Ecological System Algorithms And Their Applications In Industrial Process Control

The phenomena of adaptive optimization in the nature have always fascinated human beings. Creatures and natural ecological systems make the optimization problems with high degree of complexity to scientists solved perfectly. Under this background, Ecological System Algorithms (ESAs) which mimic the mechanism of the nature and creatures?behaviors have emerged in the field of intelligent computation. This category of novel optimization algorithms emulates the natural ecological systems that entirely depend upon the instinct of organisms to adapt their surrounding environment via unconscious searching ability. There are several distinctive features in the ESAs different from conventional optimization approaches (e.g. Mathematical Programming, Dynamical Programming etc.). Three kinds of typical ESAs, i.e., Genetic Algorithms (GAs), Ant Colony Algorithms (ACAs) and Immune Algorithms (lAs) are studied for their basic theories and features in this dissertation. Based on this research, some improvements to these algorithms are proposed according to the problems emerging from practical application areas. The validity and the necessity of these improvements have been proved by some case studies. The achievements in the research work of this dissertation include: 1 .Extending the study of GAs on the aspect of multi-objective optimization, parallel computation and hybrid-variable treatment. A Parallel Multi-objective Genetic Algorithm is proposed based on the Pareto optimality. An objective ranking evaluation technique is developed to associate the tradeoff information to a better solution with preference articulation. A novel double-layer chromosome coding method is presented to express the system hybridness. Computation time is at least reduced to 10 percent of its original value by adopting a hierarchical decomposed parallel computing technique. This approach is general, and capable of flexibly dealing with different types of variables, realizing the interactive multi-objective decision making, and increasing the speed of computation by the parallel implementation. 2. Studying the other interesting approach in the ESAs, i.e. the ant colony algorithm. An Adaptive Ant Colony Algorithm is proposed in this dissertation to tackle continuous-space optimization problems, using a new objective-function-based heuristic pheromone assignment approach for pheromone update and an elitist strategy to filtrate solution candidates. The proposed algorithm guarantees the assignment of pheromone proportional to the optimality of the solutions by adaptive pheromone update strategy, and the correct distribution of the pheromone in the currently optimal path according to the solution encoding information. Study shows that this approach has the reinforcement ability from previous searching, and correctly direct the latter Iv ABSTRACT searching. Thus it avoids the ineffective search and improves the searching efficiency. A numerical example with multiple extremes is solved to validate the effectiveness of the proposed algorithm. 3. On analyzing the mechanism of the vaccination in the immune system, a new approach ?the Self-recognized Full-process Immune Algorithm is proposed to solve complex optimization problems. This approach can make self-adjustment of the immune responses along with the cultivation period of antibodies, to boost or restrict the antibody generation, gradually enhancing the system recovery ability and decreasing the activity of the antigen, with more ef...