Signals are an important part of the urban rail transit system. Signals being in functioning condition is key to rail transit safety. Predicting rail transit signal failures ahead of time has significant benefits with regard to operating safety and efficiency. This paper proposes a machine learning method for predicting urban rail transit signal failures 1 month in advance, based on records of past failures and maintenance events. Because signal failure is a relatively rare event, imbalanced data mining techniques are used to address its prediction. A case study based on data provided by a major rail transit agency in the United States is developed to illustrate the application of the proposed machine learning method. The results show that our model can be used to identify approximately one-third of signal failures 1 month ahead of time by focusing on 10% of locations on the network. This method can be used by rail transit agencies as a risk screening and ranking tool to identify high-risk hot spots for prioritized inspection and maintenance, given limited resources.
Safety is always a top priority during the development of high-speed rail systems. Wheel, as one of the most important components of a high-speed train, requires frequent inspections and maintenances in order to prevent operation failure of the train or even catastrophes from happening. The Residual Useful Life (RUL) has been extensively studied in the research field of health management for industrial components or systems. It is critical to understand the RUL of the high-speed train wheel for informed maintenance and replacement. Given an accurate estimation of train wheel RUL, rail companies can optimize maintenance and repair schedules more economically while ensuring safety. In this paper, we develop a machine learning based methodology to estimate the RUL of high-speed train wheels, using actual, in-field vibration data from sensors mounted on bogies. To our best knowledge, few studies have used data from in-field sensors mounted directly on train components to train and test a model for train wheel RUL estimation. A traditional time-domain signal processing method is implemented to extract characteristic features from the vibration data. Various machine learning models are introduced and applied to validate our proposed method. The estimation results conform to the empirical data, and can be used to infer the RUL of the wheel based on vehicle-mounted vibration sensor data for rail safety management.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.