Parametric study of the penalty function in a genetic algorithm applied to the support arrangement for piping systems

A genetic algorithm (GA) is applied to the support arrangement of the piping systems in the nuclear power plant. In a GA, the form of the fitness function is the most important theme. It has a large influence on the result of support arrangement. There are many factors to solve for this problem; for example, stress of the piping system, the number of supports, strength of the supports, and so on. The authors attempt to select a more reasonable fitness function by trial and error. They illustrate the results of numerical experiments using an actual piping system. They found that the sigmoid function is good for the stress penalty function and that the balance of fitness function terms is important. The design results using the GA applied compare favorably with the results of the designer.