Characterization of train fleet wheel condition in a metro

Abstract

The wheel condition of a metro fleet (>100 trains) is characterized based on train passby noise and wheel maintenance records over a four-year period. The initial objective was to identify wheel flats from noise measurements. The base unit of measurement is the maximum value of LAeq,1s from individual train passbys. These quarterly operation data are processed and ranked on a normalized scale, to provide an indication of the fleet performance. The change in the distribution of the train fleet wheel condition, on the normalized scale, is charted over time with the use of histograms. Further analyses are conducted, with the dataset sub-divided into two parts: the outliers (noisy trains) and the remaining data points. The outlier information is reviewed with wheel maintenance records, in an attempt to generalize descriptors like wheel defect severity, gestation period and location. The remaining data points are used to characterize the performance of each train type within the fleet. A relative performance comparison between the four train types in the fleet is attempted. The noise-based analyses indicate that there is a differentiation in terms of likelihood of wheel defect formation for a given time period and train passby noise emission. Using actual noise measurements and the derived characteristics of each train type, train traffic scenarios are simulated and compared to published National Environment Agency (NEA) guideline. The same baseline information is also used to simulate wheel maintenance optimization, with the objective of not exceeding the NEA noise emission guideline. It incorporates the increased train traffic expected with the new train signaling system. The paper is concluded with discussions on potential maintenance optimization and expected conditions with the increased train traffic.