Fatigue is the most dangerous failure mode for mechanical components subject to alternating loads. Due to repeated loading and unloading, one or several cracks can be initiated and propagated through the cross section of the structure. Once a critical crack length is exceeded, the structure will catastrophically fail even for stress level much lower than the design stress limit. Non-destructive inspections may be performed at predetermined time intervals in order to detect the cracks. Alternatively, a continuous monitoring of the dynamic response of the structure can allow real-time cracks detection and corrective maintenance procedures might be taken in case the monitoring procedure identifies a crack. In this paper, Bayesian model updating procedures is adopted for the detection of crack location and length on a suspension arm, normally used by automotive industry. Experimental data of the damaged structure (frequency response function) are simulated using a high-fidelity numerical model of the arm. A second, coarse-model represents the model to be updated where cracks of random locations and dimensions are introduced. The idea underlining the approach is to identify the most probable model consistent with the observations. The likelihood is the key mathematical formulation to include the experimental knowledge in the updating of the probabilistic model. Different likelihood functions can be used based on different mathematical assumptions. In this work, the effect of different likelihood functions will be compared to verify the capability of Bayesian procedure for system health monitoring. The different likelihoods will be categorized according to the accuracy of the results and the efficiency of the numerical procedure.