Giuseppe Carlo Marano, Cristoforo Demartino, Rita
Greco, Bruno Briseghella, Alessandra Fiore
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 1
Outline
Use Opensees for structural optimization
Develop a Matlab – OpenSees connection using GA
Deal with constraints in a «smart» way
Application to simple steel structure
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 2
Structural Optimization
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
( ){ }
( )
( )
min
0 1, ,
0 1, ,
,
q q
r r
l u
f
g q n
h r n
≤ =
= =
∈   
x
x
x
x x x


Paradigm
• Classical
• Non-classical
Evolutive algorhitms for structural
optimization
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
Artificial Neural
Networks (ANN)
Evolutive/Genetic
Algorithms (GA)
Soft computing
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
•One seeks the solution of a problem in the form of strings of numbers (traditionally binary,
although the best representations are usually those that reflect something about the problem
being solved), by applying operators such as recombination and mutation;
Genetic algorithm
•Here the solutions are in the form of computer programs, and their fitness is determined by
their ability to solve a computational problem.Genetic programming
•Similar to genetic programming, but the structure of the program is fixed and its numerical
parameters are allowed to evolve;
Evolutionary programming
•Works with vectors of real numbers as representations of solutions, and typically uses self-
adaptive mutation rates;Evolution strategy
•Based on vector differences and is therefore primarily suited for numerical optimization
problems.Differential evolution
•Based on the ideas of animal flocking behaviour. Also primarily suited for numerical
optimization problems.Particle swarm optimization
•Based on the ideas of ant foraging by pheromone communication to form paths. Primarily
suited for combinatorial optimization problems.Ant colony optimization
•Based on the ideas of weed colony behavior in searching and finding a suitable place for
growth and reproduction.
Invasive weed optimization
algorithm
•Based on the ideas of musicians' behavior in searching for better harmonies. This algorithm is
suitable for combinatorial optimization as well as parameter optimization.Harmony search
•Based on information theory. Used for maximization of manufacturing yield, mean fitness or
average information. See for instance Entropy in thermodynamics and information theory.Gaussian adaptation
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 7
1 1 2 2
, , , ,l u l u l u l u
j j n n
x x x x x x x x = ⊗ ⊗ ⊗ ⊆           Ω   
( ) ( ){ }1
max 0, 0
q rn n
k k
i p i
p
g
+
=
Φ ≥∑x x
( ){ }min f x
Structural optimization using soft computing
X1,x2,….Xn
Verification
( ) ( ){ }1
max 0, 0
q rn n
k k
i p i
p
g
+
=
Φ ≥∑x x
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
Constraint Violation
x1
feasible
x2
infeasible
( ) 0k
iΦ =x
( ) 0k
iΦ >x
Penalty functions-based methods
Methods based on special operators and representations
Methods based on repair algorithms
Methods based on the separation between OF and
constraints
Hybrid methods
Objective Function
Constraints’ violation
Both two feasible
Feasible and infeasible
Both two infeasible
denotes that (k-1)xi
Pb is dominated bykxi.
, ,
k k k
i i jk
b i j k k k
j j k
if
if

= 

x x x
x
x x x


( )
( ) ( )
( )( ) ( )( ) ( )
( )( )
( )( ) ( )
( )( )
( ) ( )
( )
1 1
1 1
1
0 0
0 0
k kk Pb k Pb
i i i i
kk Pb kk Pb
i i i i
kk Pb
i i
f f − −
− −
−
 < ∧ Φ = ∧ Φ =

