City Bus Reliability Measurement Based on Sparse Field Data Supported by Selected State Space Models

Abstract

Means of transport are an important part of today’s cities. Bus transport in particular is considered to be a reliable mode of transport. In cooperation with a city’s transport company, we process in this article data collected from two fleets of buses. The data records are related to the failures of individual bus subsystems. We focus on the study of data from engine and brake subsystems, the consequences of failures of which are the most serious in relation to traffic safety. The data are seemingly austere, as the records only contain information such as “operating/fault” during a given month (no known causes, mechanisms, or other more precise time information about the failure). On the basis of such sparse data, however, it is still possible to estimate the trend or predict the development of certain measures over time. For the study and subsequent prediction, we used approaches based on state space models. Specifically, we worked with a linear trend model and a periodic component model. For both fleets of buses, we have also analyzed what the respective model and its prediction could look like if we knew selected and more detailed time information about the failures. This model therefore provides a general idea of the rate of occurrence of failure trend development, expected number of failures within single months, and respective bus subsystem failure occurrence forecasts. Based on this information, operators and entrepreneurs can rationalize the processes related to operations, maintenance, and repair planning.