A high-speed rail turnout intelligent fault prediction method based on hybrid deep learning

Abstract

The invention provides a high-speed rail turnout intelligent fault prediction method based on hybrid deep learning. Characteristics of turnout current curve data are automatically extracted by using anoise reduction auto-encoder; the unlabeled feature data is clustered based on a density clustering method; clustering clusters under different turnout fault standards are selected as normal data bycombining expert knowledge; finally, normal data under different turnout fault standards are used for training single classification support vector machines respectively, output results of the two single classification support vector machines are crossed, a turnout fault detection and prediction four-quadrant model is formed, and fault detection and prediction are conducted on to-be-detected data;The method and the system have the advantages that the characteristics of the turnout action current curve are efficiently and objectively extracted; The workload of manual one-by-one data labeling is reduced, and the method has the advantages of obtaining normal data reliably and efficiently; And turnout fault detection and prediction can be carried out only by inputting normal data under different fault standards.

Description

technical field [0001] The invention belongs to the field of turnout fault prediction, in particular to an intelligent fault prediction method for high-speed rail turnouts based on hybrid deep learning. Background technique [0002] As the key ground signal equipment of the railway, the turnout is in direct contact with the high-speed train, and is used to realize the transfer or cross-line operation of the train moving forward. The quality of its working status will directly determine the safety of the high-speed rail operation. At present, the operation status of my country's high-speed railway turnouts is characterized by a large number, frequent operation, harsh environment, and large load impact. [0003] At present, the traditional turnout maintenance method is mainly based on manual experience. The staff compares the turnout action current curve monitored by the computer monitoring method with the current curve obtained by the expert summary, so as to determine the wo...

AI Technical Summary

Problems solved by technology

At the same time, the existing fault detection methods based on analytical models, expert methods and support vector machines, because the feature extraction of turnout operating current curve data in the early stage depends on manual experience, there are certain uncertainties and low efficiency, so there is an urgent need for a A Highly Intelligent Method for Predicting Turnout Faults

Images