Location:#0215, Jiuli campus
The U.S. railroad companies spend billions of dollars every year on railroad infrastructure maintenance in order to ensure operational efficiency and safety on their extensive railroad networks. In previous practice, the decision making process for maintenance planning was largely manual and mostly relied on expert knowledge and experience. This talk will present mathematical models and solution algorithms for two interrelated optimization problems: maintenance project scheduling, and resource planning. At the beginning of every fiscal year, the railroad identifies thousands of needed infrastructure maintenance jobs. We cluster these jobs into hundreds of projects, schedule the projects for the fiscal year, and then plan resources to support these projects. The goal is to minimize the (expected) system-wide cost, taking into account many types of business and technical constraints as well as the risk of probabilistic operational disruptions. A time-space network model and a multiple neighborhood search algorithm were developed to solve the project scheduling problem. An integer programming formulation of a reliable location-routing problem and customized solution approaches (e.g., Lagrangian relaxation with embedded column generation and local search) were developed for resource planning. These proposed models have been implemented in practice and have helped the railroad save tens of millions of dollars each year.