An unbalanced data-oriented class-aware representation learning method for unmanned aerial vehicle fault diagnosis

By constructing a class-aware adaptive representation learning model and utilizing data augmentation and feature fusion techniques, the problem of insufficient feature discrimination in imbalanced data was solved, thereby improving the accuracy of UAV fault diagnosis.

CN122087551BActive Publication Date: 2026-07-07GUIZHOU UNIV
View PDF 3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
GUIZHOU UNIV
Filing Date
2026-04-27
Publication Date
2026-07-07

AI Technical Summary

Technical Problem

Existing technologies cannot fully exploit the class-aware features and temporal differential information of fault samples, resulting in insufficient feature discrimination for imbalanced data, large classification bias, and reduced accuracy of UAV fault diagnosis.

Method used

A class-aware adaptive representation learning model is constructed. Data augmentation is performed through a data generation module, and original temporal and differential temporal features are extracted through a feature extraction module. These features are then weighted and fused through an adaptive gating mechanism. The model parameters are optimized by combining contrastive learning and classification loss to improve feature discrimination and classification performance.

Benefits of technology

It effectively improves the feature discrimination and classification accuracy in imbalanced data, reduces classification bias caused by insufficient feature mining, and significantly improves the accuracy of UAV fault diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122087551B_ABST
    Figure CN122087551B_ABST
Patent Text Reader

Abstract

The application relates to the technical field of unmanned aerial vehicle fault diagnosis, in particular to a class perception representation learning unmanned aerial vehicle fault diagnosis method for unbalanced data, which comprises the following steps: constructing a class perception adaptive representation learning model; acquiring original time sequence signal data of an unmanned aerial vehicle, and performing data enhancement on the original time sequence signal data of the unmanned aerial vehicle; performing feature extraction on the time sequence signal data after data enhancement; performing adaptive adjustment on differential time sequence features according to an adaptive gating mechanism; optimizing contrast feature representation through a class weighting contrast supervision learning mechanism, displaying inter-class regularization constraints and class center ordered arrangement set constraints; performing classification mapping on fused time sequence features; training model parameters of the class perception adaptive representation learning model, and analyzing original time sequence signal data of a to-be-tested unmanned aerial vehicle. The application can fully mine class perception features and time sequence differential information of fault samples, and improves the precision of unmanned aerial vehicle fault diagnosis.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Unmanned aerial vehicle control surface intelligent fault diagnosis method and device under unbalanced small sample

    CN117171681A

  • Unmanned aerial vehicle fault diagnosis method and system fusing random convolution data enhancement and deep learning

    CN119577439A

  • Bearing unbalance fault diagnosis method and system cooperatively guided by double classifiers

    CN119827155A