This thesis treats the fatigue assessment process of steel bridges. The purpose is to identify areas with potential of enhancement with an overall aim of attaining a longer service life. The aging bridge stock in Sweden and in many other developed countries is an impending economical burden. Many bridges have reached their expected service life and increased axle loads, speeds, and traffic intensity further accelerate the deterioration. An immediate replacement of all our bridges approaching their expected service life will not be possible. For economical and environmental reasons, effort should be put on extending the theoretically safe service life as far as possible.  Fatigue is one of the major reasons limiting the service life of steel bridges .A specific example is the Söderström Bridge in Stockholm, Sweden. Fatigue cracks have been found in the webs of the main beams and theoretical assessments have shown an exhausted service life. As a mean to reduce the uncertainties in the theoretical assessments a monitoring campaign was started in 2008 and continued in periods until 2011. The first continuous period of 43 days of measurements in 2008 forms the experimental foundation for this thesis. A fatigue life prediction involves (i) an estimation of the load effect, (ii) an estimation of the resistance, and (iii) the selection of a prediction model. This thesis treats in some sense all three of them. Considering part (i), the load effect, a theoretical study on the influence of dynamics has been performed. The quasi static approach suggested in the standards does not necessarily reflect the true dynamic behavior of the structure. Performing a dynamic moving load analysis gives for all cases studied a more favorable fatigue life. A further enhancement of the assessment is to perform in situ measurements. The uncertainties related to dynamics, loads, and structural behavior are thereby captured in the response. Routines for processing the measured response and performing life predictions are treated, moreover the quality of the measured response. The resistance, part (ii), in form of the fatigue endurance has a funda- mental influence on the fatigue life. As an attempt to reach a more adequate fatigue endurance a refined assessment of a typical joint is performed using linear elastic fracture mechanics (LEFM). A descriptive detail category is suggested which renders a somewhat more favorable resistance. Part (iii), the prediction model, is treated considering the safety format. A reliability-based model is suggested which enables a consideration of the uncertainties in each stochastic variable. Thereby, all aforementioned parts and uncertainties can be combined within the same prediction. The reliability-based model is used for a code calibration of partial safety factors to be used in semi-probabilistic assessments according to the standards. The implementation of the suggested procedures is shown with a numerical example. The outcome should not be generalized but the example clearly shows an increase in fatigue life with the proposed enhancements of the assessment procedure. / <p>QC 20130917</p>