You are on page 1of 5

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309427071

Using the Ant Colony Optimization method to find an optimal solution in


drilling process

Conference Paper · October 2016

CITATION READS

1 277

1 author:

Danijela Pezer
University Department of Professional Studies, University of Split, Croatia
21 PUBLICATIONS   36 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Danijela Pezer on 26 October 2016.

The user has requested enhancement of the downloaded file.


Using the Ant Colony Optimization
method to find an optimal solution in
drilling process

Danijela Pezer
Mechanical Engineering Faculty in Slavonski Brod, Trg I.B. Mazuranic 2, 35000
Slavonski Brod, Croatia, E-mail:dpezer@sfsb.hr

Abstract
The aim of each production process is to reduce the processing time, which ultimately leads to
reduction of the total costs in production. Considering the stated objective, Ant Colony Optimization
(ACO) method for searching of the optimal tool path in the process of drilling, was used in this paper.
In this case the efficiency of production depends on the time required for drilling. ACO method is
used to solve NP hard combinatorial optimization problems inspired by the behavior of different
species of ants in searching for the shortest route. One of the most extensively studied combinatorial
problem is the Traveling Salesman Problem (TSP). Analysis of this technological task is also based
on this principle, and we consider an open tour in drilling process where at the end of the process of
drilling, tool does not return to the starting point (which depends on the random selection of the
proposed algorithm). With the proposed ACO algorithm, a satisfactory solution was achieved in a
relatively short time, and the algorithm has shown that is reliable to use. The solution achieved with
ACO algorithm was obtained by MATLAB program, and results are compared with the result
obtained by manual programming. Simulation of drilling was performed by educational program
EMCO WinNC for Sinumerik 840D Mill control unit.

Keywords: Ant Colony Optimization, Tool path planing, CNC programming, Drilling

1. INTRODUCTION set of arcs. For each edge, or arc i, j is

Today, many metaheuristic methods for assigned a value d ij which can indicate the
solving combinatorial optimization problems are distance, time, price or other interest factors
used. In this paper, for finding the shortest tool associated with edges, respectively arcs. For
path in the process of holes drilling, the Ant standard TSP problem the assumption is that the
Colony Optimization (ACO) metaheuristic square cost matrix, D ^d ij : i, j V , i z j ` is
method was used. The problem is based on the symmetric d ij d ji , respectively, distance is
principle of a Traveling Salesman.
equal in both directions. Another standard
Traveling Salesman Problem (TSP) can be assumption is that the distance matrix D d ij
represented as a fully weighted undirected graph
satisfies the triangle inequality in the case when
G V , E if it is symmetric, and fully weighted
d ij d d ik  d kj , , za  i, j, k  V .
directed graph G V , A if it is asymmetric. The
set V ^1, 2, ...n ` is the set of vertex, which The objective of the problem is to find a
minimal Hamiltonian cycle, which closes the
represents the cities, E ^ i, j : i, j V , i  j `
tour after each n V vertexes of G visited only
represents the set of edges which connecting all
vertices, while A ^ i, j : i, j V , i z j ` represents once.

8
In many papers the possibility of applying ƒ Initialize of pheromone trails (on each way
ACO [1,2] and other algorithms, such as Particle is set the amount of pheromone that is
Swarm Optimization [3], genetic algorithms [4], equal to the reciprocal value of the path
and more recently with the use of the bat length found by using a greedy algorithm)
algorithm [5], cuckoo search algorithm [6] etc.,
ƒ Route selection - an ant selects the next
for the various production problems based on
hole from the group of holes that have not
the principle of a Traveling Salesman described
yet been drilled. Since that is a fully
above are investigated.
connected graph, from each hole it can be
reached to any remaining hole with the
2. ANT COLONY probability:
W ijD u K ijE (1)
OPTIMIZATION METHOD Pij
¦ W ikD u K ikE
Ant Colony Optimization (ACO) is classified kN

as swarm intelligence, inspired by the behaviour Pij - the probability of path choosing from
of different species (ants, bees, birds, fish, bats) hole i to hole j
during the search for food in nature. N - a set of holes that haven't been drilled
Except the Ant Colony Optimization, the yet
other algorithms that imitate the phenomenon of W ij - the intensity of the pheromone on the
swarm intelligence, such as a Particle Swarm path from hole i to hole j
Optimization (PSO), bee algorithm, firefly K ij - the value of heuristic function on the
algorithm, bat algorithm, etc. have been
path from hole i to hole j
developed.
Į ± constant (pheromone trail)
Ant Colony Algorithm imitate the movement ȕ ± constant (visibility between holes).
of ants in the nature, during the search for the
shortest route to the food, or finding the optimal For Į 0 algorithm acts like Greedy
solution for a given prREOHP6LQFHDQWVKDYHQ¶W algorithm.
well-developed eyesight, their mutual
For ȕ=0 the search is based only on the
communication is based on pheromone trails.
pheromone trail.
The amount of pheromone on the ground
determines the probability to be followed by ƒ Updating the trails (occurs in two phases):
other ants. o Evaporation phase
The basic algorithmic structure [7] of Ant The pheromone trails are updated for all
Colony is presented by Algorithm 1. paths using relation
Algorithm 1. General algorithm for ACO W ij 1  U W ij , i , j  >1, n@
Initialize the pheromone trails; (2)
Repeat where is
For each ant Do
Solution construction using the U - evaporation constant, U  >0,1@
pheromone trail; Evaporation is possible to apply after each
Update the pheromone trails:
ant constructs a solution or after each iteration in
Evaporation;
Reinforcement; which the solution constructed by all ants (as is
Until Stopping criteria the case in this paper).
Output: Best solution(s) found o Reinforcement phase
The pheromone trail is updated according to
The steps of the algorithm work [7]:
the last constructed solution and can be applied

