A hybridized genetic algorithm for cost estimation in bridge maintenance systems

Abstract

A hybridized genetic algorithm is proposed to determine a repair schedule for a network of bridges. The schedule aims for the lowest overall cost while maintaining each bridge at satisfactory quality conditions. Appreciation, deterioration, and cost models are employed to model real-life behaviour. To reduce the computational time, pre-processing algorithms are used to determine an initial genome that is closer to the optimal solution rather than a randomly generated genome. A post-processing algorithm that locates a local optimal solution from the output of the genetic algorithm is employed for further reduction of computational costs. Experimental work was carried out to demonstrate the effectiveness of the proposed approach in determining the bridge repair schedule. The addition of a pre-processing algorithm improves the results if the simulation period is constrained. If the simulation is ran sufficiently long all pre-processing algorithms converge to the same optimal solution. If a pre-processing algorithm is not implemented, however, the simulation period increases significantly. The cost and deterioration tests also indicate that certain pre-processing algorithms are better suited for larger bridge networks. The local search performed on the genetic algorithm output is always seen as a positive add-on to further improve results.