16. Cost Optimization and High Available Heterogeneous Series-Parallel Redundant System Design Using Genetic Algorithms

Heterogeneous redundant series-parallel systems allow the mixing of components within the same subsystem. This diversity feature may improve the overall characteristics of the system compared with the homogeneous case in term of less susceptibility against so called common-cause failures and reduced cost. That means they guarantee longer availability and are quite suitable for systems that are designed to perform continuous processes. But the main challenging task is to determine the optimal design that corresponds to the minimal investment costs and satisfies the predefined constraints. This kind of combinatorial optimization tasks is perfectly solved using heuristic methods, since those approaches showed stability, powerfulness, and computing effectiveness in solving such matters. This task is more complex than the homogeneous case since the search space is getting larger due to the fact that every component available and that can be deployed in a subsystem has to be taken into account. This fact leads definitely to greater chromosome length and makes the search more time consuming. The algorithm has been implemented in Matlab and three different existing models (Levitin, Lisnianski, and Ouzineb) have been considered for a comparison with the homogeneous case and for validation purposes.