Optimization of multi-pass turning operations using ant colony system

1
Optimization of multi-pass turning operations
using ant colony system

• Authors K. Vijayakumar, G. Prabhaharan, P.
  Asokan, R. Saravanan
• 2003
• Presented by Kent Fairbourn
• 10-31-07

2
Purpose

• Compare the Ant colony algorithm with the Genetic
  algorithm and simulated annealing.
• Find a faster method for producing a global
  optimization
• Optimize per unit cost for turning operations
• Handbooks are insufficient do not consider
  economic aspects of machining

3
Genetic Algorithm

• Used to find exact or approximate solutions to
  optimization problems
• Population of solutions evolves toward a better
  solution
• Each generation involves mutation (combination)
  of fit solutions
• Terminates at set number of generations or a
  solution of predetermined fitness
• Weaknesses Finds local rather than global
  solutions. Cant adjust for long term benefit.

4
Simulated Annealing

• Approximated a global optimum for a given search
  function
• Solutions are mixed as parts and compared to
  nearby solutions
• Compares all solutions to one another
• Time consuming Many iterations required

5
Real Life Ant Colony Behavior

• Ant wander for food, find it and return to their
  colony, leaving a pheromone trail
• Other ants follow same trail
• Longer trails lose pheromone density
• Shorter trails prevail and are used by all

6
Ant Colony Optimization (ACO)

• Each virtual ant takes a random path
• Paths are evaluated and mutated or replaced
• Values associated with successful solutions
  receive more virtual pheromones (Trail Value)

7
Governing Equations

• UC CM Ci CR CT
• Goal Max F(X) -UC(X,n,d)
• (X is the machining parameter set, n is the
  number of passes, and d is the depth of cut)

8
Optimization Parameters

• Cut Speed
• Feed Rate
• Depth of Cut
• Tool Life
• Operating Constraints
• Power
• Stable Cutting Region
• Chip-Tool Interface
• Finish Machining
• Surface Finish

9
Parameters Used
10
Parameter Values

• Each parameter has an upper and lower bound for
  the virtual ants to choose from.
• Values taken from Chen and Tsais Simulated
  Annealing Model

11
ACO
12
ACO
13
ACO applet

• Example courtesy of Mark Sinclair

14
ResultsACO vs. SA vs. GA
Method Unit Cost () Runs Difference
ACO 1.6262 1-15 ------
GA 1.7610 1-100 -7.65
SA 2.2959 901-1000 -8.28
ACO Optimum Machining Parameters
Cutting Speed (m/min) Feed (mm/rev)
Roughing Cut 103.05 0.9
Finishing Cut 162.02 0.24
15
Conclusions

• ACO obtains a near optimal solution among a large
  solution base in reasonable time
• Superior to GA and SA
• Generic algorithm can be applied to varying
  parameters and constraints
• Also applicable to Milling and Threading
  Operations

16
References