Swarm Intelligence (SI) in Manufacturing

Abstract

Manufacturing is a field of Engineering which is constantly changing. Due to this fact more and more complicated problems appear and have to be solved by engineers. Some of them are difficult to solve by usual numerical methods, as equations, so a new area is developing at present using Artificial Intelligence (AI) techniques.

There are many different tools like Genetic Algorithms, Fuzzy Logic Systems… but in this thesis we are going to focus on the Swarm Intelligence (SI) which tries to mimic the behaviour of different types of swarms to solve real problems.

In this Thesis we are going to study the three main SI techniques (Ant ColonyOptimization (ACO), Particle Swarm Optimization (PSO) and Bee Colony Algorithms(BCA)) analyzing the fields where they are more useful, their biological metaphor and other interesting aspects.

Finally we are going to choose a BCA technique called Artificial Bee Colony (ABC) algorithm which mimics the behaviour social bees and we are going to analyze it more in deep. We are also going to solve a real problem about Template Matching for Printed Circuit Board Inspection. This looks like a promising field for BCA and it will show ushow effective could BCA be for solving real problems and so, if it’s interesting to keepon researching on this technique in the future.

In our experiment we use a BCA algorithm to make the Template Matching and we check the velocity, robustness and computational cost of it. We also check which is the optimum number of bees used in the program to improve both robustness and velocity. According to the results the employed algorithm fits with the studied problem mostly due to the low computational cost of the program.