The Design And Application Of Object-Oriented Genetic Algorithm Platform

Genetic algorithms appear the excellent performance on solving the optimization of complex problem by a number of researchers interested in. But for the implementation with specific code and the application in project, there are a lot of inconveniences which limits GA's further promotion in the field of industrial optimization. So the work of this thesis tried to solve the difficulty and inconvenience which GA has in the practice to improve the application efficiency.This article described the history of development of genetic algorithm, the scope of application and the result achieved in software development about GA, and then argued the basic principles and technology issues. Based on these analyses, genetic algorithm platform was implemented with the help of object-oriented programming ideas and methods. The platform was designed and developed with the sub-level, modular development methods to achieve the goal of efficient compute, easy use and smooth relocation from experiment environment to field application. The platform provided a genetic computing framework, pre-build common coded type, such as binary-coded, real-coded and symbol-coded, and other related genetic operator. A new method of use with GA– engine driver– was proposed. Graphic tools which are used in adjusting parameters, analysis the result were also provided with this platform. The adoption of the platform, researchers can focus on business-related part of the adjustment algorithm, reduce the most common work and enhance efficiency.In the program implementation, it made full use of template features in C++ language to use evolution object type as template parameter. The other component like genetic operator could be determined during the period of program compiling, which had better running efficiency than traditional runtime polymorphism to solve the problem about the undetermined genetic code type. Object-oriented method and appropriate design pattern were used in code implementation to separate program and data flow, reduce the coupling between objects, and enhance the platform scalability and code maintainability. The cross-platform feature of standard C++ and wxPython library helped the platform library and graphic tools could be run in multiple operating systems such as Windows, Linux.This article tried to solve some problems such MaxOne, function optimization and TSP, that showed general steps and the consistency method used based this platform to develop genetic algorithms. Some genetic operator related with evolution object type are argued and designed in experimental simulation section. The experiment about how the various combinations of genetic operator and parameter affect the optimization result found that the parameters chosen carefully could increase the effect."No free lunch in optimization"was concluded. The result of experiment also helped users chose suitable parameter. Finally, with the help of this platform, the genetic algorithm was applied in steel charge design problem to reduce the steel mark and redundant steel. An improved symbol-coded and heuristic mutation operator was proposed to solve this problem. The program prototype and comparative experiment were finished rapidly with the use of this platform. The experimental and practice programs all showed this platform's application value and application prospect.