Superconducting magnet quench detection method based on feature fusion hierarchical normal model

CN116087844BActive Publication Date: 2026-06-19HIWING TECH ACAD OF CASIC

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Patents(China)
Current Assignee / Owner: HIWING TECH ACAD OF CASIC
Filing Date: 2021-11-04
Publication Date: 2026-06-19

AI Technical Summary

Technical Problem

Existing methods for detecting quench in superconducting magnets have poor noise resistance, the threshold selection is affected by noise disturbances, and the generalization ability of single-parameter analysis methods is limited, resulting in decreased detection accuracy and delayed detection results.

Method used

A feature fusion hierarchical normal model is adopted. By preprocessing the voltage time series signal and magnetic field strength time series signal of the superconducting magnet, the low-level features of the mean field strength, voltage zero crossing rate and Mel cepstral coefficient are extracted. The feature layer normal sub-model and decision layer normal sub-model of the hierarchical normal model are constructed to determine the threshold range of the cumulative distribution function value of quench.

Benefits of technology

It effectively reduces the false detection rate and false negative rate caused by parameter fluctuations, improves detection accuracy and real-time performance, and avoids the problems of decreased detection accuracy and delay in single-parameter feature combined with data analysis methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN116087844B_ABST

Patent Text Reader

Abstract

This invention relates to the field of superconducting magnet technology and discloses a method for detecting quenching in superconducting magnets based on a feature fusion hierarchical normal model. The method includes: preprocessing the voltage and magnetic field strength time-series signals of the superconducting magnet to obtain multiple signal frames of the voltage and magnetic field strength signals; extracting features from the multiple signal frames of the voltage and magnetic field strength signals and fusing the extracted features to obtain fused features; constructing a feature-layer normal sub-model of a hierarchical normal model based on the fused features; obtaining the mean of the normalized probability density function of the samples based on the feature-layer normal sub-model; constructing a decision-layer normal sub-model of the hierarchical normal model based on the mean of the normalized probability density function of the samples; constructing a decision-layer normal sub-model of the hierarchical normal model based on the mean of the normalized probability density function of the samples; and determining whether the superconducting magnet under test has lost quenching based on a determined threshold range.

Need to check novelty before this filing date? Find Prior Art