---

Corrosion of reinforcement due to frequently applied deicing salts is the major source of deterioration of concrete bridge decks, e.g. severe cracking and spalling of the concrete cover. Since crack width is easily recordable in routine visual inspections there is a motivation to use it as an appropriate indicator of condition of RC bridge elements in decision making process of bridge management. While few existing research in literature dealing with spatial variation of corrosion-induced cracking of RC structures is based on empirical models, in this paper the extent and likelihood of severe cracking of a hypothetical bridge deck during its lifetime is calculated based on a recently proposed analytical model for corrosion-induced crack width. Random field theory has been utilized to account for spatial variations of surface chloride concentration, as environmental parameter, and concrete compressive strength and cover depth as design parameters. This analysis enables to track evolution of cracking process, spatially and temporally, and predict the time for the first repair of bridge deck based on acceptable extent of cracked area. Furthermore based on a sensitivity analysis it is concluded that increasing cover depth has a very promising effect in delaying corrosion phenomenon and extension of the service life of bridge decks.

---

The purpose of this study is to compare the results of different probabilistic methods such as the perturbation method, Stochastic Finite Element Method (SFEM) and Monte Carlo Method. These methods were used to study the convergence of direct approach for slope stability analysis and are developed for a linear soil behavior. In this study, two dimensional random fields are used and both the First Order Reliability Method (FORM) and Limited Step Length Iteration Method (LSLIM) have been adopted to evaluate the reliability index. The study found that the perturbation method of the second order is easy to apply using the field’s theory because accuracy is reached even with different coefficients of variation of input variables, while the spectral finite element method yields accurate results only for high levels of solution development.