One of the common approaches in system reliability optimization is the use of redundant components. It has been proven that the redundant component allocation problem is NP-hard and involves selecting redundant components to optimize system reliability based on pre-defined constraints. In this paper, the maximization of reliability in series-parallel systems is addressed by adding redundant components subject to weight and cost constraints. In the allocation of redundant components, the existence of multiple types for each component is considered, meaning that in addition to determining the number of components, it is also necessary to select the appropriate type from the available options. This problem is modeled as a three-level graph, and an Ant Colony Optimization (ACO) algorithm is employed to solve it. The search capability of the proposed algorithm is enhanced by a local search method in the neighborhood of feasible points, and a dynamic penalty function is used to guide solutions toward feasible regions. The application of this algorithm is demonstrated in optimizing the reliability of a mechanical gearbox system. Numerical results obtained from solving sample problems indicate the considerable efficiency of the proposed algorithm compared to previous approaches, achieving not only the maximization of reliability but also minimizing the required weight and cost.