In this thesis, we propose a modified genetic algorithm (referred to as GA) for large-scaled reliability redundancy allocation problems (referred to as RRA) with series-parallel systems which were proven NP-hard. Traditional GAs do not perform well in large-scaled RRA without applying any frontier search technique. To obtain high quality solutions, it is critical to search the region near the boundary of the feasible region, but when considering large-scaled problems, it is difficult to achieve with such traditional GAs. In the proposed algorithm, a large-scaled RRA problem is first narrowed down to an approximate core problem by applying some particular variable reduction strategies, and then the corresponding approximate core problem is solved by a modified hybrid genetic algorithm (referred to as HGA) first proposed by Huang. The proposed algorithm can obtain high quality solutions for large-scaled problems in a reasonable time and can be adopted into various problems by altering some mechanisms.