In this thesis, the application of Bayesian analysis within the condition-based maintenance framework is presented. This maintenance procedure is performed through two phases, diagnostics, and prognosis. Each phase requires distribution as a model. Special emphasis was on the diagnostic phase, i.e. finding suitable technical indicator. A technical indicator is some measurable physical value, that is well correlated with the damage process and is monotonous. In the diagnostic framework, this value should be detected during system operation without load monitoring. Therefore, in this paper, system damping is proposed as a technical indicator. This value is directly correlated to damage, but it is measured indirectly, and therefore it requires additional signal processing. This inherent value is often directly related to damage processes. The shortcoming is that is not a directly measurable quantity, thus special and often complex signal processing is required. This is particularly demanding under unknown load conditions. The solution to the above-mentioned problems is found within the Bayesian framework and pre-processing the signal by the random decrement technique. This technique reconstructs the free decay signals just based on sufficiently long system response. Then, by estimating the damper on a simple dynamic model the system condition is estimated. All this can be achieved with just one sensor. This technique will not always reconstruct the response properly, so a diagnostic signal acceptance criterion is developed. This value is derived directly from the Bayesian analysis, which makes it easy to implement. An additional advantage of this combination, random decrement technique, and Bayesian analysis is the free decay reconstruction correction. Also, it is explained how these results can be used to estimate the remaining useful life within the same framework. In order to test this framework on the real signal, a second-order mechanical experimental setup with an air bearing was developed and verified in detail. A key feature of the setup is minimal internal resistance that does not affect the result and derived conclusions. Also, system parameters are easy to change and control, especially damping. It is shown how this setup can be used to set up more complex experimental models.