ABSTRACT
ABSTRACT: Bridge management systems currently face the great challenge to balance the limited available funds and increasing needs for bridge maintenance, repair and rehabilitation activities. This provides an excellent opportunity for applying advanced mathematical programming techniques in this field. This paper presents a dynamic programming (DP) procedure for identifying optimal bridge maintenance plans that are associated with the minimum total maintenance costs and satisfy both bridge safety and condition requirements for a targeted lifetime period. Since the effective service lifetime of a bridge without maintenance may not be long enough to reach a target level, it can be extended by applying sequential maintenance actions. The effect of any maintenance action on bridge safety and condition profiles can be classified as (a) improvement of current safety and/or condition indices, (b) delay in deterioration occurrence, (c) reduction of deterioration rates and (d) combinations of the three previously mentioned effects. Any combination of the maintenance action candidates, which can extend the bridge effective service lifetime to the targeted level, can be regarded as a feasible bridge maintenance plan. These feasible maintenance plans may
require performing different combinations of the maintenance action candidates at different application times, resulting in different life-cycle maintenance costs. In order to find an optimal plan, the life-cycle maintenance cost for each feasible maintenance plan must be converted to the net present value (NPV), using a reasonable discount rate. An optimal maintenance plan in this study is the feasible plan that has a minimum life-cycle maintenance cost in terms of NPV. Theoretically, an optimal bridge maintenance plan can be obtained by comparing the life-cycle maintenance cost for all feasible plans, using the enumeration methods. However, the advanced mathematical optimization techniques such as DP and genetic algorithms (GAs) can provide more efficient approaches particularly for stochastic systems that involve random variables in computations.
