Bolt Pattern Design Optimization Project

Nitin Chandra 2010127 Kunal Ji Baranwal 2010098

Optimization Problem
The goal of this problem is to find the bolt

locations that maximize the force P that can be carried by the bolted joint before it fails. Failure occurs when the shear stresses in any one of the bolts exceeds the yield stress of the bolt. In order simplify the calculations, we will reformulate this problem as an equivalent optimization problem, which is to find the bolt locations that minimize the fraction of the force P that is felt by any one of the bolts.

Formulation of Objective Function

The shear force, from the applied load, on each

of the bolts: The applied force vector for the force pulling the plates apart is: = 1/ 2 + 1/ 2
The direct shear load is distributed evenly

between the three bolts in the direction of the applied force. This force on each bolt is given by the equation: =/3 1/ 2 + 1/ 2

 Forces induced by the effective moment between

and the centroid of the bolt pattern also exist ( ). In order to find these forces, the moment vector, =
Where is moment arm, or position vector from

the centroid of the bolt pattern to the applied load: = +

=
= [ 2
1 +2 +3 3

and =

1 +2 +3 3

Finally, substituting and we get:

+ ]

 we can now calculate how it is distributed among

the three bolts. This can be done by first finding the moment arm between each and the centroid of the bolt pattern: = + These moment arms are then crossed with the vector to find the direction of the moment force for each bolt: 1 = ( )

1 2 + 2

[ + ]

Additionally, we must know the magnitude of each

of these forces, which can be found by using the following equation: .

 Now that we have both the magnitude and

direction, we can use their product to find : =

With both and known, we can now find the

magnitude of the resultant force on each of the bolts:

= +
= | |

Constraints
To ensure that the bolt pattern does not fail by

tear out (the bolts being too close to the edge of the plate), a safety factor of 1.25 diameters should be used. 1.25d < , < W 1.25d
Additionally, to ensure that the bolt heads do not

interfere with each other, we will add a constraint that the distance between any two bolts must be at least as great as the diameter of the bolt heads: 1,2,3
2

>

for i , j =

Data
= 1.0 in

= 4.75 in
P =1 (assume unit force for optimization problem) Length of steel plates, L = 5:5 in

Width of steel plates, W = 2 in

Bolt diameter, d = 0:25 in Bolt head diameter, = 0:5 in Minor bolt diameter area, A = 0:0269 2 Minimum bolt proof strength, Sy = 85kpsi

Solution Methods
Gradient Based Methods
Generalized Reduced Gradient Method ( GRG )

Evolutionary Methods
Particle Swarm Optimization ( PSO ) Genetic Algorithms

GRG

PSO
Our Code

MATLABs PSO toolbox

PSO Algorithm ( Pseudo-Code )

Genetic Algorithms
Our Code

Matlabs GA toolbox
Fitness: To evaluate the fitness, each design must be analyzed to evaluate the objective f (minimized) and constraints gi <= 0 (i = 1 to m). Feasibilty of Design:
Fitness :

GA: New Generations

The genetic algorithm goes through a four-step

process to create a new generation from the current generation: 1) Selection

1) 2) 1) 1) 1)

Tournament Selection Roulette Wheel Selection Blend crossover To introduce diversity into the population of designs. Children must compete with their parents to survive to the next generation.

2) Crossover 3) Mutation 4) Elitism

GA: New Generations

The genetic algorithm goes through a four-step process to create

a new generation from the current generation:

1) Selection
1)

Suppose our tournament size is three. We randomly select three designs from the current generation, and the most fit of the three becomes the mother design. Then we randomly select three more designs from the current generation, and the most fit of the three becomes the father design. One may vary the fitness pressure by changing the tournament size.
1)

2) Crossover
A crossover probability is specified by the user. A random number between zero and one is generated, and if it is less than the crossover probability, crossover is performed. Otherwise, the mother and father designs, without modification, become the two children designs.

2)
3) Mutation
1)

4) Elitism
1) The new generation is combined with the previous generation to produce a combined generation of

Results
MATLAB Tool Box Results

[ 1.6875 0.3126 1.5898 0.3132 1.6875 [1.6875 0.3125 1.6061 0.3125 1.6875 1 1.5867] Genetic Algorithm (0.860372 seconds) PSO (1.042512 seconds)

Results
Our Code

[ 1.55216 0.57395 1.47057 0.40754 1.53735 1.68749] PSO

References
An improved particle swarm optimizer for

mechanical design optimization problems (2007)

S. He , E. Prempain & Q. H. Wua Department of Electrical Engineering and

Electronics, The University of Liverpool, Liverpool

Thank You