Traction converter fault diagnosis method based on gradient improvement decision tree

Abstract

The invention relates to the field of rail transit, and particularly discloses a traction converter fault diagnosis method based on a gradient improvement decision tree. The method comprises the following steps: carrying out recursive dichotomy on a traction converter fault data sample, and constructing a fault classification model in advance by adopting a gradient improvement decision tree learning model; and fault data of the traction converter are obtained and are substituted into the classification model for fault diagnosis. The invention provides a recursive dichotomy fault diagnosis overall framework, and a traction converter fault diagnosis method based on wavelet packet decomposition and a gradient improvement decision tree is constructed in the overall framework. The fault diagnosis method is high in fault diagnosis precision, can completely meet the actual situations that few samples with positioned faults of the traction converter exist, the data dimension of the data sampleis high, distribution of different types of samples is unbalanced, and module faults are difficult to distinguish, and has wide popularization value.

Description

technical field [0001] The invention relates to the field of rail transit, and more specifically, to a fault diagnosis method for a traction converter based on a gradient lifting decision tree. Background technique [0002] The converter is a key component of the traction system of an electric locomotive. A fault in the converter will lead to paralysis of the train operation, and in severe cases, it will affect the entire railway line. Therefore, once the converter fails, it is necessary to locate the cause of the fault immediately and take relevant measures to solve the fault. [0003] For example, the four-quadrant input overcurrent of the converter is a common fault cause in the fault of the converter. The fault causes of the four-quadrant input overcurrent of the converter include abnormal input, abnormal components, abnormal detection, and other abnormalities. , and abnormal network voltage fluctuations, reverse connection of synchronous signals, reverse connection of ...

AI Technical Summary

Problems solved by technology

[0017] In order to overcome at least one of the defects described in the above-mentioned prior art, the present invention provides a method that can finely distinguish module faults, and can solve the problem of difficult fault diagnosis caused by few data samples, high data sample dimensions, and unbalanced sample distribution. Traction Converter Fault Diagnosis Method Based on Gradient Boosting Decision Tree

Images