Railway rolling stock planning is a basic scheduling in railway transport, which assigns physical train units to given time table services and determines a roster of the train units. This planning also involves a scheduling of periodical inspection for the train units. We have proposed an Ant Colony Optimization (ACO) based approach to solve this planning problem. In this paper, local search methods are introduced to enhance the proposed ACO's performance for tackling a large-scale problem. The effectiveness of the enhanced ACO is demonstrated through numerical experiments with instance problems made from real railway lines.