993.
[3] R. Bellman “On a routing problem”. Quarterly of Applied Mathematics, Vol. 16 No.1, 1958, pp 87-90.
[4] A. Haith “Vertical alignment of sewer and drainage systems by dynamic programming”. Master thesis, Massachusetts Institute of Technology,
Boston, Massachusetts, USA, 1966.
[5] V.S. Kulkarni & P. Khanna “Pumped wastewater collection systems optimization”. Journal of Environmental Engineering, Vol. 111 No. 5,
1985, pp 589–601.
[6] G. Li & R. Matthew “New approach for optimization of urban drainage systems”. Journal of Environmental Engineering, Vol. 116 No. 5, 1990,
pp 927–944.
[7] P. K. Swamee & A. K. Sharma “Optimal design of a sewer line using linear programming”. Applied Mathematical Modelling, Vol. 37 No.6,
2013, pp 4430-4439.
[8] I. Navarro “Optimized network of urban drainage design”. Master thesis, Universidad de los Andes, Bogotá, Colombia., 2009.
[9] D. Butler & J. W. Davies “Urban Drainage”. 3th Edition. Spon Press, New York, USA, 2011.
[10] R. Paredes Sedal Plus V16.60 [online] Available at: http://sedalplus.com/ [Accessed 15 Jun. 2016].

You might also like