97
CHAPTER 5
DESIGN OF A RETRACTABLE ROOF STRUCTURE
5.1 General
The objective of this chapter is to perform the optimal design of a retractable roof structure
using the validated genetic algorithm. A predefined configuration of the movable panels
and the supporting structure is generated using a C-program developed as part of the
present research work. The structural design of retractable roof is carried-out against load
combination due to dead loads, live loads and wind loads. While carrying-out analysis
against wind, the permeability of roof structure is considered with roof full open, half open
and full closed conditions. Limit state method as per IS: 800-2007 is adopted for structural
steel design. The objective function minimizes the weight subjected to the constraints
imposed by strength and serviceability requirements. The problem formulation and the
step-by-step procedure for optimal design using real coded genetic algorithm with hybrid
mutation operator is explained in chapter 3.
5.2 Design of Movable Panels
The mechanism of movable roof system is attained by the sequence as detailed under. As
shown in Figure 3.4, panel B, panel C, panel E and panel F are the movable panels of the
retractable roof structure. The panels C and B move on the main arches whereas F and E
move on the quarter arches.
5.2.1 Panel C
The configuration of the retractable roof movable panel C generated by using a program
written in ‘C’ programming language is shown in Figure 5.1.
98
Problem statement of retractable roof movable panel C is as follows:
Number of members: 365, Number of joints: 172
The material properties: Elastic modulus, E= 2x105MPa, ϒ=2767.99 kg/m
3.
Allowable displacement: 50 mm. No. of member groups for each movable panel – 8 nos.
The design is as per IS: 800-2007.
The discrete design variable matrix - (the cross-sectional areas of members in sq. mm) as
follows:
[121 155 181 184 202 241 258 311 331 377 400 414 460 488 523 563 582 650
744 789 831 874 1010 1080 1120 1240 1270 1280 1470 1550 1730 1850 1860
1910 2050 2060 2090 2250 2270 2280 2320 2440 2490 2710 2760 2870 3190
3210 3480 3760 3950 4420 4950 6340 8650]
Loads on the movable panel C:
1. Dead load: Self weight of the members.
2. Live load: Joint load of 10 KN per each joint of movable panel.
3. Wind load: A horizontal point load of 20KN and a vertical point load of 15KN are
applied at each joint of the movable panel.
Figure 5.1 configuration of movable panel C
X Y Z
99
Optimal Design of panel C:
Optimal design of movable panel C is carried out using a population size of 20, 30 and 40
respectively. The results showing the optimal design for these population sizes are listed in
the Table 5.1 along with the corresponding design variables.
From the Table 5.1, the optimal weight of panel C for population size of 20, 30 and 40 is
739.16KN, 713.93KN and 655.77KN respectively. For calculating the loads on retractable
roof supporting structure the most optimum weight of 655.77KN is considered.
Table 5.1 Results of movable panel C
Design Variables
Population
Size(20)
(mm2)
Population
Size(30)
(mm2)
Population
Size(40)
(mm2)
A1 331.0 331.0 331.0
A2 331.0 331.0 331.0
A3 744.0 582.0 744.0
A4 331.0 331.0 331.0
A5 874.0 874.0 331.0
A6 582.0 582.0 582.0
A7 582.0 582.0 582.0
A8 241.0 241.0 241.0
Optimum Weight(KN): 739.16 713.93 655.77
The tabulated results are arrived at by using the two stage design procedure explained in
the section 3.9 of chapter 3. In the first stage the variation between the % of weak
individuals and weight of the truss is plotted by keeping the crossover probability constant
at 0.45. The second stage variation between the probabilities of crossover and weight of the
truss is plotted by keeping the % of weak individuals obtained from the first stage. Graphs
showing the variation of fitness with change in % of weak individuals and with change in
probability of crossover are presented in Figure 5.2 through Figure 5.4.
100
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 20 gives that 20% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.5 as shown in Figure 5.2.
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.2: Movable panel C design problem with population size – 20
101
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 30 gives that 20% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.55 as shown in Figure 5.3.
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.3: Movable panel C design problem with population size – 30
102
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 40 gives that 20% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.6 as shown in Figure 5.4.
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.4: Movable panel C design problem with population size – 40
103
5.2.2 Panel B
The configuration of the retractable roof movable panel B generated by using a program
written in ‘C’ programming language is shown in Figure 5.5.
Problem statement of retractable roof movable panel B is as follows:
Number of members: 283
Number of joints: 134
The material properties: Elastic modulus, E= 2x105MPa, ϒ=2767.99 kg/m
3.
Allowable displacement: 50 mm. No. of member groups for each movable panel – 8 nos.