9
after each step of the heuristics construction for The first hole from which starts drilling
each ant, after an each ant constructs a solution depends on the random selection of the
or after all ants construct a solution in which algorithm, and when drilling operation is
there are several strategies: finished WKH WRRO GRHVQ¶t return to the starting
position. The results were compared with the
¾ reinforcement based on quality ± the
results achieved by manual programming. For a
value of pheromone is updated
given problem of the tool path optimization,
according to the k best solutions
minimizing the length of tool path was realized
(k=1:m; m ± number of ants) with value
using the MATLAB software and the achieved
proportional to the quality of solutions:
software solution was simulated in EMCO
W iS * i W iS * i  ', i  >1, n@ WinNC program for Sinumerik 840D Mill
(3)
control unit (Fig. 2).
where is ' 1 / f S .

¾ reinforcement based on the range ± the k


best solutions updating an appropriate
components of trails to the amount
which depends on their ranking
¾ reinforcement based on a worst solution
± the worst found solution reduces the
appropriate parameters of trails
¾ elitistic reinforcement ± increasing the
value of pheromone for the best solution
found until then. Fig. 2. Simulation of drilling in the Sinumerik 840D
ƒ By stopping the process, the best
accomplished solution, or sequential The program was performed on a work
represantation of possible order of station HP EliteBook 8560w, i72820 QM CPU,
execution of drilling, according to the 2.3GHz, 16 GB of RAM.
assigned number of holes was displyed. The algorithm was started multiple (500
Termination criterion is defined by the times) with different combinations of parameters
number of iterations.
Į(the importance of the pheromone trail), ȕ(the
importance of visibility), ȡ (evaporation factor)
and nAnt (number of ants ± population size). As
3. EXPERIMENTAL PART
the optimal parameters for a given problem are
According to the given task (Fig. 1) the determineted the following parameters Į 
sequence of tool path holes drilling (156 holes) ȕ=4, ȡ=0,45, Q=1, nAnt=70, maximal number of
was optimized by ACO algorithm. iteration maxIter=900, by which a minimal total
tool path are achieved in the amount of
573,1939 mm (in 584 iteration).
The results of ACO algorithm were
compared with results obtained by manual
programming and total tool path length in the
amount of 629,3353 mm was achieved, and the
results are presented in the Table 1.

Fig. 1. Prismatic workpiece

10
Table 1. The results achieved by ACO algorithm and manual programming
Method Max. number of iteration Num. of population Best cost, mm CPU time, s
Manual programming --- --- 629,3353 ---
ACO algorithm 900 70 573,1939 738,9419

The optimal tool path achieved by ACO 5. REFERENCES


algorithm and manual programming is shown on
[1] Z. Q. Li; X. Wang; Y. F. Dong, ACO-
Fig. 3.
Based Holes Machining Path Optimization
Using Helical Milling Operation. Advanced
Materials Research, 834-836 (2014),
pp.1386-1390
[2] X. Hong; L. Yuan; Z. Kaifu; Y. Jianfeng;
L. Zhenxing; S. Jianbin, Multi-objective
Optimization Method for Automatic
Drilling and Riveting Sequence Planning.
Chinese Journal of Aeronautics, 23(6)
Fig. 3. Optimal tool path
(2010), pp. 734-742
The impact of the number of iteration in
[3] W. Elloumia; H. El Abeda; A. Abrahama;
relation to the value of the minimum total tool
A. M. Alimi, A comparative study of the
path length is shown on Fig. 4.
improvement of performance using a PSO
modified by ACO applied to TSP. Applied
Soft Computing, 25 (2014), pp.234±241
[4] J. Abu Qudeiri, Optimization of Operation
Sequence in CNC Machine Tools Using
Genetic Algorithm. Journal of Advanced
Mechanical Design, Systems and
Manufacturing, 1(2) (2007), pp. 272-282

Fig. 4. Fitness function depending on the number of [5] E. Osaba; X.S. Yang; F. Diaz; P. Lopez-
iteration Garcia; R. Carballedo, An improved
discrete bat algorithm for symmetric and
4. CONCLUSIONS asymmetric Traveling Salesman Problems.
Engineering applications of artificial
The aim of the optimization of the total tool
intelligence, 48 (2016), 59-71
path length is to reduce the total time, and thus
the production costs. According to the [6] W. C. E. Lim; G. Kanagaraj; S.G.
technological task, by ACO algorithm was Ponnambalam, A hybrid cuckoo search-
achieved suboptimal solution which is efficient genetic algorithm for hole-making
in relation to the solution achieved by manual sequence optimization. Journal of
programming. The proposed algorithm certainly intelligent manufacturing, 27(2) (2016),
can have a greater use or efficiency at the pp.417-429
problem with a larger number of holes (up to [7] E.G. Talbi, Metaheuristics: from design to
several thousand). The aim of this paper is to implementation. University of Lille ±
show that the proposed algorithm in a relatively CNRS-INRIA, John Wiley & Sons Inc.,
short time finds (sub)optimal solution and that is Hoboken, New Jersey, (2009).
reliable for use on the potential problems based
on the principle of the TSP.

11

View publication stats

You might also like