An integrated ANN-GA for reliability based inspection of concrete

In most concrete bridge decks subject to deicing slats or constructed in chloride-laden environments, corrosion has caused serviceability damage in the form of severe cracking and/or spalling of the concrete cover. In this paper, whilst an analytical model is used for the simulation of corrosion induced crack width, random fields are utilized accounting for the spatial variability of the concrete material and environmental factors. Then, using the Monte Carlo simulation method, the extent of the damage is simulated as a dependent random variable during the service life of the bridge deck. Finally, for finding optimum reliability-based inspection plans, with minimum life cycle costs, an integrated Artificial Neural Network-Genetic Algorithm (ANN-GA) approach is proposed. The applicability of this method is investigated on a hypothetical bridge deck. It is concluded that the employed approach can well handle the high computational complexity of the problem in finding optimum inspection plans.