Identification of ballast condition using SmartRock and pattern recognition

Abstract

In railroad, unfavorable ballast performances (e.g., ballast fouling, loss of lateral confinement) can lead to deterioration of upper structures such as the rail and tie. Therefore, accurate and timely monitoring of ballast condition is critical for rail safety operation and effective maintenance. In this paper, a series of ballast box tests were conducted to investigate the ballast particle movement pattern inside railway ballast under different ballast, loading, moisture, and shoulder confinement conditions. Eight wireless embedded devices – “SmartRocks” – were used in the laboratory tests in three different locations to study different ballast movement patterns under different conditions. A statistical Autoregressive (AR) model with X-bar control chart method was used to identify changes in particle movement patterns under different conditions. The results show that 1) the ballast particle movements are much more sensitive to moisture content for the fouled ballast than for the clean ballast; and 2) the AR model is capable of identifying ballast fouling and shoulder instability. In addition, a threshold value of 20% for the percentage of outliers of ballast particle movement patterns is suggested for the test conditions considered in this study. This study represents a preliminary step towards developing a reliable ballast condition identification index and further field studies are needed.