 ∨

⇔ Φ = ∧ Φ >

∨

Φ < Φ
x x x x
x x x x
x x

x1
feasible
x2
infeasible
Global optimum
0k k Pb
µΦ= ≥Φ
( ) ( ) ( ){ }1
k Pb k Pb k Pb k Pb
i N=Φ Φ ΦΦ x x x 
( )
( ) ( )
( )( ) ( )( ) ( )
( )( )
( )( ) ( )
( )( )
( )( ) ( )
( )( ) ( ) ( )
( )( )
1 1
1 1
1 1
k kk Pb k k k Pb k k
i i i i
k kk Pb k k k Pb k k
i i i i
k kk k k Pb k k k Pb
i i i i
f f µ µ
µ µ
µ µ
− −
Φ Φ
− −
Φ Φ
− −
Φ Φ
 < ∧ Φ ≤ Ψ ∧ Φ ≤ Ψ

 ∨


⇔ Φ ≤ Ψ ∧ Φ > Ψ

∨

Φ > Ψ ∧ Φ > Ψ ∧ Φ < Φ
x x x x
x x x x
x x x x

( )
[ ] ( )01 0,1 0, , ;k k
k L
L
η
η + 
Ψ= − ∈ = ∈ 
 
 
x1
feasible
x2
infeasible
17
Application to static analysis of steel structure
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 18
3
33
A s
s
ε
−
≤
] ]
3
33,38
A s
s
ε
−
∈
] ]
3
38,42
A s
s
ε
−
∈
3
42
A s
s
ε
−
>
Section Class
•1
•2
•3
•4
235
yf
ε =
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 19
Buckling resistance of steel bars
,
1.0Ed
b Rd
N
N
≤
1
,
y
b Rd
M
Af
N χ
γ
=
1
,
eff y
b Rd
M
A f
N χ
γ
=
1-2-3 4
Structural optimization
More than minimizing
weight!
Technological/constructive
constraints
New class of “shapes”
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 21
Optimal topology (?)
Conclusions
Interface OpenSees with EA
• External Optimization algorithm
• External constraint evaluation
• Constrained handling
Futures
• Parallel computing
• General Constrained handling
• More complex problems
Porto, June 20 2017
OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR
STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 22
Conclusions

OpenSees solver with a differential evolutionary algorithm for structural optimization of hollow sections steel structures

