that all of the sensors essentially picked up the wheel fl at signal simultaneously. Th is realization lead to a soft ware design that sums and averages the signals from each sensor. In this operation, random background noise tends to be self-canceling while the repeated signals accumulate. Th e result is a dramatic, low-cost increase in the signal to noise. But more was needed. Th ey had to learn the language of wheelstrains. To do that required the application of cepstrum cepstral analysis. Cepstrum cepstral analysis is one of the foundational building blocks of the language recognition soft ware now ubiquitous in cellphones. By applying the technology of language recognition to the sound generated by passing trains, the system learned to identify the whisper of fl at spots that were too small to be heard by the human ear at train speeds as low as 8 mph but which, upon subsequent inspection, were found to be there. Th is capacity is enhanced by the use of cross-statistical analysis using data from multiple sources to uncover conditions which could not be discerned from a single-source data analysis. By concentrating the system’s development work on the signal processing component as opposed to the hardware, the IEM system wound up generating the benefi ts of lower costs, increased eff ectiveness, extended range and improved reliability. In daily operations, the wheel fl at detection system automatically alerts train controllers when it detects a fl at wheel. Th en, depending on the severity of the fl at spot, the train controllers can order the train out of service before it enters the critical zone, or can issue a go-slow order to the train reducing the noise generated by the fl at spot. Calibration is currently taking place to develop algorithms relating impact severity with train By the Numbers 8 MPH the speed at which the sound detection software can spot a flat spot 2.5 years of research and develop to create a whisper quiet system 24/7/365 Sound Transit is on the go FEBRUARY 2017 | MassTransitmag.com | Mass Transit | 21 light rail vehicles, especially on wet rail very oft en result in the development of “slid fl at spots” on the wheel. When the wheel slides, a fl at spot develops on the wheel and then each time that wheel revolves, that fl at spot acts as a jack hammer, pounding the rail, damaging the suspension equipment and sending out both audible and inaudible signals through the air and ground. Without early intervention, fl at spots tend to grow each time the train stops. Since there is no proven technology to avoid the development of fl at spots, the only option is to detect them and repair them by truing the wheel. Since fl at spots are the most signifi cant factor that could result in Sound Transit becoming out of compliance with the terms of its contract with the university, the fi rst order of business was to create a fl at spot detector which could provide maintenance managers and train operators a head’s up prior to a wheel fl at reaching or even approaching a critical condition. Th e wheel fl at systems are located miles from the UW campus, providing Sound Transit an early warning capability to take action before the prior to the train reaching any sensitive buildings. Learning the Language of Trains Wheels To meet the performance specifi cations required by the Sound Transit application, IEM developed signal processing algorithms which resulted in higher sensitivity and responsiveness than any other technology available. Th is increased sensitivity creates the ability to reduce the number of sensors which reduces cost while enhancing its ability to detect smaller signals which both allows the detection of smaller fl at spots and the reliable detection of fl at spots at slower train speeds. Th e fi rst step was to realize that the signal created by a fl at spot propagates through the rail so rapidly (compared with the forward motion of the wheels) THE VIBRATION monitoring system. Sound Transit Sound Transit
MassTransit_February_2017
To see the actual publication please follow the link above