The design is as per IS: 800-2007.
The discrete design variable matrix - (the cross-sectional areas of members in sq. mm) as
follows:
[121 155 181 184 202 241 258 311 331 377 400 414 460 488 523 563 582 650
744 789 831 874 1010 1080 1120 1240 1270 1280 1470 1550 1730 1850 1860
1910 2050 2060 2090 2250 2270 2280 2320 2440 2490 2710 2760 2870 3190
3210 3480 3760 3950 4420 4950 6340 8650]
Loads on the movable panel B:
1. Dead load: Self weight of the members.
2. Live load: Joint load of 10 KN per each joint of movable panel.
3. Wind load: A horizontal point load of 20KN and a vertical point load of 15KN are
applied at each joint of the movable panel.
104
Figure 5.5 configuration of movable panel B
Optimal Design of Panel B:
Optimal design of movable panel B is carried out using a population size of 20, 30 and 40
respectively. The results for these population sizes are listed in the Table 5.2 along with the
corresponding design variables. The results of the Table 5.2 are arrived at by using the two
stage design procedure explained in the section 3.9 of chapter 3. From the Table 5.2 the
optimum weight of panel B for population size of 20, 30 and 40 is 659.87KN, 650.25KN
and 638.87KN respectively. For calculating the loads on retractable roof supporting
structure the most optimum weight of 638.39KN is considered.
Table 5.2 Results of movable panel B
Design Variables
Population
Size(20)
(mm2)
Population
Size(30)
(mm2)
Population
Size(40)
(mm2)
A1 523.0 523.0 523.0
A2 400.0 400.0 400.0
A3 202.0 202.0 202.0
A4 202.0 202.0 202.0
A5 311.0 311.0 311.0
A6 2270.0 2270.0 2270.0
A7 414.0 311.0 184.0
A8 414.0 414.0 414.0
Optimum Weight(KN): 659.87 650.25 638.39
X Y Z
105
Graphs showing the variation of fitness with change in % of weak individuals and with
change in probability of crossover are presented in Figure5.6 through Figure 5.8.
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.6: Movable panel B design problem with population size – 20
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 20 gives that 20% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.55 as shown in Figure 5.6.
106
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.7: Movable panel B design problem with population size – 30
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 30 gives that 20% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.55 as shown in Figure 5.7.
107
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.8: Movable panel B design problem with population size – 40
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 40 gives that 15% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.6 as shown in Figure 5.8.
108
5.2.3 Panel F
The configuration of the retractable roof movable panel C generated by using a program
written in ‘C’ programming language is shown in Figure 5.9.
Problem statement of retractable roof movable panel F is as follows:
Number of members: 283
Number of joints: 134
The material properties: Elastic modulus, E= 2x105MPa, ϒ=2767.99 kg/m
3.
Allowable displacement: 50 mm. No. of member groups for each movable panel – 8 nos.
The design is as per IS: 800-2007.
The discrete design variable matrix - (the cross-sectional areas of members in sq. mm) as
follows:
[121 155 181 184 202 241 258 311 331 377 400 414 460 488 523 563 582 650
744 789 831 874 1010 1080 1120 1240 1270 1280 1470 1550 1730 1850 1860
1910 2050 2060 2090 2250 2270 2280 2320 2440 2490 2710 2760 2870 3190
3210 3480 3760 3950 4420 4950 6340 8650]
Loads on the movable panel F:
1. Dead load: Self weight of the members.
2. Live load: Joint load of 10 KN per each joint of movable panel.
3. Wind load: A horizontal point load of 20KN and a vertical point load of 15KN are
applied at each joint of the movable panel.
109
Figure 5.9 configuration of movable panel F
Optimal Design of Panel F:
Optimal design of movable panel F is carried out using a population size of 20, 30 and 40
respectively. The results for these population sizes are listed in the Table 5.3 along with the
corresponding design variables. The results of the Table 5.3 are arrived at by using the two
stage design procedure explained in the section 3.9 of chapter 3.
The optimum weight of panel F for population size of 20, 30 and 40 are 718.78KN,
715.20KN and 712.79KN respectively. For calculating the loads on retractable roof
supporting structure the most optimum weight of 712.79 is considered.
Table 5.3 Results of movable panel F
Design Variables
Population
Size(20)
(mm2)
Population
Size(30)
(mm2)
Population
Size(40)
(mm2)
A1 184.0 184.0 184.0
A2 311.0 377.0 377.0
A3 1240.0 1240.0 1240.0
A4 311.0 311.0 311.0
A5 331.0 331.0 155.0
A6 488.0 488.0 488.0
A7 1270.0 1270.0 1270.0
A8 400.0 331.0 400.0
Optimum Weight(KN): 718.78 715.20 712.79
X Y Z
110
Graphs showing the variation of fitness with change in % of weak individuals and with
change in probability of crossover are presented in Figure 5.10 through Figure 5.12.
