Optimization Algorithm Support Platform Design And Implementation Based On Ontology

Optimization algorithm is an applied technique on various optimization problems in engineering practice. People have proposed many meta-heuristic algorithms, e.g. Genetic Algorithm, Ant System. Optimization algorithm is developing toward systematic, and going to form a unified and integrated theory system. Practical research may lead to more efficient support to the progress due to absent of theory direction. Besides, there are many necessary but boring works having nothing to do with creation, which could be assisted by computer. Therefore, research on software platform supporting optimization algorithm research is valuable both theoretically and practically.First, optimization algorithm unified model is constructed. There are proposed that algorithm be divided into operators, which operate data in work space, and the modeling principle based on flow. The model divides algorithm flow into 3 stages, 4 areas and 6 kinds of operators. Then algorithm research could be carried out based on operators forming new algorithm by modifying and combining operators.Second, Petri net is applied as a modeling tool. Petri net, concisely symbolized, is fit to algorithm in two points, one emphasizes the locality of system state change rather than globality which agrees with the case that operators operate data in work space, the other emphasizes elimination of transition control inherence which makes it easy to research algorithm structure.Third, the concept of Algorithm Project is proposed. To one problem, researchers solve it with different flows and parameters, get different results and evaluations. They form a run record, which is the basis of comparison of algorithm. It is consist of one problem and several run records.Finally, applying Object-Oriented method, Optimization Algorithm Support Platform is implemented on Windows OS based on the model above. Several usual optimization algorithms are implemented on the platform, which proved the validity of the model and the platform. It is helpful to manage operators abstracted from various algorithms and mass of data generated during research, make it possible to concentrate on creation in research, reduce complexity and improve quality and efficiency of algorithm research.