Concrete viaducts are an important part of urban rail transit systems but they produce considerable noise, thereby affecting the communities living nearby. The vibration generated at the wheel–rail interface is transmitted along the rail and also onto the bridge; hence, noise is radiated from both the rail and the bridge. To facilitate noise prediction, it is desirable to develop a model that takes into account the generation and transmission of vibration in the train–track–bridge system. The vibration and the associated noise of the track–bridge system are computed with a unified vibroacoustic model using a wavenumber domain finite element and boundary element method. An important aspect to note is the frequency-dependent stiffness of a typical rail fastener utilized on bridges due to the resonance of the baseplate between the two rubber pads. In this study, to allow for this effect, a multilayer fastener model is proposed. The proposed procedure is applied to a viaduct with a U-shaped section and compared with field measurements during the passage of trains. The elastic modulus and damping of the rubber pads and the equivalent loss factor of the rail are chosen by fitting the calculated track decay rates to those estimated from the measured rail accelerations under train passages. The wheel–rail combined roughness is also derived from the measured rail vibration. A comparison is then made between the simulated and measured bridge vibration to verify the proposed method as well as the parameters used in the track–bridge system. The predicted noise levels are also compared with the measured results. The effects of the fastener model, fastener stiffness, bridge damping, and interference between multiple wheels are then discussed. It has been found that the bridge noise has a non-negligible effect on the total A-weighted noise levels in the region beneath the bridge and up to 30 m away from the track.

Acoustic monitoring has recently shown great potential in the diagnosis of infrastructure condition. However, due to the severe noise interference in acoustic signals, meaningful features tend to be difficult to infer. It creates a considerable obstacle for an extensive application of acoustic monitoring. To tackle this problem, we propose an acceleration-guided acoustic signal denoising framework (AG-ASDF) based on learnable wavelet transform to automatically denoise the acoustic signal and extract the relevant features based on the acceleration signal. This denoising framework requires the acceleration signal only for the training stage. Therefore, only acoustic sensors (non-intrusive) need to be installed during the application phase, which is convenient and crucial for the condition monitoring of safety-critical infrastructure. A comparative study is conducted among the proposed AG-ASDF and other feature learning / extraction methods, by using a multi-class support vector machine to evaluate the detection effectiveness of slab track condition based on acoustic signals. Different healthy and unhealthy states of slab tracks are imitated with three types of slab track supporting conditions in a railway test line. The classification based on the proposed AG-ASDF features outperforms other feature extraction and learning methods with a significant accuracy improvement.

Rolling noise from running trains is significantly influenced by the wheel–rail combined roughness and the dynamic properties of the track. To facilitate predictions of vibration and noise, it is desirable to be able to determine these parameters accurately from field measurements. In this study, an inverse method for the determination of these parameters is adopted and enhanced. A track model that is based on a wavenumber finite element model of the free rail coupled to discrete supports, which allows for the pinned–pinned mode and cross–sectional deformation of the rail, has been used. The rail vibration induced by hammer impacts and the vibration during train passages are simulated using this model, and these results are then applied to illustrate the accuracy of the direct and indirect methods for the estimation of track decay rate. These methods are compared in a case study for a ballasted track for which hammer impact and train pass–by measurements have been obtained. Other track parameters can also be extracted from the measured data by using the advanced track model. Thereafter, a more complete method is adopted to estimate the wheel–rail combined roughness from the measured rail vibration under train passages. A comparison is made among the estimated roughness levels obtained from this full method, an existing simplified method and the direct measurement method. It is found that the simplified method overestimates the roughness around the pinned–pinned resonance frequency, but gives a good estimation if the track decay rates of the loaded track are used.