Automatic Piping Arrangement Design

Considering Piping Supports and Curved

Surfaces of Building Blocks

Hajime Kimura
Dep. of Marine Systems Engineering,
Kyushu University
CL7: International Environment System Engineering VII
 Motivations 3D-CAD systems reduce human efforts,

However, piping arrangement design

processes still depend on human power.

To react to the design change of another section,

modification or re-starting of the piping arrangement is too often.

Automatic Piping arrangement Design system It is torture for

the designers!
will save more time and money!
 Conventional Regular-Grid Approaches
The target design space is divided into regular grids
where the mesh size is lager than the pipe’s diameter.
The grid world is translated into a weighted graph. Obstacle
looking on the pipe arrangement problem as a
routing problem in the weighted graph
Obstacle
solved by “Dijkstra’s method”

Disadvantages are …
× The mesh size is restricted to be larger than
pipe’s diameters
× Directions of the pipes are restricted along XYZ axes.

Although the gap between

the obstacle is larger than Obstacle Goal
the diameter of the pipe, …

… conventional
methods cannot put
pipes in this case
Start Obstacle
Regular grid methods cannot consider in the following situations:

Ship Hull

Pipe
Pipes are often arranged
along the curved ship’s hull.

Pipes are arranged so as to make use of

the same supports in the different pipes.
Preparation to overcome the defects of Regular Grid Methods
Geometrically, all the pipes
with two bends can connect
arbitrary two design points
which have positions and
C
directions!
A

D B

【Example 2】
The
【Example 1】 condition：
The condition: Using only 90
AC:CD:DB＝１：２：１ degree elbows
Constraints of the real pipe’s shapes

Pipe Bender：
(1) It bends pipes with arbitrary angles,
but there exists a limit of a maximum
angle.

HYPREX: YPB-704F PIPE BENDER (2) The radius of the bend is

constrained by the pipe bender’s molds.

(3) When it bends pipes, straight parts

These troublesome constraints between the bend are required to
prevents automatic piping design! grasp the pipe. 8
A new approach to automatic design of piping

In actual piping design, the pipes are put

If the pipes are put in the middle
the position that is certain distance away
of the target design space, it is
from structures. burdensome to arrange piping
supports!

１） At all the position of the supports, put design

points which indicate position and direction of the
way of the pipe for each pipe.

２） Make a weighted graph checking all the pair of

design points which can connect by a pipe with two
bends that can make using the pipe bender.

３） Generate piping path using a shortest path

search algorithm in the weighted graph.
The new approach: (3) Generate piping path
Goal Goal

Start Start

Design Points which Pipe

have local coordinates Arrangement

Design Points are connected by pipes which have

two bends along the axes of the local coordinates.

Shortest path Search

Weighted Graph In the graphical network
 Dijkstra’s Method

This method can …

 find the shortest path in a directed and weighted graph
 guarantee to find the shortest path with minimum costs 𝟐

Where is the path with minimum costs The answer is …

between C1 and C2?
C4 C4 minimum costs
7 7
C2 7 C2 7
3 3 16
5 6 5 6
C1 3 C5 5 C7 C1 3 C5 5 C7
10 7 10 7
C3 C3
8 8
8 8
C6 C6
Water flow can solve mazes!

Water flow ＝ A path from the start to the goal Start

Water do not move on the

way to dead ends.

Goal
 Start

Goal
Electric current flow can
also solve mazes!
Start:
Current exists only on the way to the goal. Connect the + Pole

No current on the way to

the dead ends.

Goal: Connect the - pole

Shortest Path Search using SciPy
SciPy is a Python-based ecosystem of open-source software
for mathematics, science, and engineering.

EXAMPLE:

Find the shortest path (having minimum cost) from C0 to C11 in the following graph:

