In this study, an experimental and numerical analysis of degradation of mechanical properties of butt welded samples of shipbuilding AH36 steel, after long term exposure to a corrosive environment in the north Adriatic lasting 6,12, 24 and 36 months, were conducted. Based on the obtained results a probabilistic model to estimate the fatigue life of free corroded marine structure was established. The samples were subjected to the standard destructive testing while the surface of the samples was tested using optical and scanning microscopy. Using the experimental results as a basis and applying stochastic processes, a model of the sample corroded surface in the weld metal, heat affected zone and the base metal was developed. In order to simulate static tensile testing a FE model was applied, that included the SMCS (stress modified critical strain ) criteria for ductile fracture with different 𝛼 parameters and hardening models for all zones of the welded samples. Results show that by the increase of the triaxiality, which depends on the number and size of the pits in pitting corrosion, the value of critical plastic deformation decreases which causes an earlier fracture. The biggest degradation of mechanical properties was found in the sea tide zone samples, where the tensile strength and the yield strength are reduced by a fourth of their original value after three years of exposure. The basis of the fatigue life probability estimate is the concept of the boundary state function while applying the Monte Carlo method. The calibration of parameters of the fracture mechanics theory using S-N analysis was performed with specially devised computer routines for the case when the initial size of the crack is unknown and for the case when it is set as a three year old corrosion pit. The probability of fatigue fracture and material parameter C from the Paris-Erdogan equation, consequently the crack propagation, are sensitive to the choice of the initial crack size. When updating failure probability after nondestructive testing, after the fourth and fifth year of service life, there are differences in probability estimates.