  • 1.
    Giuseppe Carlo Marano,Cristoforo Demartino, Rita Greco, Bruno Briseghella, Alessandra Fiore Porto, June 20 2017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 1
  • 2.
    Outline Use Opensees forstructural optimization Develop a Matlab – OpenSees connection using GA Deal with constraints in a «smart» way Application to simple steel structure Porto, June 20 2017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 2
  • 3.
    Structural Optimization Porto, June20 2017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
  • 4.
    ( ){ } () ( ) min 0 1, , 0 1, , , q q r r l u f g q n h r n ≤ = = = ∈    x x x x x x   Paradigm • Classical • Non-classical Evolutive algorhitms for structural optimization Porto, June 20 2017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
  • 5.
    Artificial Neural Networks (ANN) Evolutive/Genetic Algorithms(GA) Soft computing Porto, June 20 2017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
  • 6.
    •One seeks thesolution of a problem in the form of strings of numbers (traditionally binary, although the best representations are usually those that reflect something about the problem being solved), by applying operators such as recombination and mutation; Genetic algorithm •Here the solutions are in the form of computer programs, and their fitness is determined by their ability to solve a computational problem.Genetic programming •Similar to genetic programming, but the structure of the program is fixed and its numerical parameters are allowed to evolve; Evolutionary programming •Works with vectors of real numbers as representations of solutions, and typically uses self- adaptive mutation rates;Evolution strategy •Based on vector differences and is therefore primarily suited for numerical optimization problems.Differential evolution •Based on the ideas of animal flocking behaviour. Also primarily suited for numerical optimization problems.Particle swarm optimization •Based on the ideas of ant foraging by pheromone communication to form paths. Primarily suited for combinatorial optimization problems.Ant colony optimization •Based on the ideas of weed colony behavior in searching and finding a suitable place for growth and reproduction. Invasive weed optimization algorithm •Based on the ideas of musicians' behavior in searching for better harmonies. This algorithm is suitable for combinatorial optimization as well as parameter optimization.Harmony search •Based on information theory. Used for maximization of manufacturing yield, mean fitness or average information. See for instance Entropy in thermodynamics and information theory.Gaussian adaptation
  • 7.
    Porto, June 202017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 7 1 1 2 2 , , , ,l u l u l u l u j j n n x x x x x x x x = ⊗ ⊗ ⊗ ⊆           Ω    ( ) ( ){ }1 max 0, 0 q rn n k k i p i p g + = Φ ≥∑x x ( ){ }min f x Structural optimization using soft computing
  • 8.
    X1,x2,….Xn Verification ( ) (){ }1 max 0, 0 q rn n k k i p i p g + = Φ ≥∑x x Porto, June 20 2017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES
  • 9.
  • 10.
    Penalty functions-based methods Methodsbased on special operators and representations Methods based on repair algorithms Methods based on the separation between OF and constraints Hybrid methods
  • 11.
  • 12.
    Both two feasible Feasibleand infeasible Both two infeasible
  • 13.
    denotes that (k-1)xi Pbis dominated bykxi. , , k k k i i jk b i j k k k j j k if if  =   x x x x x x x   ( ) ( ) ( ) ( )( ) ( )( ) ( ) ( )( ) ( )( ) ( ) ( )( ) ( ) ( ) ( ) 1 1 1 1 1 0 0 0 0 k kk Pb k Pb i i i i kk Pb kk Pb i i i i kk Pb i i f f − − − − −  < ∧ Φ = ∧ Φ =   ∨  ⇔ Φ = ∧ Φ >  ∨  Φ < Φ x x x x x x x x x x 
  • 14.
  • 15.
    0k k Pb µΦ=≥Φ ( ) ( ) ( ){ }1 k Pb k Pb k Pb k Pb i N=Φ Φ ΦΦ x x x  ( ) ( ) ( ) ( )( ) ( )( ) ( ) ( )( ) ( )( ) ( ) ( )( ) ( )( ) ( ) ( )( ) ( ) ( ) ( )( ) 1 1 1 1 1 1 k kk Pb k k k Pb k k i i i i k kk Pb k k k Pb k k i i i i k kk k k Pb k k k Pb i i i i f f µ µ µ µ µ µ − − Φ Φ − − Φ Φ − − Φ Φ  < ∧ Φ ≤ Ψ ∧ Φ ≤ Ψ   ∨   ⇔ Φ ≤ Ψ ∧ Φ > Ψ  ∨  Φ > Ψ ∧ Φ > Ψ ∧ Φ < Φ x x x x x x x x x x x x  ( ) [ ] ( )01 0,1 0, , ;k k k L L η η +  Ψ= − ∈ = ∈     
  • 16.
  • 17.
    17 Application to staticanalysis of steel structure
  • 18.
    Porto, June 202017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 18 3 33 A s s ε − ≤ ] ] 3 33,38 A s s ε − ∈ ] ] 3 38,42 A s s ε − ∈ 3 42 A s s ε − > Section Class •1 •2 •3 •4 235 yf ε =
  • 19.
    Porto, June 202017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 19 Buckling resistance of steel bars , 1.0Ed b Rd N N ≤ 1 , y b Rd M Af N χ γ = 1 , eff y b Rd M A f N χ γ = 1-2-3 4
  • 20.
    Structural optimization More thanminimizing weight! Technological/constructive constraints New class of “shapes”
  • 21.
    Porto, June 202017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 21 Optimal topology (?)
  • 22.
    Conclusions Interface OpenSees withEA • External Optimization algorithm • External constraint evaluation • Constrained handling Futures • Parallel computing • General Constrained handling • More complex problems Porto, June 20 2017 OPENSEES SOLVER WITH A DIFFERENTIAL EVOLUTIONARY ALGORITHM FOR STRUCTURAL OPTIMIZATION OF HOLLOW SECTIONS STEEL STRUCTURES 22 Conclusions