C3 15 C6
7 10
7 5
C1 3 C9
5 6 10
7
C0 3 5 C7 15 4 C11
10 C4 8
7 10 6
C2 5 C10
8 10
8
C5 10 C8
 A program implementation
import numpy as np
for finding the shortest
from scipy.sparse.csgraph import shortest_path
path by Python
To Node j
#To: 0 1 2 3 4 5 6 7 8 9 10 11
graph = [[ 0, 5,10, 0, 0, 0, 0, 0, 0, 0, 0, 0], Adjacency matrix of the
[ 5, 0, 3, 7, 6, 0, 0, 0, 0, 0, 0, 0], directed graph of the example
[10, 3, 0, 0, 7, 8, 0, 0, 0, 0, 0, 0], shown in the previous slide
[ 0, 7, 0, 0, 3, 0,15, 7, 0, 0, 0, 0], Cost = 5 from Node 0 to 1
[ 0, 6, 7, 3, 0, 0, 0, 5,10, 0, 0, 0],
[ 0, 0, 8, 0, 0, 0, 0, 8,10, 0, 0, 0], Cost = 3 from Node 1 to 2
From Node i …
[ 0, 0, 0,15, 0, 0, 0, 5, 0,10,15, 0],
[ 0, 0, 0, 7, 5, 8, 5, 0, 5, 7, 8, 0],
[ 0, 0, 0, 0,10,10, 0, 5, 0, 0,10, 0],
[ 0, 0, 0, 0, 0, 0,10, 7, 0, 0, 4,10],
[ 0, 0, 0, 0, 0, 0,15, 8,10, 4, 0, 6],
[ 0, 0, 0, 0, 0, 0, 0, 0, 0,10, 6, 0]] Solver
a = np.array( graph )
(cost, path) = shortest_path( a, return_predecessors=True)
print("cost matrix-------------")
print( cost )
print("predecessors------------")
print( path ) 20
>python graph.py
cost matrix-------------
Results
[[ 0. 5. 8. 12. 11. 16. 21. 16. 21. 23. 24. 30.]
[ 5. 0. 3. 7. 6. 11. 16. 11. 16. 18. 19. 25.] Costs of the shortest path
[ 8. 3. 0. 10. 7. 8. 17. 12. 17. 19. 20. 26.] from Node i to Node J
[12. 7. 10. 0. 3. 15. 12. 7. 12. 14. 15. 21.]
[11. 6. 7. 3. 0. 13. 10. 5. 10. 12. 13. 19.]
[16. 11. 8. 15. 13. 0. 13. 8. 10. 15. 16. 22.] Node matrix immediately
before j on the shortest path
[21. 16. 17. 12. 10. 13. 0. 5. 10. 10. 13. 19.]
starting from node i to node j
[16. 11. 12. 7. 5. 8. 5. 0. 5. 7. 8. 14.]
[21. 16. 17. 12. 10. 10. 10. 5. 0. 12. 10. 16.]
[23. 18. 19. 14. 12. 15. 10. 7. 12. 0. 4. 10.]
The shortest path
[24. 19. 20. 15. 13. 16. 13. 8. 10. 4. 0. 6.]
From Node 0 to 11
[30. 25. 26. 21. 19. 22. 19. 14. 16. 10. 6. 0.]]
predecessors------------
[[-9999 0 1 1 1 2 7 4 4 7 7 10]
[ 1 -9999 1 1 1 2 7 4 4 7 7 10]
[ 1 2 -9999 1 2 2 7 4 4 7 7 10] There are
[ 1 3 1 -9999 3 7 7 3 7 7 7 10] two routes
leading to
[ 1 4 4 4 -9999 7 7 4 4 7 7 10] node 8, 4
[ 1 2 5 7 7 -9999 7 5 5 7 7 10] and 7, but
[ 1 4 4 7 7 7 -9999 6 7 6 7 10] only 4 is
displayed.
[ 1 4 4 7 7 7 7 -9999 7 7 7 10]
[ 1 4 4 7 8 8 7 8 -9999 7 8 10]
[ 1 4 4 7 7 7 9 9 7 -9999 9 9]
[ 1 4 4 7 7 7 7 10 10 10 -9999 10]
[ 1 4 4 7 7 7 7 10 10 11 11 -9999]] 21
Correct Answer 12 21
C3 15 C6
7 10
５ 23
7 5
C1 3 C9
5 6 10
16 7
C0 3 11 5 C7 15 4 C11
10 C4 8
8 7 10 6
C2 5 C10
8 10 24
16 8 21
C5 10 C8

Solving a simple maze in the same way …

Multiple Pipe Routing

 From the largest diameter to the smallest diameter

 Random choice from pipes with the same diameters

Random choice Random choice

 Order of Routing

Order of routing in real arrangement = From the largest to the smallest

G G
Obstacle

Pipe B

S G

Pipe C

Obstacle
S S
 Order of Routing

Order of routing in real arrangement = From the largest to the smallest

G G
Obstacle

Pipe B

S Pipe A G

Pipe C

Obstacle
S S
 Order of Routing

Order of routing in real arrangement = From the largest to the smallest

G G
Obstacle

Pipe B

S Pipe A G

Pipe C

Obstacle
S S
 Order of Routing

Order of routing in real arrangement = From the largest to the smallest

G G
Obstacle

Pipe B

S Pipe A G

Pipe C

Obstacle
S S
Largely different solutions depending on the order of arranging pipes

ΦA = Φ B

30
 Path selection of one pipe

Both paths are

optimum for the pipe
B in this stage!

ΦA < Φ B

Optimum Paths for one pipe are multiple,

But we cannot find the best for the other pipes．
Interference avoidance of multiple pipes

It is for a wire routing method

Touch and Cross Method: for electronic circuits.

First, the system arranges the

shortest piping paths
without considering interferences.
Approach: Interference avoidance of multiple pipes

It is for a wire routing method

Touch and Cross Method: for electronic circuits.

First, the system arranges the

shortest piping paths
without considering interference.

Then, the system can acquire

information about the crowding area.
Approach: Interference avoidance of multiple pipes

It is for a wire routing method

Touch and Cross Method: for electronic circuits.