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.10: Movable panel F design problem with population size – 20
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 20 gives that 20% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.5 as shown in Figure 5.10.
111
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.11: Movable panel F design problem with population size – 30
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 30 gives that 15% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.55 as shown in Figure 5.11.
112
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.12: Movable panel F design problem with population size – 40
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 40 gives that 20% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.55 as shown in Figure 5.12.
113
5.2.4 Panel E
The configuration of the retractable roof movable panel C generated by using a program
written in ‘C’ programming language is shown in Figure 5.13. Problem statement of
retractable roof movable panel E is as follows:
Number of members: 283, Number of joints: 134
The material properties: Elastic modulus, E= 2x105MPa, ϒ=2767.99 kg/m
3.
Allowable displacement: 50 mm. No. of member groups for each movable panel – 8 nos.
The design is as per IS: 800-2007.
The discrete design variable matrix - (the cross-sectional areas of members in sq. mm) as
follows:
[121 155 181 184 202 241 258 311 331 377 400 414 460 488 523 563 582 650
744 789 831 874 1010 1080 1120 1240 1270 1280 1470 1550 1730 1850 1860
1910 2050 2060 2090 2250 2270 2280 2320 2440 2490 2710 2760 2870 3190
3210 3480 3760 3950 4420 4950 6340 8650]
Loads on the movable panel E:
1. Dead load: Self weight of the members.
2. Live load: Joint load of 10 KN per each joint of movable panel.
3. Wind load: A horizontal point load of 20KN and a vertical point load of 15KN are
applied at each joint of the movable panel.
114
Figure 5.13 configuration of movable panel E
Optimal Design of Panel E:
Optimal design of movable panel E is carried out using a population size of 20, 30 and 40
respectively. The results for these population sizes are listed in the Table 5.4 along with the
corresponding design variables. The results of the Table 5.4 are arrived at by using the two
stage design procedure explained in the section 3.9 of chapter 3. The optimum weight of
panel E for population size of 20, 30 and 40 are 699.31KN, 647.84KN and 636.51KN
respectively. For calculating the loads on retractable roof supporting structure the most
optimum weight of 636.51KN is considered.
Table 5.4 Results of movable panel E
Design Variables
Population
Size(20)
(mm2)
Population
Size(30)
(mm2)
Population
Size(40)
(mm2)
A1 184.0 523.0 184.0
A2 1240.0 650.0 1240.0
A3 582.0 414.0 241.0
A4 155.0 155.0 155.0
A5 377.0 377.0 377.0
A6 488.0 563.0 488.0
A7 1270.0 1270.0 1270.0
A8 400.0 400.0 400.0
Optimum Weight(KN): 699.31 647.84 636.51
X Y Z
115
Graphs showing the variation of fitness with change in % of weak individuals and with
change in probability of crossover are presented in Figure5.14 through Figure 5.16.
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.14: Movable panel E design problem with population size – 20
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 20 gives that 15% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.5 as shown in Figure 5.14.
116
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.15: Movable panel E design problem with population size – 30
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 30 gives that 15% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.6 as shown in Figure 5.15.
117
Figure a: Variation of fitness with change in % of weak individuals
Figure b: Variation of fitness with change in probability of crossover
Figure 5.16: Movable panel E design problem with population size – 40
Execution of the genetic algorithm, with varied % of weak individuals for population size
of 40 gives that 15% of weak individuals can be allowed to get optimum fitness at
crossover probability of 0.65 as shown in Figure 5.16.
118
5.3 Design of supporting structure
The configuration of the retractable roof supporting structure generated by using a program
written in ‘C’ programming language is shown in Figure 5.17.
Problem statement of retractable roof structure is as follows:
Stadium diameter: 100 m., Height: 20 m., No. of members in retractable supporting
structure: 6628, No. of joints in retractable supporting structure: 1968, No. of degrees of
freedom of retractable supporting structure: 11448. The design is as per IS: 800-2007.
The material properties: Elastic modulus, E= 2x105MPa, ϒ=2767.99 kg/m
3.
Allowable displacement: 50 mm.
No. of member groups in the retractable roof supporting structure – 26 nos.
