GA Truss-1 

Home Up FrameGA GA Truss-1 GA Truss-2

 

Optimization of Trusses-1
Genetic Algorithm Optimization in Structural Mechanics
S.I.Rodin
Department of Material Science and Mechanics

A genetic algorithm approach is used for optimization of trusses so as to minimize the weight of the structure. Island ring topology is used to achieve solution.

The fitness function determines the "goodness" of a given design based on a weighted combination of full weight of the structure, the amount of hinges and some geometrical restrictions:

fitness = f (weight) * f (geometry) ,

where

f (weight) - dependence upon weight, it may be linear;

f (geometry) - necessary geometrical and physical restrictions.

Usual methods are applied to determine geometrical stability, stresses and cross section areas for compressed, tensile and zero steel bars with flexibility limits (compressed bars = 120, tensile bars = 200 and zero bars = 150).

Usual - it means standard methods that engineers use to calculate the stresses and necessary cross-section areas of bars of steel trusses.

 This solution is rather general - hinges are scattered at random with random ties between them. Structure may have any number of support ties at any position. Mutations are applied to the position of hinges and ties. Finite element method is used to calculate displacements and stresses for any negative or zero degree of freedom.

 Loads are shown on the trusses and there may be any number of load combinations - like it's in real construction design.

Flexure limit is taken 1/1000 of span.

 In all cases fitness function k and weight of the best truss is compared to the weight of base steel standard model (the first picture in some examples). The best structures among population of 5 islands each with the subpopulation size of 100 members with 1000 generations are shown. Optimal trusses are shown (blue - tensile bars, red - compressed bars, black – zero bars). Optimal – not the best, it’s always possible to get better solution.

 There are possible two ways: to choose a shape with given ties and optimize it by evolutionary algorithm (as it’s shown on Figure 2) or to use random position of hinges with random ties between them for every member of first generation.

It’s evolution and every new start of the computer – new results, with some differences bigger or smaller from time to time.

Sure, not all best structures may be used - there is a possibility to choose among the best in every island.

 Though initial structures have different negative or zero degree of freedom (SS on the pictures) the best results of evolution in most cases are statically determinate structures.

 

 

Examples of Genetic Algorithm Optimization

 


Figure 1. Free symmetrical shapes ( span = 12 m; forces - each = 100 kN ).

 


a.


b.                                                                     c.

Figure 2. Optimal truss for chosen shape (a) with fixed number of hinges 
with different flexure limits (b, c) ( span = 1500 in; forces - each = 50 klb ).


Figure 3. Optimal continuous frame beams (length = 40 m; forces - each = 100 kN).

 
a. A max = 0.02 m 2                                                     b. A max = 0.03 m 2
Figure 4. Optimal trusses with different limits on maximum cross section area
( span = 40 m; forces - each = 100 kN ).




Figure 6. Optimal arch trusses with different flexure limits
( span = 40 m; forces - each = 100 kN ).

 

 


E-mail:  stullia@mail.ru

Home ] Up ] NN and GA ] FrameGA3D ] GA Irrigation ] GaWaterPump ] GA Oil Pump ] DownloadDemo ]