Trade-off analysis for infrastructure management: new approaches to cross-asset challenges

As transportation infrastructure managers pursue performance-based management, increased scrutiny is rightfully imposed by the stakeholders (tax-payers and their legislative representatives). The performance of public transportation agencies is not evaluated on a single asset type (i.e. on Pavement, or Bridges alone) but on the system as a whole (cross-assets). Many commercial software packages are focused on managing a single asset type making cross-asset analysis difficult. However, a main problem for transportation agency managers is how to split available budget among different types of assets to provide the best overall performance to the public stakeholders. This paper focuses on two different approaches to the Cross-Asset Problem (CAP) and demonstrates, using real data examples, how optimal budget distribution for various asset types can be found for large agencies. The paper formulates individual asset management as an integer optimization problem (IP). It then discusses two approaches to the CAP. The first approach assumes that overall agency performance can be expressed as a linear combination of individual asset type performances. In this case the CAP can be formulated as an Integer Optimization Problem (IP) where its objective and constraints are defined as a simple linear combination of the objectives/constraints for corresponding asset types. Model formulation, running times and optimal budget distributions using real transportation agency data are presented for this approach. The second approach requires no assumptions on the CAPs objective formula and considers separate asset type problems as a black box which, for a given budget distribution, returns the best overall performance for that asset type. A Derivative Free Optimization algorithm is presented for this setup, showing running time and final budget distributions for several examples. Finally, the paper outlines the advantages and disadvantages of each approach and provides guidance on when each approach should be used. (C) 2016 Published by Elsevier B.V.