Design method of tiny fault estimation system used for CRH2 high-speed railway inverter

Abstract

The invention discloses a design method of a tiny fault estimation system used for a CRH2 high-speed railway inverter, and belongs to the field of fault diagnosis and fault estimation. The method specifically comprises the following steps of in a first stage, building an offline data model: 1) collecting steady state running data of a high-speed train and preprocessing offline data; 2) calculatinga score matrix of a rotation subspace where the preprocessed data is located; and 3) determining a nonlinear relationship between a fault estimation value and a fault magnitude, thereby performing online estimation; and in a second stage, performing online fault estimation: 4) processing online data; 5) calculating online score vectors; and 6) estimating the existent fault magnitude by utilizinga fault estimator. The invention designs a multivariate statistical analysis-based estimation method. The method combines Kullback Leibler divergence and a neural network approximation method, can realize relatively accurate fault estimation, and has a certain advantage in sensitivity to tiny fault estimation.

Description

technical field [0001] The invention relates to a design method of a micro-fault estimation system for a CRH2 type high-speed rail inverter, belonging to the field of fault diagnosis and fault estimation. Background technique [0002] Due to its high efficiency, high-speed railway has developed rapidly in the past two decades. As the executive mechanism of the motor, the inverter is one of the important equipment of the high-speed rail EMU. Generally, the control law is formed by collecting the information of its sensors. Typically, since VFDs have a direct interaction with the working environment, their sensors are susceptible to faults, which can lead to abnormal operation of the motor, loss of traction and, in severe cases, even emergency braking. Therefore, it is necessary to realize real-time detection and fault estimation of sensor faults to improve the reliability of high-speed railways. [0003] In recent decades, there have been many studies on fault detection and...

AI Technical Summary

Problems solved by technology

However, due to the complex operating environment and the influence of unknown noise, it cannot be represented by an exact function

Furthermore, expert knowledge about high-speed trains is not fully accessible

Knowledge-based fault diagnosis schemes are often not effectively used when minor faults occur

Images