Railway deterioration has an immediate impact on our daily life. The correct functioning of trams, metros and trains is required to regularly transport hundreds of millions of passengers and tons of goods. A key aspect to guarantee the massive passenger and freight transport is to prevent or, at least, slow down railway deterioration. The main goal of this dissertation is to have a better understanding of track vibrations and interactions between components. In turn, such deeper understanding will enable us to slow down the deterioration through new, optimized track designs and maintenance measures, thus extending the service life of railway tracks and, consequently, lower their life-cycle costs. Therefore, we focus on measuring and modeling the vertical dynamics of railway tracks. The first part of the dissertation focuses on the in-depth analysis of extensive field hammer test measurements. Hammer tests are simple and inexpensive, yet they provide valuable information about the characteristic frequencies of tracks. To assess the potential of hammer tests to be employed for track deterioration investigation, we designed and conducted a feasibility study on insulated rail joints (IRJs) in the field. First, a reference dynamic response is defined and then it is compared to the response of three different damaged IRJs. Three characteristic frequency bands related to the damaged IRJs are derived independently of the type of damage. In view of these promising results, a Frequency Response Function (FRF)-based statistical method is proposed to identify characteristic frequencies of railway track defects. The method compares a damaged track state to a healthy state, which, in this case, is defined following the concepts of control charts employed in process monitoring. The FRF-based statistical method is tested at squats of different severity in two tracks types. For both squats and damaged IRJs, the identified characteristic frequencies agree with those found with an extensively validated vehicle-borne detection system (i.e. Axle Box Acceleration (ABA) system). This means we are indeed able to identify characteristic frequencies of defects using hammer tests. In the second part of the dissertation, a three-dimensional finite element (3D FE) model of tracks with monoblock sleepers is developed to study in-detail track vertical dynamics. This sleeper type is used worldwide, but its dynamic behavior is often not accurately considered in track models. To study the track dynamics, hammer tests are numerically reproduced applying an Implicit–Explicit FE procedure. First, the equilibrium state of the track is calculated and then the response of the track to hammer excitation is simulated in the time domain. Next, our 3D FE model and field measurements are combined by fitting simulations to measurements, so that (1) the model is validated and the accuracy to reproduce measurements is determined, (2) the in-service track parameters are derived, and (3) insight is gained into the contribution of components to track dynamics and the effect of simplifications in modeling. After validating our 3D FE model with measurements and performing a comprehensive analysis, we find that the frequency response between 300 and 3000 Hz is defined by seven characteristic features and vibration modes. This is an important result for understanding track dynamics, given that the tracks with biblock sleepers only have four such features. The bending modes of monoblock sleeper and the stronger coupling between the two rails cause two of the additional features. The third additional feature occurs in the frequency range dominated by the rail-railpad-sleeper interaction. With the 3D FE model as basis, the influence of the representation of this interaction into the numerically calculated vertical dynamics is investigated. For this purpose, four fastening representations are developed: (1) commonly used spring-damper pair, (2) area covering spring-damper pairs, (3) solid railpad connected to the rail, and (4) solid railpad in frictional contact with the rail and fixed to the support by preloaded springs, which represent the clamps. Their comparison shows that the overall numerical reproduction of the measurements improves the more realistic is the representation of the fastening. If the accuracy of the fastening models is quantified, the model with solid railpads and clamps reproduce the seven characteristics at a maximum frequency difference of 6%, whereas for the conventional model, the difference can be as high as 27%. The lateral and longitudinal dimensions of the railpad, and the lateral and longitudinal constraints between the rail and support applied by the fastening, are both relevant aspects to consider when modeling fastenings. By examining field measurements and numerical models, useful information is gained for track design and for the development of maintenance measures. Regarding measurements, if field hammer tests are analyzed by employing the FRF-based statistical method, characteristic frequencies of track defects can be identified and can become valuable input data for the development of vehicle-born detection systems, such as ABA systems. Concerning the 3D FE model, a deep insight into the vertical dynamics of tracks with monoblock sleepers is obtained and the comprehensive study of fastening representation has indicated the need to model this track component more realistically, so that the track dynamics at high frequencies can be correctly reproduced. This finding should be considered in vehicle-track models to improve the reproduction of vehicle-track high frequency dynamics. A more realistic fastening modeling is especially required to represent tracks with rail defects since the defect's development may be related to the condition of the fastening. Thus, the advanced fastening models presented in this research are expected to significantly contribute to the investigation of fastening degradation and track defects. In addition, the evolution of in-service track parameters (i.e. stiffness and damping of railpad and ballast) identified during the fitting process can be monitored and it may become valuable input data for the planning of maintenance. Overall, our combination of measurements and numerical models contributes to the understanding of track dynamics. Furthermore, we have provided models and tools that can be used for the investigation and monitoring of track deterioration.