Figure: 5.17 Configuration of retractable roof supporting structure
The discrete design variable matrix - (the cross-sectional areas of members in sq. mm) as
follows:
[121 155 181 184 202 241 258 311 331 377 400 414 460 488 523 563 582 650
744 789 831 874 1010 1080 1120 1240 1270 1280 1470 1550 1730 1850 1860
1910 2050 2060 2090 2250 2270 2280 2320 2440 2490 2710 2760 2870 3190
3210 3480 3760 3950 4420 4950 6340 8650]
119
Loads on the supporting structure:
1. Dead load: Self weight of the members.
2. Live load: (a). Wheel loads from the movable panels on the respective joints.
Wheel load of Panel C: 281 KN (vertical) and 172KN (horizontal), Wheel load of
Panel B: 289 KN (vertical) and 167.5KN (horizontal), Wheel load of Panel F: 298.5
KN (vertical) and 167.5KN (horizontal), Wheel load of Panel E: 289 KN (vertical)
and 167.5KN (horizontal) (b). Joint load of 10 KN per joint at other joints.
3. Wind load: A horizontal point load of 20KN and a vertical point load of 15KN are
applied at each joint of upper surface.
Table 5.5: Optimal design of supporting structure
Design Variable Roof full open
(mm2)
Roof full closed
(mm2)
Roof half open
(mm2)
A1 831 460 460
A2 3950 2760 2760
A3 377 2050 874
A4 1240 181 181
A5 874 874 874
A6 258 582 582
A7 831 2090 2090
A8 3260 241 241
A9 2870 1850 3210
A10 311 1010 414
A11 1280 1850 1470
A12 3760 121 121
A13 789 3260 3260
A14 2250 4950 4950
A15 3480 414 414
A16 582 1850 1730
A17 2050 4420 181
A18 1910 202 202
A19 400 2060 2060
A20 2320 563 563
A21 2760 2870 2870
A22 789 650 650
A23 3950 1270 1270
A24 241 1550 1550
A25 563 2440 2440
A26 582 563 563
Weight(KN): 50937.78 57762.27 53299.88
120
Optimal design of supporting structure:
Optimal design of supporting structure for both movable and immovable panels and the
main frame work on which the panels move or rest are designed using the genetic
algorithm. Optimal design of the retractable roof supporting structure is carried out with
the roof in full open position, half open position and full closed position. The genetic
algorithm is executed according to the two stage process adopted ealier by considering
population size as forty for all the above cases of roof positions. In the first stage the
variation between the % of weak individuals and weight of the truss is plotted by keeping
the crossover probability constant at 0.45. The second stage variation between the
probabilities of crossover and weight of the truss is plotted by keeping the % of weak
individuals obtained from the first stage.
For the roof full open position of the supporting structure, execution of the genetic
algorithm with varied % of weak individuals for population size of forty gives that 15% of
weak individuals can be allowed to get optimum fitness at crossover probability of 0.45 as
shown in Figure 5.18.
Figure a: Variation of fitness with change in % of weak individuals
121
Figure b: Variation of fitness with change in probability of crossover
Figure 5.18: Retractable roof full open position with population size – 40
For the roof full closed position of the supporting structure, execution of the genetic
algorithm with varied % of weak individuals for population size of forty gives that 15% of
weak individuals can be allowed to get optimum fitness at crossover probability of 0.45 as
shown in Figure 5.19.
Figure a: Variation of fitness with change in % of weak individuals
122
Figure b: Variation of fitness with change in probability of crossover
Figure 5.19: Retractable roof full closed position with population size – 40
For the roof half open position of the supporting structure, execution of the genetic
algorithm with varied % of weak individuals for population size of 40 gives that 20% of
weak individuals can be allowed to get optimum fitness at crossover probability of 0.45 as
shown in Figure 5.20.
Figure a: Variation of fitness with change in % of weak individuals
123
Figure b: Variation of fitness with change in probability of crossover
Figure 5.20: Retractable roof half open position with population size – 40
5.4 Optimal design of retractable roof structure
Optimal design of the retractable roof movable panels and the supporting structure is
carried out using real coded genetic algorithm with hybrid mutation operator. Table: 5.1
through Table 5.4 shows the results of optimum design of movable panels. The optimum
weight of movable panels C and B which move on the main arches is 655.77 KN and
638.39 KN respectively. The optimum weight of movable panels F and E which move on
the quarter arches is 712.79 KN and 636.51 KN respectively. These optimal weights of
movable panels are obtained when the genetic algorithm is executed with a population size
of 40. Therefore, the optimal weight of retractable roof (movable panels) is 7985.52 KN.
Table: 5.5 shows the results of optimum design of retractable roof supporting structure
when the roof is full open, full closed and half open conditions. The optimum weight of
retractable roof supporting structure when the roof is in the above three positions is
50937.78 KN, 57762.27 KN and 53299.88 KN respectively.