Exploring Acceleration Data Resulting from Passenger Discomfort in a Railway Vehicle by Using Machine Learning

Rail transport continues to be a popular public choice for business and personal travel. Passenger comfort is, however, an important factor that will influence an individual’s choice to use the railway as a frequent mode of transport. Passenger comfort is affected by many factors, including temperature, noise, and vibration. Vibrations of a railway vehicle experienced by passengers play an important role in evaluating and measuring the comfort level. The railway system, including vehicles (rolling stock) and infrastructure (track), is made up of key components which can affect the ride quality. Passenger discomfort is a challenging factor to overcome as it can have multiple root causes. Identifying the root cause of car-body vibrations, usually resulting in poor ride and passenger discomfort, is a challenging task. Indeed, it might be related to various possible rolling stock and track component degradation conditions. Despite this, the literature seldom links the two into a cause/effect tree. To build this cause/effect tree, key components related to rolling stock and track affecting passenger comfort are selected and used for scenario-based simulations, and passenger comfort is evaluated following the EN12299 standard. Acceleration data from these scenarios, including track irregularities and wheel flats used as inputs to railway vehicle dynamic simulations, are collected for a machine learning model to analyse and differentiate between the two condition signals. Through specific signatures of these degraded conditions, the proposed model helps identify which degraded components are most likely to cause passenger discomfort.