Department of Civil & Environmental Engineering, School of Engineering, Rutgers, The State University of New Jersey, 96 Frelinghuysen Road, Room 607, Piscataway, NJ 08854, USA
Dr. Xiang Liu
Department of Civil & Environmental Engineering, School of Engineering, Rutgers, The State University of New Jersey, 96 Frelinghuysen Road, Room 607, Piscataway, NJ 08854, USA
Safety is the top priority for every rail system in the world. A widely used measure for rail safety is the accident rate, which is the number of train accidents normalized by traffic exposure. Of interest in rail safety research is understanding the temporal trend of accident rates, the significant factors affecting the trend, and how to predict accident rates. This paper uses a negative binomial regression model to present a statistical analysis of U.S. Class I railroad freight train derailment rates on main tracks by year and accident cause for 2000 to 2012. The accident and traffic data used in the analysis come from FRA. The analysis led to several observations. There is a significant temporal decline in freight train derailment rate (-5.9% per year). The rate of change in accident rate varied by accident cause. Rates of freight train derailment caused by broken rails or welds and track geometry defects declined by 6% and 5% annually, respectively; the rate of derailment caused by bearing failure decreased by 11% annually; and rate of derailment caused by train handling errors fell by 7% annually. The regression model is used to project train derailment rates by accident causes and can be used to evaluate the safety benefit of potential accident prevention strategies. This research provides policy makers and practitioners with a statistical method for analyzing the temporal trend of train accident rate for development of rail safety policy and practice.
26 July 2017 in Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit
In a railroad track, the inherent and small geometrical deviations in the position of rails from their ideal design states constitute imperfections that can have a significant impact on the safety and the rate of degradation of the rail system. These deviations are measured by various technologies and further assessed using various algorithms and statistical techniques to quantify the condition of the system. This paper reviews the existing research regarding the collection of track geometry data, analysis of degradation, and the associated safety and maintenance decisions. The knowledge gaps in the existing literature are identified and possible future research directions are suggested. The review can be used as a reference by practitioners and researchers to determine optimal practices for assuring the safety of tracks.
31 December 2022 in Accident Analysis & Prevention
Since the GPS is unavailable in underground environment, it is extremely challenging to measure the speed and location of a metro train. This paper proposes a novel data-fusion approach for speed estimation and location calibration of a metro train in underground environment, simply using the data from the 3-axis accelerometers in smartphones. Firstly, we place multiple smartphones in different cars of a train to measure the longitudinal, lateral and vertical accelerations, then propose a method to transform the measured accelerations from the coordinate systems of smartphones to that of the metro train. In the data fusion model, the initial estimations of train speed and position are obtained by the integral and double integral of the longitudinal accelerations. The lateral and vertical accelerations are used to provide absolute reference for speed estimation, where the local time delay and waveform similarity between the measured accelerations in different smartphones are defined and estimated to obtain the time-delay-based speed. Finally, a more accurate estimation of train speed is obtained by fusing the integral-based speed and the time-delay-based speed. A case study is conducted on Chengdu Metro Line 7 in Chengdu, China. The results show that, taking the interval length between adjacent stations as ground truth, our data-fusion approach achieves higher accuracy than the direct integral method, with the relative errors reduced from 9.5% to 1.6%.
13 July 2022 in Construction and Building Materials