The identification of the acoustic Green’s function (or, equivalently, the dynamic transfer function) of a medium is of interest to many fields, including structural testing, inspections and health monitoring. This paper will focus on the passive identification utilizing pairs of receivers and exploiting dynamic excitations that occur naturally in the structure. Several opportunities for this passive extraction exist, including: bridges under traffic excitation, buildings under seismic excitation, oceans under natural flows, and railroad tracks under train wheels excitations, among many others. A special signal processing approach is proposed to ensure that the Green’s function (or the transfer function) identification occurs without the influence of the random and generally unknown excitation and without the influence of uncorrelated noise that may affect the receivers. In particular, a special version of Welch’s periodogram technique is proposed where averages of the two outputs are taken both for the same time segments (“intra-segment” averaging) and for different time segments (“inter-segment” averaging) in order to eliminate the influence of noise at both receivers, in addition to eliminating the excitation source spectrum. It will be demonstrated, both analytically and experimentally, that this special signal processing is optimum for robust dual-output passive transfer function estimation. This technique will be then applied to the high-speed inspection of rail tracks by passive extraction of the rail acoustic Green’s function in the ultrasonic regime from the natural train wheel excitations. In this application, the dynamic outputs are collected by pairs of non-contact air-coupled receivers that have a 2-in stand-off from the rail surface. Changes in the passively-extracted Green’s function are then related to the presence of internal flaws (e.g. cracks) in the rail. Previously. a prototype based on this concept has been built and tested at the Transportation Technology Center (TTC) in Pueblo, Colorado, at speeds up to 80 mph. These speeds are unprecedented in the field of rail inspections, that are today carried out at ~30 mph at most by specialized test vehicles. This paper presents preliminary results from a second field test performed in the Fall of 2018 at TTC using a revised prototype design an speeds up to 40 mph. The successful development of this technique would revolutionize many aspects of rail maintenance by, for example, allowing regular trains to perform the inspections with no traffic disruption and great opportunity for redundancy due to the multiple train passes on the same track.
Ultrasonic rail inspection is the most commonly implemented method for detecting internal rail defects. While the conventional ultrasound wheel probe has gained its popularity within rail maintenance community, it suffers from the limited test speeds (25 mph at most). This paper presents the state-of-the-art developments in ultrasonic rail inspection technique that utilizes non-contact receivers and no active transmitters. The transfer function between two points of the rail is reconstructed by deconvolutions of multiple pairs of receivers that sense the acoustics naturally excited in the rail by the running wheels. The deconvolution process eliminates the random effect of the excitation to reconstruct a stable acoustic transfer function of the rail. A fixed bulk delay and frequency selection technique are introduced to facilitate the power spectral density estimation for robust transfer function reconstruction. A multivariate analysis based on selected features extracted from various frequency bands is conducted on the signals recorded by multiple sensor pairs respectively. Furthermore, damage index traces based on data from different sensor pairs provide system redundancy for improved reliability with the voting logic for damage detection. The proposed approach lends itself to extremely high testing speeds (as fast as the service train speed, e.g. 60 mph and above), that would enable the real-time evaluation of rail track integrity at train operational speeds. A prototype based on this passive-only inspection idea has been constructed and tested with the DOTX216 testing vehicle of the Federal Railroad Administration at the Transportation Technology Center (TTC) in Pueblo, CO in September 2016. Test runs were made at various speeds from 25 mph to 80 mph (the maximum speed allowed on the test track). The results show the feasibility of stable reconstruction of the transfer function from the random wheel excitation, as well as the detection of joints and welds present in the track. Some tests were also conducted on TTC Defect Farm showing the potential for defect defection.