Optimization of Machine System Designs Using Hierarchical Decomposition Based on Criteria Influence

Machine product designs routinely have so many mutually related characteristics that common design optimization methods often result in an unsatisfactory local optimum solution. In order to overcome this problem, this paper proposes a design optimization method based on the clarification of the conflicting and cooperative relationships among the characteristics. First of all, each performance characteristic is divided into simpler basic characteristics according to its structure. Next, the relationships among the basic characteristics are systematically identified and clarified. Then, based on this clarification, the optimization problem is expressed using hierarchical constructions of these basic characteristics and design variables related to the most basic characteristics. Finally, an optimization strategy and detailed hierarchical optimization procedures are constructed, after clarifying the influence levels of each basic characteristic upon the objective functions and setting a core characteristic for the product under consideration. Here, optimizations are sequentially repeated starting with the basic optimal unit group at the bottom hierarchical level and proceeding to higher levels by the hierarchical genetic algorithms. Then, the Pareto optimum solutions at the top hierarchical level are obtained. With the proposed optimization methods, optimization can be more easily applied after the optimization problems have been simplified by decomposition. In doing so, the volume of design spaces for each optimization is reduced, while useful and unique rules and laws may be uncovered. The optimization strategy expressed by the hierarchical structures can be used for the optimization of similar product designs, which realize these breakthroughs, yielding improved product performances. The proposed method is applied to a machine-tool structural model.