Failure Recognition for Switch Machines Based on Machine Learning

Abstract

With the faster and faster development of urban rail transit, the scheduling of the operation has become increasingly tight. In light of this background, higher demands on the reliability and maintainability of metro signaling equipment were proposed. At present, the switch machine, which contributes the highest failure rate in the operation of urban rail transit lines, has caught the attention of the maintenance companies, since its failure and disrepair may directly affect the punctuality and the occurrence of accidents. When the switch malfunctioned or operated abnormally, some differences will be revealed on the curve. Consequently, important evidence for the normal operation of the switch machine is whether the characteristic curve of the switch is displayed normally. On the basis of understanding the characteristics of common failures, analyzing the characteristic curve for normal operations of the switch machine can contribute to the proactive determination of whether there might be an impending fault with the machine; or the quicker localization and diagnosis of the cause after the failure.