Failure history analysis using multidimensional scaling and neural networks in railway systems

Abstract

Urban mobility is one of the main problems faced by big cities. The electric multiple unit (EMU) is the best alternative since it transports a high volume of people at a low cost. Failures during operation cause delays and inconvenience for passengers and operators. This article deals with failures in DC traction motors and their control systems in a transport company operating urban trains. The study focuses on identifying the main causes and consequences of these failures through a statistical analysis. Initially, the study proposes multidimensional scaling to analyze these observations and concludes with the use of neural networks. The acquired results are useful for actions in traction motors during preventive maintenance and in corrective maintenance.