O framework NP-Opt e suas aplicações a problemas de otimização

One of the greatest difficulties in working with combinatorial optimization is that the methods require specific procedures for each problem. Even general purpose methods, like Evolutionary Algorithms, require operators developed specifically for the problems. Suppose, for example, that someone has developed a code to solve Single Machine Scheduling problems. If such person decides to change and address a Flowshop problem, most of this code will have to be changed, wasting precious time due to a low reutilization of the software components. In order to overcome this problem, we developed an optimization framework, named Np-Opt, which allows a high degree of code reutilization, facilitating the migration from any type of problem to another. The core of this framework is composed of two evolutionary algorithms: a Genetic and a Memetic. In this thesis we will examine thoroughly all the main characteristics of the framework and present results for four different NP problems that are incorporated to the NP-Opt: Single Machine Scheduling, Permutational Flowshop with families of jobs, Gate Matrix Layout and Gene Ordering. A fifth problem, which was not included in the framework due to the need of accessing external packages, but whose implementation uses NP-Opt classes is also commented. It is named Capacitor Allocation in Distribution Networks, and is described as an external application of the NP-Opt.