Re-arrange pipes with the order

of the pipe’s length.
Approach: Interference avoidance of multiple pipes

It is for a wire routing method

Touch and Cross Method: for electronic circuits.

Re-arrange pipes with the order

of the pipe’s length.
Approach: Interference avoidance of multiple pipes

It is for a wire routing method

Touch and Cross Method: for electronic circuits.

Re-arrange pipes with the order

of the pipe’s length.
Touch and Cross Method for Branches
First, the system arranges
the shortest pipes without
considering interference.
Touch and Cross Method for Branches
First, the system arranges
the shortest pipes without
considering interference.

Re-arrange to overlap
with the same pipelines.
Touch and Cross Method for Branches
First, the system arranges Re-arrange to overlap
the shortest pipes without with the same pipelines.
considering interference.

Re-arrange to overlap
with the same pipelines.
Touch and Cross Method for Branches
First, the system arranges Re-arrange to overlap
the shortest pipes without with the same pipelines.
considering interference.

Re-arrange to overlap
with the same pipelines.

Re-arrange to overlap
with the same pipelines.
Touch and Cross Method for Branches
First, the system arranges Re-arrange to overlap
the shortest pipes without with the same pipelines.
considering interference.

Re-arrange to overlap
with the same pipelines.

Re-arrange to overlap
with the same pipelines.
Computer Simulations

【Experiment 1】 Arrangement along curved structures

Φ600 bend’s radius： 2400 or 1200
Φ400 bend’s radius： 1600 or 1200
Φ100 bend’s radius： 300 or 150
The minimum length of the straight pipe between
elbows： Over 200 in φ600 and 400 pipes and 100
in φ100 pipe.

The angle of the bend is 90 deg only in Φ100

The others are arbitrary (maximum is 90 deg)

Size of structures： half of a cylinder, radius 3000

with a plane which is the length 200

Arrangement of the design points：

interval 400 in the height direction,
In the direction of radius in the cylinder,
away 200 from the wall, interval 400,
in the direction to the circumference,
divide 180 deg. into 20, then 870 points
Almost the same solutions in all trials
45
It takes 2~3 minutes, sometimes failed.
 Results of Experiment (2)-1
COST=1370167

For Ship Design（Arbitrary angle elbows）

Φ200 bend’s radius： 400 or 300
Φ150 bend’s radius: 300 or 150 COST=1171743
Φ100 bend’s radius： 200 or 100

The minimum length of the straight pipe between elbows：

Over 100 in all diameters of pipes.

Maximum angle of the bend is 90 degrees

COST=1192755
Cost of the pipe per unit length:
Φ200：40
Φ150：22.5
Φ100：10

Cost of bends : 1000 constant that does not depend

on the diameter nor angles COST=1318435

In all trials, different Solutions are found. 52

 Results of Experiment (2)-2
COST=1368599

For Building Design （90 degree elbows）

Φ200 bend’s radius： 400 or 300
Φ150 bend’s radius: 300 or 150 COST=1320220
Φ100 bend’s radius： 200 or 100

The minimum length of the straight pipe between elbows：

Over 100 in all diameters of pipes.

The angle of the bends is only 90 degree.

COST=1281864
Cost of the pipe per unit length:
Φ200：40
Φ150：22.5
No detour
Φ100：10

Cost of bends : 1000 constant COST=1181384

that does not depend on the diameter nor angles

In all trials, different Solutions are found.

53
In 10 trials, failed 4 times （truncated 4 min.）
It can use the
same support

54
Discussion

The number of design points

In experiment 2, the system can find feasible solutions,
However, when the number of design points are
decrease, It could not find any solutions.
→ Appropriate arrangement of design points
is very important!
Future work

Different solutions in trials

Solutions are largely depend on the initial path
→ Improve touch and cross method so as to find best solutions certainly

The Solution in the constraint of 90 degree elbows:

Although the elbows are 90 deg, some pipes are arranged diagonally → Real piping design
56
 arranged diagonally

A solution in the constraint of 90 degree elbows

57
Piping with Branches
Design Points: 1288
Obstacle Boxes: 28
Computation time: 18min.

Branch

Branch

Branch

58
Path Planning of Multiple Flying Drones
Design Points: 2000
Obstacle Boxes: 7
Computation time: 25min.
Conclusion

 On the routing problem for one pipe, to deal with supports and
arrangement along curved structure, a new approach is proposed:
Design points that indicates direction and position are arranged in advance
as candidates of waypoints of the pipe, and generate weighted graph
checking whether two design points can connect by a pipe which satisfies
the constraint of the pipe bender. After that, generate piping path from
the shortest path search in the weighted graph.

Arranged along
the structures

 For multiple piping, Touch and cross method is also applied

to avoid different pipelines, and to overlap the same pipelines. 65
Remarks

This research was sponsored by the Japan Society for the

Promotion of Science (JSPS) Grants-in-Aid for Scientific Research
(B) 23360388.

This technology is Patent pending by Kyushu University.

(Japanese Patent Application No. 2017-96845)

If you want to get the source code of the programs,

Please contract with Kyushu university to use the license.